Home

Entwicklung und Umsetzung von Strategien zur Qualitätssicherung

1. V250 SENSOR APV Rinv 100Q F Vimp Istrip Rpinhole V125 9 ER iR x A LA dd Ce yi E N this point e E e N L a common for det Rpoly 1 8 MQ e ail L a all 128 chans bias FR Y O Y N DCU VSS common for all channels Rret 22kQ base line restorer VSS FEH Figure 5 15 Pinholes and the APV25 circuit Raymond 2001a group for example has reported a characteristic increase of sensors leakage current Ileak with respect to the number of pinholes Coldewey 2000 This implies that a pinhole strip will probably drive a higher current Igrip than a regular strip Unfortunately a strip current I strip of about 0 5 4A would be sufficient to produce a potential connected to the APV25s input channel equal to the virtual ground In this configuration the pinhole will become invisible 5 5 3 6 Bad Poly Resistors Typical bad poly resistors have a too small or infinite value With a significantly smaller value for Rpyoiy the p implant strip gets sensitive to noise on the bias line mainly expected to be induced by CM from the other strips The bias resistance noise scales with 1 Rpoly and thus the noise increases for bad poly resistors but its contribution stays small as long as the poly resistor is not completely shorted see Tab 4 5 on page 44 Because the field configuration does not alter the signals within the silicon are collected properly Therefore not deterior
2. example fluctuates due to electron emission statistics This shot noise is represented by a current noise generator in parallel with the detector Due to thermal velocity fluctuations of the charge carriers resistors exhibit thermal noise This kind of noise can be modelled either as a voltage source or current generator Generally resistors shunting the input act as noise current sources and resistors in series with the input act as noise voltage sources why some people in the detector community refer to current and voltage noise as parallel and series noise Since the bias resistor effectively shunts the input as the capacitor Cy passes current fluctuations to ground it acts as a current generator inb and its noise current has the same effect as the shot noise current from the detector Any other shunt resistances can be incorporated in the same way Conversely the series resistor R acts as a voltage generator The electronic noise of the amplifier is described fully by a combination of voltage and current sources at its input shown as ena and ina Johns and Marin 1997 Thermal and shot noise have a white frequency distribution i e the spectral power densities dP df amp di df x de df are constant with the magnitudes E 4kT ia 2elieak i e2 4kTRs 4 9 Rpoly where e is the electronic charge Ijeq the detector bias current k th
3. Figure 7 1 Pedestal for ia the different readout modes Br we N me ee e ar de show only small differences 9 E E in shape In general a au a y pe Peak mode pedestal is about ge ja aaa a m ef i 20 ADC higher compared to A E je A Un u Deconvolution mode Inverter 180 fl yap E pul we N mode pedestals are further 1706 ty 3 more smaller on a scale of 160 E 5 ADC compared to the cor E g responding non inverter mode m P SAK wo Cal Wo Iny Mode zg The APV25 borders chan m 2 nn F nels are clearly visible in the 130E Dec wo Cal w Inv Mode en pedestals and in general the HA Tr be ET 400 500 3 pedestal increases with the DB channel number channel number which depends mainly in the shielding All these effects results in a decreased signal quality Therefore the cable between the FEH and the FED is kept as short as possible and differential signal transmission on a cable with low capacity and resistivity and good shielding is used The noise measured will always be the quadratic sum of all the independent noise sources and therefore it is impossible to disentangle the modules noise and the readout systems con tributions However the latter can be reduced to a level where its quadratic contribution can be more or less neglected Although the APV25 s internal CM suppression compare Sec 4 2 1 2 on page 30 results in a very stable behaviour a reduced noise level is preferable and re
4. 3 5 3 2 5 E 2 a Figure 7 47 APV25 edge 2 noise in Peak and Deconvo 2 es lution mode shows a signifi ES cant difference While in Peak 25 mode all edge channels of the APV25s have to be flagged noisy only the sensor edges are noisy in Deconvolution mode i i i i 100 200 300 400 500 DB channel number a N Peak wo Inv Mode wo Cal Module 30200020000503 7 7 6 Noisy Strips Channels with an increased noise that can not be related to one of the other strip defect types are flagged as noisy channels Beside bad APV25 preamplifiers this will include channels with strongly increased leakage currents or channel which are going into breakdown as well as most probably channels with bad poly resistors On the tested prototypes no candidate for this type of noisy channels was found An important group of noisy channels are the APV25 border channels which show in the Peak modes an increased noise see Fig 7 47 This border channel effect is enhanced for the first and the last channel of the module sensor edge effect Both effects were discovered during the first prototype testing in Karlsruhe Improvements of the test systems CM noise showed that the APV25 border effect scales at least partially with the CM noise Furthermore Fig 7 45 on the preceding page shows that the additional noise of the border channels is independent from the modules leakage current The exact mechanism of this APV25 boarder c
5. Ls on de 20 E S B L 33 C l 8 C a L 25 10 F ae io ee ES Fone Sartre ane RE EN FERNEN RN ERST DURFTE HUN 100 200 300 400 500 82 0 100 200 300 400 500 Bias voltage V Bias voltage V Figure 7 23 SNR in Peak mode as func Figure 7 24 SNR in Peak mode as func tion of bias voltage at room temperature tion of bias voltage at room temperature taken with cosmic rays on the FTS taken with a Sr source on the DTS 7 5 2 Signal to Noise Ratio vs Bias Voltage The SNR is expected to depend slightly on the bias voltage Below full depletion the SNR scales with the depth of the depletion zone which increases with the square root of the bias voltage compare Eq B 2 After full depletion the SNR still increases with the bias voltage because the increasing electrical field changes the charge collection time This effect is quite strong close above the depletion voltage where the field just reached the backplane Later the slopes decreases significantly but stays below saturation for the bias voltages applied see Eq B 4 Therefore the modules are run over depleted during qualification test to get the charge collection saturated Figures 7 24 and 7 25 on the next page shows the measured SNR vs voltage characteristic The change in the slope visible in Fig 7 23 occurs at a bias voltage which is about a factor of 1 3 larger than the sensors depletion voltage in good agreement with test beam experiences see e g Fri
6. Broken Strips 2 Missing Bonds Pinholes hon ds Bad Poly Resistors Noisy StriPs o o 7 7 1 7 7 2 7 7 3 7 7 4 7 7 9 7 7 6 7 7 7 7 8 High Leakage Current Behaviour of Pinholes 7 9 Scratches Conclusion Appendix Infrared LED signals General ASICs Faults Acronyms and Abbreviations Silicon strip detector characteristic B I Leakage current carrier velocities 2 222220 B II Depletion Voltage and effective Doping Concentration B II Charge Karlsruhe Readout libraries Channel numbering schemes I2C bus protocol Front End Hybrid schematic Fast Test Station schematic List of Figures List of Tab References les CONTENTS 123 128 oe Gis ne tn N HUNG 128 doy Wee tht Uh Sut Bl pe oh die 128 ginal weer eer a ue Spied 128 130 136 137 137 140 141 144 145 Zusammenfassung Zusammenfassung Entwicklung und Umsetzung von Strategien zur Qualitatssicherung von CMS Silizium Mikrostreifenspurdetektormodulen Le Physikvam GAGs cd Ai hah es feat aida I 2 Der CMS Detektor 134 4 64 2b been ee lesen deshalb I 3 Der Silizium Mikrostreifenspurdetektor 0 0 00 cece eee eee eee eee tenet nees I 4 Qualit tskontrolle f r CMS Spurdetektormodule 00 e eee eee eee ee eee V 5 Die Karlsruher Teststationen 0 ccc cece cece eee eee ene nee e ne en VI 6 Testsystem Charakterist
7. GND y vss Figure 4 12 Connec tions of the DCU on CMS end cap modules Fried 2001 Reset 2Cadd 12Cbus 4 2 2 Optical Links Optical data transmission is essential for the CMS tracking system to transfer data from and to the tracking volume with minimal contribution to the material budget Furthermore an optical system has the advantage of being inherent immune to electrical interference and allows galvanic decoupling of front and back ends Thus the data acquisition system of the CMS tracker will make use of optical fibre links for analogue readout of detector signals from the front end electronics 40000 fibers in total and for digital transmission 1000 fibers in total of trigger timing and control signal between the front end communication and control modules CCUMs and the back end FECs The basic technology used for the digital and for the analogue links is very similar edge emitting lasers operating at 1310nm epitaxial positive intrinsic negative PIN photodiodes and single mode optical fibres Vasey et al 1998 4 2 2 1 LLD The linear laser driver LLD array ASIC converts differential input voltage into a single ended output current added to a preset DC current This DC current allows correct biasing of the laser diode above its threshold in the linear region of its characteristic The absolute value of the bias current can be varied over a wide range 0 to 55 mA in order to maintain the correct funct
8. D P Roy Higgs and SUSY searches at LHC An overview eprint hep ph 0303106 2003 A Rubini Linux Device Drivers O Reilly 1998 W Schwerdtfeger Entwicklung von Systemkomponenten zur Qualit tssicherung von CMS Silizium Streifen Detektoren Diplomarbeit Universitat Karlsruhe TH Institut fiir Ex perimentelle Kernphysik May 2002 H J Simonis 2003 private communication H Spieler Signal Processing ICFA Instrumentation School Fuare South Africa March 2001 http www physics lbl gov spieler ICFA_School ICFA2001 pdf REFERENCES 149 B Surrow Automated silicon module assembly for the CMS silicon tracker Nucl Instrum Meth A461 251 252 2001 S M Sze Physics of Semiconductor Devices John Wiley amp Son September 1981 J Troska et al Prototype Analogue Optohybrids for the CMS Outer Barrel and Endcap Tracker NIM 2002 to be published Proceedings of IEEE Norfolk R Turchetta et al Design and results from the APV25 a Deep Submicron CMOS Front End Chip for the CMS Tracker Nucl Instrum Meth A466 359 365 2001 Universell Universe II TM User Manual http www tundra com 1998 F Vasey et al Development of radiation hard optical links for the CMS tracker at CERN IEEE Trans Nuc Science 45 3 331 337 1998 NSS 1997 Proceedings M Waldschmitt Fertigstellung einer Teststation zur Qualifikation von CMS Siliziumstreifen dektormodulen Diplomarbeit Universit t Karlsruhe TH Institut f
9. Sr source is given in Fig 7 18 The difference compared to cosmic ray results from the different geometrical acceptance For the source tests the scintillator are placed directly below the silicon which enables also the detection of tracks that are affected by multiple scattering while for cosmic rays tracks affected by multiple scattering are lost because they will never reach the second scintillator which is about a meter below the sensor Furthermore broader clusters have an increased probability of signal truncation caused by cluster algorithms three sigma cut on collected channels 7 4 3 Signal to Noise Ratio Fits The energy loss of a particle in a thin absorber are described by a Landau distribution as discussed in Sec 4 3 1 on page 38 While the discussion in Sec 4 3 1 deals with the idealis tic energy loss and its probability distribution a measurement is also affected by noise and resolution effects Therefore the measured energy loss distribution data is fitted by a Landau distribution convoluted with a narrow normal distribution representing electronics noise and intrinsic detector fluctuations This results in a minor broadening of the peak and a slight 100 7 4 Signal Performance Entries 744 a 140 Mean 38 19 x7 Indf 32 59 34 E Width 1 337 0 2022 u 120 MP 34 53 0 3701 Area 2509 95 61 100 GSigma 5 556 0 3512 80 Figure 7 19 SNR in Peak mode at a bias voltage of300 V at a tempe
10. Entries 7 Test System Performance and Module Qualification Studies 99 Entries 9407 Entries 744 a 4500 Mean 2 775 z Mean 2176 E RMS 0 8786 Ly RMS 1 381 i 4000 Fr E C 3500 H 3000 F 2500 E 2000 E de 1500 i 8 H F E 1000 3 6 E H 28 8 El ms a o H HE REFERRAL A ies 0 12 8 10 12 2 Cluster size Cluster size Figure 7 17 Cluster size of cosmic ray Figure 7 18 Cluster size of Sr measure measurement show an average spread of ment show an average spread of signals signals over two channels over three channels by taking the square root of the averaged squared noise n 2 S y 2 Netuster ai N 7 1 cluster 1 channels which gives noisy channels a bigger weight without getting dominated of the most noisy channel Furthermore the clustering algorithm rejects channels with a noise of 50 above modules mean noise which causes in Fig 7 15 on page 97 a small gap in the hit map between channel 255 and 258 for example The different cluster noise algorithms produce significant differences mainly for larger cluster sizes which are typical for gaseous detectors while this effect is very much reduced for silicon detectors where diffusion and charge sharing effects are smaller compared to gaseous detectors Figure 7 17 shows the cluster size distribution measured for cosmic rays on the DTS indicating an average cluster size of 2 2 strips The same plot for a
11. ing for the LHC clock and first A level trigger signals Placidi 1 t Li t et al 2000 L1 reject une This concept simply allows to reduce the number of physical channels necessary to transmit the clock and trigger information to the detector However at the receiving end a special purpose circuit is required to extract from the encoded signal the original information The tracker phase locked loop PLL ASIC implements this function it extracts the LHC clock from the encoded signal and decodes the first level trigger decisions An important function of any clock and trigger distribution network is the ability to correct the timing of the clock and trigger signals according to the geographical position of the different subsystems inside the detector This function is also implemented in the PLL The ASIC contains an internal 4 Silicon Strip Tracker and its End cap Modules 29 Mi Analog Output ml Figure 4 7 Block diagram of the APV25 Each preamplifier is connected to a 192 cells deep pipeline to store data during the time needed for trigger decision Access to the pipeline is controlled by a read write pointer logic saving cells awaiting for readout from being over written An APSP can be used to de convolute the shaping behaviour of the preamplifier and shaper stage before sending the data through a 128 1 multiplexer stage to the DAQ system Internal register access is performed using an I C interface Furthermore for funct
12. ration 1998 technology with improved signal to noise ratio SNR performance This enables usage of large area modules in the outer part of the tracker Modules with strip length of up to 20 6cm comparing to 12 3cm in the inner part produce 50 more noise This will be compensated by increasing the sensors thickness from 320 ym as used in the inner layers to 500 um in the outer ones resulting in the same SNR Pitches range from 80 um to 183 um in the barrel and 80 5 um to 205 um in the end caps The stereo layers are built out of two completely independent single sided detection units mounted back to back one of them precisely measures the main coordinate the other rotated by 100 mrad with respect to the first gives a sufficient measurement of the second coordinate The tracker has to operate in a high radiation environment for at least 10 years maintain ing a satisfactory global performance despite expected changes in the material characteristics due to irradiation The level of irradiation coming from primary interaction will be very high around the collision region in addition a high flux of backscattered neutrons evaporated from nuclear interactions in the material surrounding the tracker will be present The innermost layer of the SST will undergo a fluence of 1 6 x 1014 1 MeV equivalent n cm during 10 years of operation To avoid potentially critical degeneration of the sensor performance due to irradiation the SST will be ke
13. Cy 2Cint and the total detector capacitance as seen from the amplifier results in Ce C strip Cro E 4 13 Ce C strip Furthermore the fraction of charge collected is Ce 4 14 Ce T strip In the specifications of the CMS sensors we have Ce 1 2pF cm um with the strip length given in cm and the implant width in um Cstrip is expected not to exceed 1 3pF cm The ratio of pitch to implant width is for all CMS sensors chosen to be 0 25 compare Sec 4 1 2 on page 24 Together with the pitches shown in Tab 4 1 on page 22 this leads to typical Cot values in the range of 14 6 25 1pF and charge coupling efficiencies of 94 9 97 9 The overall noise bandwidth depends on the time constant i e the characteristic time Ts Noise with a 1 f spectrum depends only on the ratio of upper to lower cutoff frequencies differentiator to integrator time constants of the CR RC shaper so for a given shaper topology the 1 f contribution to Qn is independent of T The contribution of noise voltage sources to Qn increases with detector capacitance and decreases with increasing shaping time while the contribution from noise currents increases with shaping time Furthermore pulse shapers can be designed to reduce the effect of current noise which will also mitigate radiation damage 4kT Q eiea R FT AAT Re e FChe Ta Foy A C 4 15 POLY As the characteristic time T is changed the total noise
14. DB channel number and 70 which are pulsed by the same calibration unit Furthermore a strong cross talk from channel 62 to its shorted neighbour 63 is visible This cross talk is in fact a clear and unique tag for shorted channels For an ideal and low ohmic short of two channels the amplitude is expected to drop by 50 For inverter off modes the sum of the signal in shorted strips results quite well to the total signal driven by the calibration unit while inverter on modes typically show a reduced sum Shorts are also visible in the channels noise due to the increased load capacitance of the APV25s input channel While this holds true for Peak without Inverter mode where an increase by a factor of 2 is visible see Fig 7 37 this effect is reduced by the APSP filter of the Deconvolution mode resulting in an increase of less than 50 The inverter furthermore changes the noise behaviour of shorted channels drastically and a shorted strip signature is no longer present in the noise data lower plot of Fig 7 37 The loss in calibration amplitude as shown in Fig 7 35 is not a unique signature of a short although it is necessary while the charge sharing as presented in Fig 7 36 is a sufficient tag for the identification of shorted channels Table 7 3 on the next page summaries the signature of shorts in noise and calibration amplitude for the different readout modes and for different bias voltages including a forward bias scenario of 9 V
15. Due to the large fluence integrated over the ten years of operation radiation damages will strongly affect the silicon sensors characteristics There are two basic mechanisms of radiation damages in silicon detectors 1 Bulk damage due to displacement of atoms from their lattice positions This leads to increased leakage current charge carrier trapping and build up of space charge which changes the required operating voltage This kind of displacement damage depends on the nonionising energy loss NIEL and the energy imparted to the recoil atoms Hence it depends on both protons and neutrons as well as on energy Dierlamm 2003 2 Surface damage occurs due to charge build up in the oxide layer which leads to an in creased surface leakage current Furthermore the inter strip isolation is affected within strip detectors Since the damage is proportional to the absorbed energy when ionisa tion dominates it is independent of particle type The increase in reverse bias current due to bulk damage is given by AJ a per unit volume where is the particle fluence and a the damage coefficient a 3 x 10717 A cm 4 Silicon Strip Tracker and its End cap Modules 41 As the reverse bias current depends strongly on temperature cooling the sensors to 10 C results in a 15 fold reduction of the current Furthermore the bulk damage induced traps for charge carriers results in a decreased charge collection efficiency After irradiatio
16. Further more the support of different stations is realised by the call of different configuration files from a main configuration file 136 APPENDIX D CHANNEL NUMBERING SCHEMES D Channel numbering schemes Within the CMS collaboration two different numbering scheme are established The first one is called the database numbering scheme and is based the numbering of the APV25s which uses the 12C addresses of the APV25s as order Their corresponding addresses are 0x40 0742 0744 0146 0148 and 074A and they increase from left to right looking for the NAIS connector towards the FEH For modules with only 4 APV25s the two middle addresses are not used This numbering scheme is used within the readout software and the database The other one is the sensor numbering scheme which uses the channel number as they are printed on the silicon sensors These are always on the opposite side of the poly resistors which is for the end cap module the wider side of the wedge shaped sensors The FEHs are for some rings of the end caps on the wider side of the sensors and for the others on the smaller side As a result the channel numbers printed on the silicon sensors increase for some rings from left to right while the decrease of others always looking from the FEH towards the silicon The sensor numbering scheme is the natural one from the bonding operators point of view because these are printed on the sensors and visible for him while from the
17. Hi erbei ist es neben den normalen Analysen des Modulverhaltens auch m glich weitere Me gr en zu erhalten indem Probenadeln radioaktive Quellen oder Laser verwendet werden Das gro z gig gew hlte Volumen dieser Station erlaubt eine vielseitige Anwen dung und kurzfristige Modifikationen Beide Stationen verf gen ber Szintillatoren und Photomultiplier zur absoluten Eichung mit kosmischer H henstrahlung Das Auslesesystem besteht aus Komponenten welche innerhalb der CMS Kolabora tion entwickelt wurden industriell verf gbar oder selbst entwickelt sind Letztere um fassen wichtige Schl sselkomponenten wie die Takt und Triggererzeugung oder die Hoch und Niederspannungsversorgung Hierbei wurde f r die Eigenentwicklung ein modulares Konzept zu Grunde gelegt welches auf einer Hauptplatine verschiedene funktionale Ein heiten zusammenf hrt vgl Abb 5 Die Auslesesoftware basiert auf einem objektorientierten Ansatz welcher in C auf einer Linux Plattform realisiert wurde F r das Benutzerinterface und zur Steuerung wird LabView verwendet Auch bei der Softwareentwicklung wurde gro er Wert auf Mod ularit t gelegt und getrennte Funktionen in unterschiedlichen Bibliotheken abgelegt vgl Abb 6 6 Testsystem Charakteristika und Qualit tsstudien Eingehende Analysen von Detektormodulen setzen eine genaue Kenntnis von Testsys tem und Testobjekt voraus Hierbei werden insbesondere schon w hrend der Entwick lungsphase des
18. If a calibration pulse is given on one of the shorted strips the other will also share part of the signal which gives a clear tag for identification bias ring AC pad coda guard ring DC pad AL St pa SiO SI N A P implant A R n bulk_ gt por breaks a 7 n layer Al Figure 5 13 Schematic of break fault in silicon strip detectors 5 5 3 3 Broken Strips are expected to show the opposite effects concerning the capaci tive load than shorted strips the APV25 amplifier will face a reduced capacitive load If the readout strip is broken part of the capacitive load is lost and also part of the charge collected compare Eq 4 14 causing a SNR deterioration But furthermore due to the metal overhang technique used a broken metal readout strip will result in a disturbed electrical field The high edge of the electric field moves back from the much more resistant oxide layer to the silicon p implant increasing the risk of a local electrical breakdown As a result the corresponding strip leakage current strip increases similar to the effect of a surface scratch compare Sec 5 5 3 1 The signature of a broken readout strip will typically contain an increased strip leakage current The calibration amplitude is expected to be slightly larger due to the reduced capacitive load and signals will only cause a reduced amplitude 58 5 5 Module Error Type Detection 5 5 3 4 M
19. The total number of readout channels pixels is about 45 x 10 The pixel detectors uses a readout system with zero suppression to reduce the amount of data to be transfered At full LHC luminosity about 1000 tracks cross the pixel barrel layers every 25 ns which translates into an occupancy of 3 3 x 1074 or a single pixel counting rate of 10 kHz 3 3 2 Silicon Strip Tracker The SST surrounds the pixel system In the central pseudo rapidity region detectors are arranged in a barrel geometry with 10 layers while in the forward regions they are arranged in 9 disks radially segmented into the so called petals Furthermore 5 stereo layers are concentrated below a radius of 90cm The detector modules are assembled using single sided sensors with the strips parallel to the beam axis in the centre or barrel region and with the strips orthogonal to the beam nominally pointing to in the two end cap regions Figure 3 5 on the following page shows a longitudinal view of the all silicon CMS tracker The overall instrumented area of the silicon strip tracker will cover 210m The possibility to construct such a large silicon tracker relies crucially on a few key elements like the usage of 6 wafer technologies and implementation of the front end readout chip in deep sub micron 18 3 3 Central Tracker Figure 3 5 Strip Tracker layout Solid lines represent double sided modules CMS Collabo
20. ae behave like a single pinhole 3 70 NR EEE tae inte En en NE EEE ER shown in Fig 7 44 Their cal E N A A ee A ibration amplitude regenerates Ss for small artifical leakage cur E E rent and decrease again after 3 ME the full amplitude has been 30 33 reached But with a dozen pin 20 RE holes also the other channels i a E E of the APV25 are effected and E i 83 show an deterioration in their ee ee o e A 100 150 200 250 300 350 400 88 calibration amplitude of about Leakage current A 25 ADC The second ASIC fault found affects again a prototype FEH with six APV25s for which also two APV25s were added later In this case an inverter mode problem was found which is shown in Fig 7 49 Most probably this defect arises from a supply problem of the inverter GND The provided GND supply of the inverter is realised via an external 1002 resistor from VDD compare Jones 2001 This inverter fault looks like a bad solder joint of that external resistor Unfortunately this particular module was also only for a few days in our lab and a prove of this assumption was not possible in the given time 7 8 High Leakage Current Behaviour of Pinholes While most of the strip errors affect only one single channel respectively two or even more for shorts pinholes can affect a complete readout chip In Sec 7 7 4 on page 113 the effect of a single pinhole is discussed and Fig 7 44 on page 114 showed the unique behaviou
21. m a Aik en Ere Oh a te 56 5 5 3 2 Shorted Strips naar need ea ea Be sr 57 5 5 3 3 Broken Stips itin sn Seek eee a ea Pi ee ee ia 57 5 5 3 4 Missing Bonds aa au Ban Pi ae di 58 5 9 3 09 P nholess a een euer te 58 5 5 3 6 Bad Poly Resistors 22 2 om onen 59 5 5 3 4 Noisy Channels lisis ro a RE 59 5 6 Module Quality Grades e 59 CONTENTS 6 Karlsruhe Test Stations 6 1 Karlsruhe Readout System n nn 6 2 Hardware Components 6 2 1 Motherboard urn ent eo a ana le oe 6 2 2 PLD Sequencer mip ne oa hola ee ee ern den ania 6 2 3 ARES Repeater laicas Ber a ar a rd 0 24 LAW Card oe utes a ie de es ee ee ets ohn Bhe A etl et amp aA 6 2 5 Infrared LED System 2 200000 000 eee eee 6 2 6 Power Pack and Peltier Control 2 2 2 2 Cr o 6 2 7 Multiplexer Device 2 2 20 2 0002 ee ee ee 6 22 38 FEED A aa ed oy Shoe Oe area 6 2 9 BE the oo en Re A AE ORAS BEG ES ES 6 251 OF 0 II OR 5 2 ra es en daha ta teh en ne en ae 6 21 MIO 2 53 3 2 Bak AA AE Be eB Bh le des 6 2 12 Slow Control Multiplexer 2 2 2 Emo 6 3 Software Layout te dr hee OA oe EER Ne tO be ee oe E 6341 DEVICE Driver ana a Pk Re es en Ba PODEVer e E RA BE ae BY 6 3 1 2 PED Driver 2 2 ae Dalai ri 6 35 27 Libraries na ea eee ee An WY EP Be Poon Eee 6 3 3 Threads ii ee En ale mu 6 3 4 Graphical User Interface 2 2 22 2 on on 6 4 East Test Station m i suit Shan an ran eee BAD ic
22. system must unambiguously identify the LHC bunch crossing with high efficiency The RPC detectors are employed as dedicated fast trigger detectors and complement the triggering capabilities of the DT and CSC detectors The muon detection system has been designed with a certain degree of redundancy in order to properly reconstruct muon tracks even in the presence of several superimposed events Giacomelli 2002 22 4 Silicon Strip Tracker and its End cap Modules The end cap modules of the Silicon Strip Tracker SST are discussion in detail and their properties and electrical behaviour is presented Starting with an overview of the different geometries and their multiplicity a short description of the modules main components is given followed by a more detailed discussion of the SST readout and control chain Hereby main attention is paid on the FEH and the application specific integrated circuits ASICs implemented on it Especially the front end amplifier chip characteristics are discussed in greater detail Finally the expected module performance is analysed by means of signal creation and detection A detailed noise analysis showing the strength of the noise contributions arising from sensors and electronics allows a precise calculation of expected signal to noise ratio SNR 4 1 Silicon Strip Tracker Modules The micro strip part of the central tracker in CMS will cover 210m and consist of 15 232 individual modules The total number
23. to bias resistance Rpoiy parallel DE channe number l noise see Sec 4 3 3 Peak wo Inv Mode w Cal 30216630200027 Noise ADC T Figure 7 30 The noise pro file of all 512 channels shown in Fig 7 29 as leakage current function gives access to the slope which is within a rea sonable range consistent with the expectation from the noise analysis see Sec 4 3 3 Al ne aa 0 200 300 400 500 600 700 800 900 Leakage current pA Peak w Cal wo Inv Mode Module 30216630200027 time of a 500 um thick silicon strip sensor Lutz 1999 This enables a pulsed usage of these LEDs which provides information on individual channel response independent and redundant from the APV25s internal calibration circuit Furthermore creating external signal in the silicon gives information about the signals coupling to the readout strip and its connection to the APV25 which is essentially useful for missing bond detection Figure 7 31 on the next page shows the signal of a single LED A collimator restricts the light cone emitted by the LED resulting in a spatial width of the signal of 15 channels at FWHM Figure 7 32 on the following page shows the envelope of all LEDs which shows only minor differences between the light cones from the LEDs and their overlap regions For comparison of data from different LED arrays like it is needed for automatic detection of missing bond positions the LED signals have to be norm
24. x 10 3mol7 g mol 0 510 998 902 21 MeV 2 817 940 285 31 fm 1 380 658 x 10723J K 8 617385 x 10 eV L MeV MeV lim n oo mn 0 577215 k g cm 0 0265X0 Z 1 2 pA mA nA uA MQ GQ GQ 22k typical 400 V 20 pF Cy 2Cint Table A 2 Symbols used within this work 128 APPENDIX B SILICON STRIP DETECTOR CHARACTERISTIC B Silicon strip detector characteristic A detailed discussion of silicon detector characteristics is surely out of scope of this work Nevertheless two of them will be stated as they have a direct connection to measurement performed on the module level For a more detail discussion of silicon detector characteristics see Lutz 1999 Sze 1981 or Moll 1999 B I Leakage current The leakage current of a reverse biased junction not close to breakdown is approximately given by gAjni Iren Td B 1 27 with the junction area A the silicons intrinsic concentration of charge carriers n the effective charge carrier lifetime To and the depletion regions thickness xg The last depends on the bias voltage applied 2K Tda 2 00 Voias B 2 Y aNa and the intrinsic charge carrier concentration n is given by Ni Z y Ng Nye Es kT B 3 with the conductive and valence band state densities Nc and Ny and the band gap E between them Since the intrinsic charge carrier concentration n is a strong function of temperature and double for every te
25. 6 5 Diagnostic Test Station nn 6 5 1 Linear Gate System 2 on n nn 6 5 1 1 An Optical Microscope 2 2 2m nn 6 5 1 2 The Laser System 2 22 22 2 CE onen 6 5 1 3 A Radioactive Source 2 2 2 En nn Test System Performance and Module Qualification Studies 7 1 Readout Modes is eme ke as ae as le hae aS pre 1 2 N ise Studies u amp Bra a an ae Be re 7 2 1 Pedestal and Raw Noise 2 2 2 CC nn nn nn 7 2 2 Common Mode Noise 2 2 2 2 aa 7 2 3 Readout Mode Dependence of Module Noise 7 3 Module Leakage Current 222222 Coon tA Signal Performance u u a ee a re 7 4 1 Calibration Signal 2 2 on nn 7 4 2 Cosmic Ray and Radioactive Source Signals Detection 7 4 2 1 Angular Acceptance 2 2 2 2 non nn 7 4 2 2 Timing Jitter 2 2222 Como nn 7 4 2 3 Clustering Algorithm 7 4 3 Signal to Noise Ratio Fits 2 0 0 00 ee ee 7 5 Signal to Noise Ratio Studies 2 ooa aa es 7 5 1 Signal to Noise Ratio vs Temperature 2 7 5 2 Signal to Noise Ratio vs Bias Voltage o 7 5 3 Signal to Noise Ratio vs Leakage Current ii 61 61 62 63 64 65 65 67 69 70 70 73 73 73 73 74 74 76 76 77 77 77 80 83 84 84 85 85 iv Y Qo 3 amp Oa 7 6 Infrared LED Studies High Leakage Current Behaviour of Modules 7 6 1 7 6 2 7 7 Module Fault Detection Studies Shorted Strips
26. APV25 Pinhole rate to be connected to more than three pinholes number of module for one sensor modules for two sensors modules affected 0 30 0 064 1760 0 78 670 16 0 2420 0 20 0 014 385 0 18 150 3 5 535 0 10 9 6 x 10 3 0 014 12 15 0 05 9 9 x 1076 0 1 Table 7 5 Gain loss probability for an APV25 For the M200 an pinhole rate of 0 2 of the channels was found The other rates allow an estimation of the effect of the pinhole rate on the production quality The fact that pinholes are frequently found as a cluster of several bad strips Hartmann 2003a will increase the number of APV25 connected to more than three pinholes Figure 7 53 Scratch induced pinholes will arise due to mishandling For these scratces all affected strips show a pinhole behaviour sensors compare Sec 4 1 2 on page 24 for the sensor design and Sec 5 5 3 1 on page 56 for the effect of the metal overhang on the leakage currents Especially the leakage current effect of surface scratches was unexpected and new Hartmann 2003a Krammer 2003 Manneli 2003 Figure 7 11 on page 95 shows the effect of a single strip breakdown which can be taken as guidance for the expected influence of a surface scratch on the silicon sensors leakage current Surface scratches increase also the risk of gain loss due to pinholes significantly If the scratch penetrates the thin oxide layer thickness 300nm the break through voltage of the
27. Ausgangsdaten bestimmt Vor den Daten wird ein 12 bit digitaler Datenkopf gesendet welcher die Nummer der Pipe linezelle und ein Fehlerbit enth lt Eine Ausgabegeschwindigkeit von 20 MHz damit halb so gro wie die Taktfrequenz erm glicht ein nachgeschaltetes 2 1 Multiplexen von zwei Auslesechips auf eine Datenleitung Der APV25 Chip generiert alle 70 Taktzyklen ein Synchronisierungsbit oder startet die Ausgabe von Daten so ein Auslesesignal empfangen wurde phase locked loop analogue pipeline voltage build in a 0 25 um process III Zusammenfassung SDA Analog Output 192 storage cells x 128 channels shapers Abbildung 3 Blockdiagramm des APV25 Chips Jeder Vorverst rker ist mit einer 192 zellentiefen Pipeline verkn pft welche die Auslesedaten f r die Zeit der Ausl seentscheidung zwischenspeichert Der Zugriff auf die Pipeline wird von einer Schreib Lesezeigerlogik kontrolliert welche Zellen die auf die Auslese warten entsprechend berspringt Das vom Vorverst rker und Pulsformer gefaltete Signal kann mittels eines APSP Filters entfaltet werden bevor es durch einen 128 1 Multiplexer und das Auslesesystem bertragen wird Zugriff auf die internen Register kann ber eine I2C Schnittstelle erfolgen Eine Kablibrationseinheit steht f r Funktionstests zur Verf gung Heier 2001 Die analoge Pipeline wird von einer Schreib Lesezeigerlogik verwaltet Der Schreib zeiger eilt hierbei dem Lesezeiger
28. CMS detector utilises three calorimeters the innermost is the ECAL placed together with the main part of the hadronic calorimeter HCAL inside of the magnet coil Furthermore the HCAL has two forward extensions installed 11m up and downstream of the interaction point 3 4 1 Electromagnetic Calorimeter CMS has made precision electromagnetic calorimetry a priority and has chosen a design based on scintillating crystals of lead tungstate PbWO CMS Collaboration 1997b Lecomte 2002 Lead tungstate has several properties which makes a good choice for this detector It is intrin sically radiation hard and therefore suitable to withstand the harsh radiation environment at the LHC Lead tungstate has a small Moli re radius of 20mm allowing separation of nearby adjacent electromagnetic showers It has a short radiation length of 8 9mm which means that the detector can be relatively compact Finally by controlling the impurities it is possible to produce crystals with a fast scintillation time of about 10 ns The ECAL is comprised of three separate sections the central barrel n lt 1 5 and two end caps 1 5 lt n lt 2 5 In the barrel region solid state avalanche photodiodes APDs are used which are not suitable for the end cap region because the irradiation induced reverse bias currents would become excessive The ECAL end caps use radiation tolerant vacuum photo triodes VPTs as photodetectors On the other hand these VPTs can n
29. FED It containes the methods APVData class representing the data of a single APV25 frame including its pipeline address transmitted in the digital header and the bunch crossing and trigger counter values from the FED Event class holds the data organised in the APVData classes for all APV25s of a FEH APPENDIX C KARLSRUHE READOUT LIBRARIES 135 Block class represents the fact that the FED is able to store a larger number of events in its DPM The FED is configured by the fed library to store up to 128 events and this will be readout and analysed block wise ShrdMemMaster class sets up the shared memory segments used to exchange the readout data with the client class used to analyse to write them to disk or the present them to the user The shared memory uses the Block class structure to exchange the data with the clients and the master is responsible to fill the Block structure and its internal Event and APVData structure with the data readout from the FED The synchronisation of the shared memory is done via semaphores signalling blocks waiting to be processed by the clients or being processed by them ShrdMemClient class is the counter part of the ShrdMemMaster class All clients use this class to get access to the shared memory and it contains Analysis library hosts the Graph the Histogram and the Analysis classes The Analysis class is one of the shared memory clients It performs an online analysis of the data taken and uses the other tw
30. Murray APVMUX User Guide Version 1 0 CERN 2000 S My CMS silicon strip detectors Int J Mod Phys A16S1C 1074 1077 2001 National Instruments DAQ PCI DIO 96 PXI 6508 PCI 6503 User Manual National In struments 1998 http www ni com pdf manuals 320938c pdf National Instruments DAQ NI 6034E 6035E 6036E User Manual National Instruments 2002 http www ni com pdf manuals 322339d pdf A Neviani APV25S0 Calibration Curuit Technical report Padova 1999 D Passeri P Ciampolini A Scorzoni and G M Bilei Physical modeling of silicon microstrip detectors Influence of the electrode geometry on critical electric fields IEEE Trans Nucl Sci 47 1468 1473 2000 Philips 12C Specification Technical report Philips semiconductors 1992 to be checked Philips PCF8584 I2C bus controller data sheet Philips http www philips com October 1997 P Placidi A Marchioro and P Moreira CMS Tracker PLL Reference Manual Version 2 0 CERN 2000 PCI 9050 1 Data Book PLX Technology http www plxtech com December 1999 PCI 9080 Data Book PLX Technology http www plxtech com January 2000 M Raymond HIP and pinhole effects on APV25 CMS General Tracker Meeting December 2001a http cmsdoc cern ch Tracker managment GTM GM_01 12 Mark_ CMShipstalk ppt M Raymond Results form lab tests of module CMS Tracker Electronics Meeting October 2001b http cmsdoc cern ch ghall TKEL TKEL_1001 Raymond_1001 pdf
31. Neviani 1999 designed to allow for on chip testing and measurement of the APV25 analogue front ends impulse response Basically the circuit generates and injects current pulses of programmable total charge into selected readout channels Furthermore the time interval between stimulation and the front end amplifiers sampling point is programmable both on a coarse 25ns and a fine 25ns 8 3 125 ns time scale The amount of charge injected is programmable in steps of 0 1fC 0 025 MIPs in the range between 0 and 25 5fC 0 6 4 MIPs however since the resistor and the capacitance defining the calibration amplitude have large tolerances the charge is not very well defined The 128 analogue channels are grouped into eight selectable 16 channel sets and the charge is injected only in the selected one Channels belonging to a given set are separated by seven Channels belonging to other ones minimising channel to channel interference effects The polarity of the charge pulse alternates with every calibration request thus inverting and non inverting modes can be tested Interfaces The APV25 has fast and slow control interfaces The slow one uses the I C industrial protocol Philips 1992 to program registers which control the chips logic like the readout mode and its analogue circuits by generating via on chip digital analogue converters DACs defined currents and voltages required by the analogue parts The fast control signals consist of th
32. RUN_CAL finished 2632 eventa done 03 04 13 03 04 13_15 44 41 RUN_CALSHAPE finished 37760 events done 15 40 35 RUN_CALSHAPE started going to take 40960 events 03 04 13715 44 42 RUN_PED started going to take 1000 events 03 04 13_15 44 49 RUN_PED finished 1062 events done 03 04 13_15 44 50 RUN_PED started going to take 1000 events 03 04 13_15 44 58 RUN_PED finished 1062 events done i 4 58 RUN_CAL started going to take 3072 events 03 04 5 11 RUN_CAL finished 2832 events done 03 04 13 15 45 12 AUN CALSHAPE started going to take 40960 events o ta E Pr Q Figure 6 14 The GUI gives control to all functionalities of the Karlsruhe readout systems The figure shows the GUI main window which gives access to different run types like Pedestal Run or Calibration Run Other windows can be launched like the Histogram View which gives access to the online analysis results or like the Temperature Control which allows the monitoring and steering of the temperature slow control part values are initialised during module connection phase and can be refered to from the processcontrol menue file controls the entries of the GUls main menu and their connection to the different sections of the processcontrol prcesscontrol file holds the sequences launched by the GUI if the corresponding menu item has been selected Within the processcontrol file complex sequences of different task can easily realised
33. Zerf lle und im System der B Mesonen werden neue Kan le zur Untersuchung der CP Verletzung zug nglich 2 Der CMS Detektor Das Design des CMS Detektors siehe Abb 1 fu t auf e einem sehr guten und redundanten Muon Detektor e optimal dazu angepa ten elektromagnetischen und hadronischen Kalorimetern e einem hochaufl senden Spurdetektor Das Design geht von einem 13 m langen und 6 m durchmessenden supraleitenden Magnet system aus welches das gr te Element in Bezug auf Ausdehnung und Gewicht darstellt Das Muonsystem ist hierbei in das Magnetjoch integriert Der Innendurchmesser des Mag neten ist hinreichend gro um sowohl das elektromagnetische als auch den gr ten Teil des hadronischen Kalorimeters HCAL in sich aufzunehmen Das ECAL setzt sich aus 83 000 Blei Wolframat PbWO Kristallen zusammen wel che aufgrund ihrer kurzen Strahlungsl nge ein kompaktes Design des ECALs bei her vorragender Schauertrennung erlaubt Das HCAL ist als Kupfer Messing Szintillator Sandwichkalorimeter realisiert Seine St rke w chst von 6 6 auf ber 10 Wechselwirkungs l ngen von der Mitte zu den Endkappen hin an II Zusammenfassung SUPERCONDUCTING CALORIMETERS COIL ECAL HCAL Scintillating PbWO4 crystals Plastic scintillator brass sandwich IRON YOKE MUON MUON BARRE ENDCAPS Drift Tube Resistive Plate Cathode Strip Chambers esc Chambers DT Chambers RPC Resistive Plate Chambers RPC Abbil
34. a lower level see Fig 7 24 on the preceding page While the deterioration of the signals amplitude may be explained by multiple scattering and cluster truncation remains the different voltage behaviour unexplained The same behaviour is visible in De convolution mode shown in Fig 7 25 for room temperature with a radioactive source and for 10 C and cosmic rays in Fig 7 26 7 5 3 Signal to Noise Ratio vs Leakage Current As the leakage current of the module increases the noise compare Sec 4 3 3 on page 41 it is expected that the SNR will decrease Therefore the SNR is measured at the DTS where the leakage current can be increased artifically up to currents of 1mA by illumination with IR LEDs see also Sec 7 6 Dec wo Inv Mode wo Cal Module 30216630300056 Dec wo Inv Mode wo Cal Module 30216630300056 104 7 6 Infrared LED Studies SNR 7 28 26 Figure 7 28 SNR in Peak mode as 24 function of leakage current taken with a Sr at a voltage of 300 V at the DTS 22 As expected the SNR deterioration is stronger in Peak mode 30 than it is in Deconvolution compare Tab 4 5 on o apt et eo Se page 44 Leakage current A N o Peak wo Inv Mode wo Cal Module 30216630300027 ST Figures 7 27 and 7 28 shows the SNR dependency on the leakage current measured for two modules in Deconvolution and Peak mode both without inverter The deterioration of the SNR for the two modules measured in Deconvolut
35. an application to The extention so stands for shared object and indicates that we have a shared library On Windows systems the extension dll for dynamic linked library is used APPENDIX C KARLSRUHE READOUT LIBRARIES 131 void FEDobj Config a constructor taking a pointer to a Config class described later on page 135 The constructor parses from the Config class some variables needed by the FEDobj class like the name of the FED device file the synchronisa tion thresholds the clock delay and a flag whether header finding or scope mode data capture should be launched later void FEDobj as destructor closes the devices and cleans up all memory allocated by the FEDobj class void start_digitisation and void stop_digitisation methods to launch or to stop the data acquisition long getNfilled unsigned short 0 method returns the number of events pending for readout in the DPM The optional argument can be used to check the number of triggers received by the FED with a number of triggers generated by the readout system This is implemented because the FED is very sensitive to thermal heat up noise pick up and cross talk on the trigger line which all may results in corrupted data capture void readFED int amp short int amp method which takes a reference to the number of events to be readout a pointer to the destination address and again a reference to an integer returning the number of bytes transfered to the gi
36. and Collection 2 2 o e e 38 4 3 2 Radiation Effects sad 2 4 la ee A a ehe 40 4 33 Nore Analysis 4 0 00 A ee eee Bsa ee ae 41 Quality Control at CMS Tracker Modules 45 5 1 Silicon Strip Tracker Production Scheme 2 2 2 2 22 nn nn 45 5 2 Sensor Quality Test Strategy 2 2 02 2 nn nn 47 5 2 1 Sensor Quality Test Centre 2 2 2 2 Co e o 47 5 2 2 Process Qualification Centre 2 2m mon 48 5 2 3 Irradiation Qualification Centre 2 2 2 o e e 49 5 3 Quality Test Strategy for the Front End Electronics 50 5 3 1 Front End Hybrid Industrial Test 2 222 0 50 5 3 2 FEH Bonding and Quality Test Centre o 51 5 4 Module Quality Test Strategy ee ee 52 5 4 1 Module Assembly Centre 2 2 2 2 oo En nn 52 5 4 2 Bonding and Module Quality Assurance Centre 53 5 4 3 Petal Integration Centre 0 0 ee ee 54 5 5 Module Error Type Detection 2 2 2 mon nn 54 5 5 1 Non electrical Errors 22 2 Coon 54 5 5 2 General ASICs Error Detection 2 22 2 2m mn nen 54 5 5 2 1 PC Connection Problems 2 2 22 22 55 5 5 2 2 Leakage Current Failures 2 2 2 Coon nn 55 5 5 2 3 APV25 Header Problems 55 5 5 2 4 APVMUX and PLL Failures 55 5 5 2 5 Low Voltage Power Consumption 56 5 5 3 Strip Error Detection 2 2 aoaaa a 56 5 9 3 17 gt Ser tches 2 22
37. calibration circuit but these are only of interest for expert debugging purposes MUX library The functionalities of the APVMUX ASIC compare Sec 4 2 1 3 on page 32 are addressed be the Mux class The light weighted Mux class supported method con tain void Mux ushort amp addr 2c bus constructor call taking a reference to an un signed short variable holding the 12C address of the chip to be addressed and a pointer to the I2c class as arguments It initialises the APVMUX to the highest resistivity of 4000 int readNumkResBits and int setNumResBits int n reads back the number of resistors or changes them according to argument given DCU library contains the methods needed to run and control the DCU The Dcu class contains void Dcu I2c i2cbus constructor taking a pointer to the 2c class as arguments The I C base address is hard coded in the current release to be 0 This may change in future int readChannel int ch takes the channel number as argument and performs a mea surement on that given channel This method returns the result as integer value A second read method int readChannels int ch takes a pointer to an integer array which is filled with the measurements of all DCU channel The latter func tion does not check if the given array has the according size which may result in serious problems if not int readID returns the unique chip ID which is a 16 digits unsigned integer value Motherboard library contains
38. cooling cycle the pro is a much easier control loop due to the inertial reaction The cooling speed can be increased by keeping the middle layer at 10 to 15 C during the phases of module exchange and testing at room temperature The reheating of the system is much quicker because it is supported by the electrical power dissipation of the peltier elements compare Fig 6 17 6 Karlsruhe Test Stations 83 6 5 Diagnostic Test Station In contrast to the small test volume of the FTS the Diagnostic Test Station DTS is kept voluminous Having a lot of available space in the test volume gives all possibilities to realise additional test functionalities within the station Figure 6 18 View on the Karlsruhe Diag nostic Test Station The large front door gives access to the large interior volume which is vibrational decoupled from its sur rounding box This decoupling is essential for the use of micromanipulation probes Photomultipliers are available for source tests one of them can be seen on inner backside as well as for cosmic rays one PM is visible on top A microscope can be used together with an installed gate sys tem to inspect the modules surfaces Fur thermore a source can be mounted or a laser fiber can be attached to perform strip scans For ESD protection several GND connection points are available for the op erator e g cable visible at the left front corner 2003 2 17 6 38pm Figure 6 18 shows a picture
39. fault of a ring 5 prototype FEH 116 Inverter fault of a ring 5 prototype module 0 4 117 Effect of a dozen pinholes on the calibration amplitudes 117 Effect of a dozen pinholes on the noise 2 a a 118 Pinhole induced gain loss of APV25 2 2 a 119 Scratch induced pinholes 2 2 2 2 En nn 120 Front End Hybrid schematic part lar era 138 Front End Hybrid schematic part 2 lara ee hae ee 139 PA of a ring 6 module 2 2 on on 140 144 LIST OF TABLES List of Tables 4 1 4 2 4 3 4 4 4 5 5 1 7 1 7 2 7 3 7 4 7 8 Al A 2 Cl D 2 Sensor types as used for the tracker barrel and end cap modules 22 End cap module types Emm nn 23 Symbols transmitted over the trigger line 2 2 EC nn nn 32 Noise sources types and relative ENC evaluation 2 222 2 2 nn 44 Noise contributions for Ring 6 Modules 2 2 2 nn nn nn 44 Module quality grade definitions 2 22 2 CC En nn nen 60 Common mode noise and slope for the different readout modes 90 SNR values measured for Peak and Deconvolution mode 101 Signature of a short 2 on on ee 109 Signature of a missing bond 0 0 0 0 eee ee eee 112 Gain loss probability for an APV25 0 Eon 120 Acronyms used within this work 2 000000 ee eee 126 Symbols used within this work 0 0 0 000000 eee eee 127 PowerPack modes lt 2 2 2 A ee Ras e
40. from LED sig E nals as they are shown in the me 8 Figs 7 31 to 7 33 on page 106 8 O OENE SAONE O NAO E The flag result corresponds to A UNE Manet A A AA E a bit sum of four LED ar C 3 rays The three channels have EEE EEE Bene A o A E 3 been tagged by all four arrays 5 E 3 which shows that the PA to z E E E E 2 sensor bond connection failed ae I a E gt EN o i 500 33 for all of them DB channel number s Bias Peak Mode Deconvolution mode voltage wo Inv w Inv wo Inv w Inv Noise Cal amp Noise Cal amp Noise Cal amp Noise Cal amp 200V tag tag tag tag tag tag tag tag 20V tag tag tag tag tag tag tag tag OV tag tag tag tag tag tag tag tag 9V invisible tag tag tag invisible tag tag tag Table 7 4 Signature of a missing bond as measured for channel 264 on Module 30200020000638 for different bias voltages 9 V means that a forward bias of 9 V has been applied to the module A tag requires a clear structure which can be identified easily and is ranked from perfect clear to still tag able Invisible reflects that not distortion can be found bond position unambiguously There are in principle three possible positions for a missing bond e the connection between the APV25 and the PA which should be tested and known from the FEH qualification e between PA and the f
41. immer um eine einstellbare Taktzahl voraus Wird nun ein Auslesesignal empfangen wird die aktuelle Position des Lesezeigers zur Auslese in einem FIFO Speicher abgelegt Dieser FIFO Speicher hat eine Tiefe von 31 Zellen wobei im Deconvolution Modus jeweils drei Positionen f r die Verarbeitung im APSP Filter gespeichert werden Weiterhin verf gt der APV25 Chip ber eine interne Kalibrationseinheit mit der un terschiedliche Ladungsmengen in den Vorverst rker eingekoppelt werden k nnen Die Zeitspanne zwischen der Eingekopplung des Signals und dem Abtastzeitpunkt kann hier bei in Schritten von 3 125 ns variiert werden Der Auslesemodus und eine interne Kalibrationseinheit sowie weitere Register des APV25 Chips werden ber eine I2C Schnittstelle kontrolliert Mit der Ausl sesignallei tung steht ein weiteres schnelles Interface zur Verf gung welches Auslesesignale Resetan weisungen oder Kalibrationsanfragen bertr gt Die verwendeten Muster haben hierbei eine L nge von drei Bit Dies die Tiefe des Auslese FIFO Speichers und die Dauer der seriellen Daten bertragung begrenzen die Auslesesignalrate APVMUX Der APVMUX Chip bernimmt das paarweise Multiplexen der APV25 Chips auf eine gemeinsame Ausgabeleitung Des Weiteren wandelt er das Stromsignal der APV25 Chips in ein Spannungssignal um wobei dieses ber verschiedene Widerst nde erfolgen kann DCU Der DCU Chip integriert einen 12 bit ADC mit I2C Interface zur Messung ver schi
42. important consequence either the Higgs boson must have a mass less than about 2 High Energy Physics at the Large Hadron Collider 5 800 GeV c or the dynamics of WW and ZZ interactions with centre of mass energies of the order of 1 TeV will reveal new structures The Higgs boson is an essential part of the analogy to the Meissner effect in supercon ductivity that leads to an excellent understanding of the masses of the electroweak gauge bosons W and Z as consequences of the electroweak symmetry breaking At tree level in the electroweak theory we have My gv 2 10 Grv2sin w MZ Miy cos By where the electroweak scale v Gp va 3 246 GeV is set by the vacuum expectation value of the Higgs field The problem is how to give mass to the weak gauge bosons W and Z without breaking gauge symmetry which is required for an renormalisable field theory In order to understand it one may consider the weak interaction Lagrangian of the charged scalar field i e T gt t F gt t 1 gt f 5 0 9 6 ao 9 6 gt wale Hate Wir Wa 22 where gt gt gt gt gt W ww 56 W 5 W gW x Wy 2 3 is the field tensor for the weak gauge bosons Wp The charged and the neutral W bosons form a SU 2 vector reflecting the non Abelian nature of this gauge group which is responsible for the last term in Eq 2 3 and leads to gauge boson self interaction Correspondingly the gauge transformation on W ha
43. of 100ns full width at half maximum Sampling the shaper output with the bunch crossing frequency Peak mode signal one can deconvolute the signal by adding three consecutive Peak mode signals with according weights Bingefors et al 1993 Hall 1994 Using this Deconvolution mode the pulse width reduces to 25 ns Furthermore the APV25 gives access to three consecutive Peak mode signals within the Multi mode The APSP adds a power consumption of 0 2mW channel and has been designed to keep an equal gain of 100 mV MIP in both peak and deconvolution mode Figure 4 8 shows the schematic circuit block diagram of the APV25 For a more detailed discussion see Friedl 2001 Jones et al 1999 When the chip receives an external trigger analogue samples are retrieved from the pipeline processed in the APSP and fed to a single output via a 128 1 analogue multiplexer Due to the used tree structure the output of this multiplexer stage is non consecutive To re trieve the physical channel number c from output sample number n the following calculation must be done c 32 n mod 4 8 int 5 31 int 5 The output data stream contains these 128 analogue data samples preceded by a digital header compare also Fig 4 11 on page 33 which consists of 3 start bits followed by an 8 bit pipeline column address location of one of the 192 time slices and an error bit The output speed can be selected to be either at full system speed 40 08 MHz o
44. of silicon wafers needed and to be tested reaches a number of 25000 Hereby CMS has to deal with 15 different sensor geometries masks and as many different module types four for the barrel region ten for the end caps see Tab 4 1 and 4 2 on the next page and one additional different geometry for the inner disks The modules represent the smallest independent units within the tracker which are integrated into rods for the barrel region and petals for the end cap sensor width length thickness pitch of of type mm mm um um strips sensors IB1 63 3 119 0 320 80 768 1536 IB2 63 3 119 0 320 120 512 1188 OB1 96 4 94 4 500 122 768 3360 OB2 96 4 94 4 500 183 512 7056 W1 TEC 64 6 87 9 87 2 320 81 112 768 288 W1 TID 63 6 93 8 112 9 320 80 5 119 768 288 W2 112 2 112 2 90 2 320 113 143 768 576 W3 64 9 83 0 112 7 320 123 158 512 640 W4 59 7 73 2 117 2 320 113 139 512 1008 W5a 98 9 112 3 84 0 500 126 142 768 1440 W5b 112 5 122 8 66 0 500 143 156 768 1440 W6a 86 1 97 4 99 0 500 163 185 512 1008 W6b 97 5 107 5 87 8 500 185 205 512 1008 W7a 74 0 82 9 109 8 500 140 156 512 1440 W7b 82 9 90 8 90 8 500 156 172 512 1440 Table 4 1 Sensor types and multiplicities used for the SST barrel and end cap modules Different thickness of the substrate is chosen to compensate noise for longer strip length of models with two sensors Hartmann 2003b 4 Silicon Strip Tracker and its
45. of the DTS Its outer dimensions are 100x70x70 cm3 wxdxh The box has a light thermal insulation with a 2 5cm thick layer of Polyurethane Although the DTS has no embedded cooling system it is possible to perform tests at low temperatures inside To do this the internal volume is subdivided and an external chiller is used to cool the active volume Temperatures of down to 20 C have been reached inside the DTS The door opens to the top which means that the lower part of the station builds a tub and that the air will not be exchanged as quick as it would be if the complete front would be opened The door of the DTS is connected to an interlock sys Figure 6 19 The switch of the tem used to kill the HV and low voltage LV supplies DTS interlock system will only be of the DUT if the door is not closed correctly Therefore closed if the DTS is closed and two switches are mounted in the DTS door which will be looked closed only if the door is closed and locked see Fig 6 19 The readout system used for the DTS is also based on the components described in Sec 6 1 on page 61 The readout electronics is placed below the station which includes the readout system the power supplies the readout PC and the controller for the stepper motors of the gate system 84 6 5 Diagnostic Test Station Figure 6 20 View inside of the Karlsruhe Diagnostic Test Station A microscope can be attached to the linear gate system which will be used for
46. on page 120 Beside a complete module V breakdown also individual strips may go into breakdown due to localised defects Figure 7 11 shows an IV curve for a sensor with a strip breakdown where the poly silicon resistor limits the current at higher voltages N o o T PTT EPA Leakage current A Figure 7 11 IV curve for a single strip breakdown taken at the QTC Karlsruhe Above a bias voltage of 250 V a single a a o e strip causes the steep increase of the Bias voltage V leakage current 7 4 Signal Performance Signal performance can be estimated using the calibration circuit but only with a very poor precision compare Sec 4 2 1 2 on page 32 More accurate results for the signal performance can be obtained using radioactive sources or cosmic ray particles While the use of the calibration circuit gives an immediate result after a few seconds of run time the more precise calibration using cosmic rays takes at least several hours of data taking 7 4 1 Calibration Signal Figure 7 12 shows the comparison of the calibration pulse shape in Peak and Deconvolution mode for a 1 MIP equivalent signal These pulse shapes are obtained by variating the time between the stimulation by the calibration unit and the sampling point of the APV25 compare Sec 4 2 1 2 The used time scale of latency steps is the natural time scale of the LHC given by the bunch crossing time of 25 ns The calibration
47. readout point of view it is the database numbering scheme which looks natural because this is correlated to the cabling and the FED channels independent of the FEH type used Ring sensor scheme database scheme opposite 1 1 768 1 768 no 2 1 768 1 768 no 3 1 512 1 512 no 4 1 512 1 512 no 5 1 768 1 768 no 6 1 512 1 512 no T 1 512 512 1 yes Table D 2 Sensor and database numbering scheme APPENDIX F FRONT END HYBRID SCHEMATIC 137 E I2C bus protocol The I C bus system is used in a wide variety of commercial applications It is based on a two wire serial protocol consisting of a data line called serial data SDA and a clock line serial clock SCL Both lines are connected with pull up resistors and have to be pulled down actively by the connected devices During idle time both lines are always on a high level thus there is no clock signal on the SCL line producing some pick up noise All transmissions start with a falling edge on the SDA line while the SCL line is still high The chip who pulled the SDA low is the master of the 12C bus for that transmission and has to serve the SCL for that complete transmission During a transmission the SDA line has to be stable while the SCL is high The transmission consists of tokens of 8 bit length The first 8 bit transmit slave address 7 MSB and a read write flag the LSB Depending on the read write bit the slave device is expected to receive or send the
48. sensible to speak of an elementary Higgs boson Failure to find a Higgs boson over the range would therefore rule out the SM Then the Higgs sector either consists of non standard Higgs bosons or the electroweak symmetry breaking occurs via some strongly coupled processes that will manifest itself in the study of WW scattering 2 3 Supersymmetry In the supersymmetric standard model a doubling of the particle spectrum is predicted For every particle that has been discovered supersymmetry predicts a supersymmetric partner that has not discovered yet The extra particles circulate in loops and protect the hierarchy from quadratic divergences But supersymmetry can not be exact otherwise the particles and their superpartners would be degenerate in mass which is clearly not the case But even with a broken symmetry different masses for particles and their superpartners the radiative correction from virtual boson and fermion loops are of opposite sign Thus since the particles and their supersymmetric partners are assumed to have the same couplings the one loop divergences will cancel The most widely quoted scheme is that of the Minimal Super Symmetric Model MSSM which consists of taking the SM and adding the corresponding supersymmetric partners In addition the MSSM contains two hypercharge Y 1 Higgs doublets which is the minimal structure for the Higgs sector of an anomaly free supersymmetric extension of the SM Fur thermore
49. stage to saturate with high currents and the common inverter stages supply drops The latter causes the chip wide gain loss currents of 200 3004A The second step taken to reduce the pinhole sensitivity of the APV25s is the exchange of the external inverter supply resistor The final FEH will utilise a inverter resistor of Riny 50 Q which is expected to increase the number of pinholes tolerable by an APV25 During the Milestone 200 compare Sec 5 2 on page 47 the Quality Test Centres QTCs measured a pinhole rate of 0 2 Hartmann 2003b Furthermore the M200 showed that pinholes frequently appear as a cluster of two neighbouring bad strips Taking the rate of 0 2 the probability of an APV25 to be connected with more than three pinholes results is given in Tab 7 5 Hereby modules with one and two sensors are treated separately While the results of Tab 7 5 are not so dramatically especially with the changes on the final FEH in mind there is another serious source for additional pinholes surface scratches 7 9 Scratches While from former experiments surface scratches are known to produce localised defects like shorts or breaks there is a complete new effect found for the CMS silicon detectors Scratch induced shorts are still visible like shown in Fig 7 34 on page 107 but they will also affect the leakage current behaviour because of the metal overhang technique used for the CMS silicon 120 7 9 Scratches Probability of an
50. the LED array system is not foreseen to be used for electrical stress tests of the silicon To do so a special LED system has been equipped with LEDs of 1050nm wavelength which turns out to be much more expensive than those ones used for the regular type of LED arrays This special system is used in the Diagnostic Test Station DTS for special stress studies see Sec 6 5 on page 83 The LED system is used collaboration wide for module quality tests due to its unique pinhole detection capability compare Sec 7 6 on page 104 6 Karlsruhe Test Stations 69 switch switch switch switch network network network network PLD external 6V 12Ah Peltier Peltier 6V 12Ah 6V 12Ah power inputead acid battery control control lead acid batteryead acid battery ALTERA MAX7064 SLC44 6 J e y J Strobe Select 8 Bit Data via Motherboard to PC Figure 6 8 Karlsruhe Power Pack is used to disconnect the readout system of the FTS from the in house power system and to control the peltier elements based cooling facility of that station 6 2 6 Power Pack and Peltier Control The power pack is a special device developed for the Fast Test Station FTS The basic idea behind the power pack is to use batteries instead of power supplies to run the electronics which minimises the noise pickup from the in house power system The second idea behind
51. the superposition of the eight calibration units compare Sec 4 2 1 2 on page 32 a even clearer tag is visible in the calibration signal like presented in Fig 7 36 on the following page Feeding a calibration pulse into one of the shorted channels will result in a signal in both of them In Fig 7 36 the pulse fed into channel 62 causes a significant smaller signal compared to the channels 54 108 7 7 Module Fault Detection Studies Sensor channel number 460 455 450 445 440 80 60 40 Figure 7 35 Shorted strips signature in calibration ampli tude for both Peak and De convolution mode The solid lines correspond to inverter on modes while the dashed ones 20 are without inverter Shorted channels show a significant loss in calibration pulse height 20 Calibration amplitude ADC 60 40 Peak w amp wo Inv Mode w Cal Module 30200020000638 Sensor channel number 460 455 450 445 440 80 A Peak Mode 60 Figure 7 36 Shorted strips signature in calibration signals for both Peak upper plot and Deconvolution lower plot modes The dashed lines are with inverter on and the solid ones correspond to inverter off modes The charge sharing due to the short results in an easy detectable signal in the neighbouring channel 40 20 Calibration amplitude ADC e a ce 60 65 70 75 Peak w amp wo Inv Mode w Cal Module 30200020000638 70 75
52. well and the bias current has to be monitored with a resolution of the order of 50nA for IV measurements An appropriate interlock system has to be implemented e Clock and trigger signals have to be generated for the FEH s ASICs the FED and external devices such as pulsers which may be used e g for infrared lasers light emitting diode LED arrays etc The electrical levels of the trigger and clock line have to be selectable according to the different logic standards and external trigger processing has to be implemented e An 12C bus has to be supported e As ADC the FED will be used because of its good performance the built in automatic header finding and last but not least the technical support within the CMS collaboration e A slow control system with temperature and humidity control including the possibility of active thermo cycles down to at least 10 C is needed e Possible usage of external signals like infrared light flashes from a laser or a LED system or signals induced by particles of radioactive sources or cosmic rays to verify module functionality e Good modularity to keep different functionalities disentangled which allows parallel development of components simplifies the development itself and eases debugging Two setups are realised with a different scope The first station is focused on performing fast functionality and qualification tests and is called Fast Test Station FTS This station has a built in cooling
53. 0 which is important for the dao ron ee le nee ne ea te 9 128 130 132 134 136 138 140 1422 2 SNR Latency 25ns Sensor channel number 500 400 300 200 Senn Ze Q FE i lt E g 80 EX ao oF 75 o s FP 70 2 H Figure 7 13 Calibration amplitude in amp 8 P 8 sE Peak wo inverter mode The overall am E 28 A EE E E 8 plitude variation is at a scale of 15 and 60 28 E gs the shape of increasing amplitudes with a a E z2 the channel number is a common phe ee d Z3 as lo Bil pe ee fe jee ed 100 200 300 400 500 nomenon of the APV25 DB channel number to 15 on a single chip and different chips have different offsets Finally the calibration signals give access to the charge sharing of the readout strips Figure 7 14 shows the signal of a single calibration group Beside every eighth channel also the first neighbouring channel show a signal on the 8 level in Peak and on the 12 level in Deconvolution mode Sensor channel number 500 480 460 440 420 400 T T T T T T o o Figure 7 14 Calibration amplitude and cross talk in Peak upper plot and Deconvolution Mode lower plot both without inverter The cross talk to the D o Y o LS LAS LN ie LL LLL LL Calibration amplitude ADC E o neighbouring channels is of the order of 0 8 for Peak mode and of the order of 20 12 for Deconvolution mode Further 50 33 more the CM correction result
54. 0 um and sensors with breaks longer than 40 um are rejected Sensor Fabrication Sensor Fabrication Center Center 100 sensors 5 sensors 5 teststructures 1 sensors 5 teststructures 5 sensors 94 sensors 5 sensors Module Assembly Centers 100 modules Figure 5 2 Sensor logistics for the main production Nearly 30000 sen sors will be tested with the given per centages during the production period Simonis 2003 48 5 2 Sensor Quality Test Strategy On a 5 10 scale the sensors are electrically qualified Hereby two global and four strip by strip measurements are performed Global IV and CV scans from 0 up to 550 V will be done to determine depletion voltage and high voltage behaviour The current drawn on a sensor has to stay below 10 uA at 450V and the increase between 450 V and 550 V has to be less than 10 4A as a breakdown criteria The strip by strip measurements are the individual leakage currents styip limit 100nA the poly silicon resistors Ryo within a range of 1 5 0 5MQ and sensor uniformity of 0 3MQ around the mean the current over the coupling dielectric Ige limit 1nA at 10V and the coupling capacities Ce The last two measurements are devoted to identify pinholes shorts breaks and to check the dielectric All these tests have been performed on a 100 scale during M200 and pre series Hartmann 2002 5 2 2 Process Qualification Centre To ensure a hom
55. 1 then CP is violated which is called CP violation in mixing or indirect CP violation because the mass eigenstates are different from the CP eigenstates or in other words B and B are not orthogonal and CP is not conserved in AB 2 amplitudes 2 High Energy Physics at the Large Hadron Collider 13 CP violation in decay 147 47 1 E 147 A 1 then CP is violated which is called CP violation in decay or direct CP violation It occurs due to interference between various terms in the decay amplitude and requires that at least two terms differ in both their strong and in their weak phases CP violation in the interference between decay and mixing A 4 1 The third type of CP violation is possible in neutral B decay into CP eigenstate final state fep If CP is conserved then not only q p 1 and A A f 1 but the relative phase between q p and A7 A f also vanishes If not this is called CP violation in the interference between decays with and without mixing because it results from the CP violation interference between BP fep and B B fer Measuring CP The most promising method of measuring CP violation is to look for an asymmetry between T B fep and rB fep where fep is a hadronic state having a well defined CP eigenvalue 1 Commonly used are two particle systems like 6 Ks CP parity 1 rt 77 CP parity 1 and pP Kg CP parity 1 The golden mode for studying CP viol
56. 2 2 Standard Model Higgs The Standard Model SM as the combination of the Quantum Chromodynamics QCD as theory of the strong interaction and the electroweak interaction as the unification of Quantum Electrodynamics QED and weak interaction has the group structure S SU 3 c 8 SU 2 Q U 1 y 2 1 Experiments over the past thirty years have shown numerous confirmations of the SU 2 L8 U 1 y electroweak theory the existence of neutral currents the necessity of charm and the existence and properties of the weak gauge bosons WF and Z Experiments have also given essential guidance to the form of the evolving standard model through the discovery of a third generation of leptons 7 1v and quarks t b Finally experiments have shown a number of big surprises that have shaped both experimental and theoretical opportunities the narrow ness of J b and 4 the unexpectedly long B lifetime the large degree of B B mixing the extreme heaviness of the top quark and evidence of neutrino oscillations Ten years of precision measurements have found no significant deviations from the pre dictions of the electroweak theory A series of quite remarkable experiments not to mention the accompanying evolution in theoretical calculations have tested the quantum corrections of the electroweak theory loop effects to a precision of one per mil The net result of this prodigious effort is that we have found no evidence for new ph
57. 25 data frames starting with their digital header while in scope mode it delivers all ADC samples taken fedpmc_start_digitisation and fedpmc_stop digitisation enable or disables data capture which is initiated on receipt of either an APV25 frame an external or software trigger depending on the run mode Technically these routines only control the data storage into the DPM while the ADC are running continously fedpmc_readout_event transfers the data captured from the DPM to a given desti nation memory address Furthermore it transfers the bunch crossing and trigger counter values corresponding to the taken events Further miscellaneous functions are disposed by the RALFED library like status reset purge DMA buffer generate software trigger or update firmware routines and even more FED library All the routines collected in the RALFED library are used to control the behaviour of the FED but they do not manage the control variables used Therefore the FEDobj class managing the parameter space of the FED is implemented between the RALFED library and the application This FEDobj class is placed in the libfed so or FED library Placing the management of the FED parameters into a C class utilises the encap sulation of C and prevents misusage of the RALFED library routines which is of importance due to the fact that the FED may hang the system if it is addressed in a wrong way The FEDobj class reduces the number of methods accessible to
58. 3 Karlsruhe motherboard schematic 2 2 2 2 En Eon nn 64 Karlsruhe sequencer card picture and block diagram 65 Karlsruhe HV card picture and block diagram 2 2 2 2 nn nn 66 Picture and block diagram of the Karlsruhe IR LED array controller card 67 Picture and block diagram of the Karlsruhe IR LED array 2 2 2 2 68 Karlsruhe Power Pack 2 2 Co onen 69 Karlsruhe multiplexer device oaoa a 70 Block diagram and picture of FED 0 0000 008 71 Block diagram and picture of Flandes 72 Basic design of the Karlsruhe readout system software 75 Snapshot of the CVO tree 66 ee nen ee ae a EA 78 Graphical User Interface toa ded aa leh A es 79 The Karlsruhe Fast Test Station hardware setup 80 View on the Karlsruhe Fast Test Station 2 22 Enno nn 81 Slow control front end showing two cooling cycles 2 22 2222 82 View on the Karlsruhe Diagnostic Test Station 2 2 2222 83 Interlock system of the Diagnostic Test Station 2 2 2 2 22m 83 View inside of the Karlsruhe Diagnostic Test Station 84 Pedestal for the different readout modes 2 o e e 87 Pedestal dependency on inverter stage o 88 Raw noise for the different readout modes 2 2 2m nn nn 89 Common mode noise and slope distribution 2 2 22 22mm 89 Common mode correction 2 2 2 22 mn 90 Noise profile after common m
59. 3 Enge enge res een tudes for all channels of the ET ee ae given APV25 are plotted For 5 60 ae ane ee all channels expect the pin E o alal ER A er ER oe eee buen BURN hole one the calibration ampli 8 E ode as comiant wre he race ae ees Gai Ga ee E a 8 channel with the pinhole first MEA rd ak a RE regenerates its calibration am A ee g E plitude before it drops again u en Ed towards higher leakage cur a 350 A00 3 3 rents Leakage current A 3 4 5 g 4 2 Figure 7 45 Pinhole search 39 in the noise with artifical leak 3 age current shows that at leak a age currents of about 30 uA the pinhole lowest and bold 2 curve vanishes in the noise plots The two channel with increased noise are the chan nels 512 and 385 APV25 edge channels see Sec 7 7 6 Peak w Cal wo Inv Mode Module 30200020000638 50 100 150 200 250 300 350 400 Leakage current A ob difference between the APV25s virtual ground and a p implant of approximately 6 V pro totype design Although this is comparable to the expected value for a pinhole on a final module it has the wrong polarity Therefore it will drive the channel in saturation but the inverter transistor will not be turned on Figure 7 46 on the facing page shows the noise under 9 V forward bias Obviously the pinhole at channel 420 can be easily identified as a problematic channel but the second pinhole at channel 162 and the shorts at channels 61 amp 62 show tha
60. 402 0402 0402 0402 0402 0402 0402 0402 a o x o El la 5 E R15 R11 v2 cis AN 0402 0402 y x f oo z 3 25 g 3 Ss A El la P S lt u E 4 a El a ta a i 4 x a a El 4 E n oe o 2 R12 o 5 E 0402 m o p oia 5 470 00 00 000 0000 R20 RS R4 nC t V 0402 0402 0402 25 100 100 100 V von x x A x 3 4 El 4 dl o 4 e m qd 4 INTERCONNECTED E 5 E 5 ta E Figure F 1 Front End Hybrid schematic part 1 APPENDIX F FRONT END HYBRID SCHEMATIC 139 REVISIONS 01 03 2001 CHANGED THERMISTOR TO 22K 0805 01 03 2001 POSSIBLE RETURN TO V125 ADDED FOR TEST 26 02 2001 C15 TO VSS 13 02 2001 SAME MODIFICATIONS THAT FINAL TIB 02 10 2000 CONNECTOR PIN ASSIGNEMENT SUBJECT TO CHANGE 02 10 2000 CHANGE CONNECTOR FOR FW AND OB 07 07 2000 REALLOCATE CONNECTOR PINS 05 07 2000 RENUMBER APVMUX PINOUT 04 07 2000 CHANGE VOLTAGE DIVIDER FOR DCU CMS_FEH_6_OB_EC_POS DESSIN LES R9 R10 V125INV REF CFH IREF IREFBIAS c8 c5 c6 c9 c10 C11 cia c13 100N 100N 100N 100N 100N 100N 100N 100N 0402 0402 0402 0402 0402 0402 0402 0402 z O 5 g el EN x EX ox 3 c eo
61. 5 for a border APV25 compared to the inner ones and their corresponding slope distributions show a typical broadening of 40 60 again compared to an inner APV25 This can partially be explained by a cross talk from the sensors bias ring Hereby especially the first and the last strips of a sensor see the fluctuations on the bias ring compare Sec 7 7 6 on page 115 threshold can be changed by configuration files 90 Readout mode Peak wo Inv Mode wo Cal Peak w Inv Mode wo Cal Peak wo Inv Mode w Cal Peak w Inv Modew Cal Dec wo Inv Mode wo Cal Dec w Inv Mode wo Cal Dec wo Inv Mode w Cal Dec w Inv Mode w Cal Common mode RMS APV1 0 86 0 51 1 13 1 31 1 03 0 85 1 33 1 13 ADC counts APV2 APV3 0 73 0 68 0 40 0 36 1 04 1 02 1 24 1 24 0 87 0 79 0 61 0 54 1 18 1 15 1 03 1 05 APV4 0 75 0 41 1 07 1 29 0 90 0 61 1 25 1 08 7 2 Noise Studies Common mode slope RMS x107 ADC counts channel APV4 APV1 7 52 6 62 14 31 14 85 9 88 10 05 15 25 14 16 APV2 APV3 4 33 3 84 3 64 3 32 11 53 11 48 10 73 10 64 8 76 5 40 5 49 5 17 11 03 10 75 11 61 11 44 8 23 8 34 12 63 14 80 11 24 12 23 13 65 12 65 Table 7 1 Common mode noise and slope for the different readout modes measured for module Figure 7 5 Common mode correction The upper plot shows the raw noise of pedestal cor rected raw data The plot in the middle is the result of the CM cor rection while the lower plot shows the co
62. 8 Mean 2 154 RMS 0 074 Entries 128 Entries Entries lc DE 22 24 26 28 3 2 Peak wo Inv Mode wo Cal Module 30200020000503 Ei io ici 3 3 4 o 16 18 2 22 24 26 28 3 32 3 4 Channel noise ADC Channel noise ADC Eii Er ser ser Figure 7 6 Noise profile of all four APV25s after common mode correction shows a more or less Gaussian shape The CM corrected noise as shown in the middle plot of Fig 7 5 increases slightly towards the edge of the module which can be explained partially by the PA Because the length of the lines on the PA differ significantly up to a factor of two this results in a different serial resistance and thus in a slight increase in noise for longer lines Figure G 1 on page 140 shows the PA for a ring 6 module The longest lines are placed on both border sides of the PA caused by the different width of the FEH and the W6B sensor Beside this large structure a smaller one in terms of length variation is connected to the individual APV25s where the length of the lines typically increases from the mid of an APV25 towards its borders The measured resistive load of the PA for a Ring 6 module changes from 1300 for the longest line channels 1 amp 512 to 709 for the shortest around channel 192 amp 350 which adds to the Aluminium readout strips resistance of 650 Beside the serial noise from the PA additional contribution will arise from the capac
63. 8 This card has 24 input output lines which are grouped by 8 bit to a port labelled A Band C This ports can be either programmed as input or output and port C can be split into two groups Typically port C is used as handshake lines for the other two ports Unfortunately switching a port from read to write or vice versa results in all lines going to low Due to this the DIO card can not be used for negative logics This is the reason why the communication between PC and motherboard uses positive logics 6 2 11 MIO The multiple input output MIO card is used to measure analogue signals like temperatures humidities voltages or currents is a PCI 6035E card from Nation Instruments National In struments 2002 It serves 16 analogue input lines which can either be used as 16 independent unipolar lines or coupled to 8 differential ones Depending on the input levels the full range of the 16 bit ADC can be selected from 10 V via 5 V and 0 5 V down to 0 05 V and the input can be sampled at a maximum rate of 200 ksample s Furthermore the MIO card has 2 analogue output lines with 12 bit resolution and a full range of 10 V and maximum current of 5mA It also has 8 digital channels TTL levels which can be selected independently as input or output line Finally the card has 8 timing input outputs which can be used to measure frequencies times counts or even to generate interrupts 6 2 12 Slow Control Multiplexer The sl
64. 98 Figure 3 3 Pixel detector layout The barrel length is 52 cm Kotlinski 2002 flux of particles necessitates appropriate pixel size to keep the occupancy sufficiently low On the other side the pixel size is influenced by the area needed for the readout chip for the pixel analogue front end and readout circuit Furthermore material budget and sufficient cooling had to be kept in mind while designing the pixel detector The pixel detector consists of three barrel layers at mean radii of 4 3 7 3 and 11 0cm The two innermost layers will be installed directly while the third one will be added before start of the high luminosity phase of LHC The pixel barrel will be 52cm long and supplemented by two end disks each side see Fig 3 3 In order to achieve optimal vertex position resolution in both the r y and the z coordinates a design with a square pixel shape of 150 x 150 um was adopted To enhance the spatial resolution by analogue signal interpolation the use of charge sharing induced by the large drift Lorentz angle 28 at AT for electrons is made Hence the detectors are deliberately not tilted in the barrel layers but titled in the end disks resulting in a turbine like geometry The whole pixel system consists out of about 1400 detector modules arranged into 4 module ladders in the barrel and 7 module blades in the disks To readout the detector about 16000 readout chips are bump bonded to the detector modules see Fig 3 4
65. Aa AN Ha dee H R1 fe EE Se vo Dam B BD bb DB e e S63MW54 90 00 00 2K e o o 3 afejeoj2121322 8 84 a S63MW5 S63MW5 J1 1K 1 NLY ONE RESISTOR INSERTED Figure F 2 Front End Hybrid schematic part 2 140 APPENDIX G PITCH ADAPTER G Pitch Adapter TEC 204 4 um HE BRUSSELS 11 2002 a nn Er emi Figure G 1 PA of a ring 6 mod ule The left picture is scaled 1 1 while the right picture show the an enlarged upper part The different length of the lines on the PA is clearly visible The separated line on top is the HV return line LIST OF FIGURES List of Figures 2 1 2 2 2 3 2 4 3 1 3 2 3 3 3 4 3 5 3 6 3 7 4 1 4 2 4 3 4 4 4 5 4 6 4 7 4 8 4 9 4 10 4 11 4 12 4 13 4 14 4 15 4 16 4 17 4 18 4 19 5 1 5 2 5 3 5 4 9 0 5 7 5 6 5 8 5 9 5 10 5 11 5 12 5 13 5 14 5 15 Higgs production in proton proton high energy collisions Branching ratios for the main decay modes of SM Higgs bosons The unitarity triangle a B B transition diapr ms a san Boe nee Compact Muon Solenoid detector dimensions and layout Central Tracker dimensions and layout 2 2 Em Eon nn Pixel detector layo t ees gt 4 au ms sr Panda en en Construction principle of a pixel detector 2 2 2 2 nn nn St
66. Auslesesystems erste Prototypen vermessen Zusammenfassung VII T H Sensors Abbildung 5 Die Karlsruher Teststationen bauen auf kleinen funktionalen Einheiten auf welche entweder direkt vom PC aus betrieben werden oder auf einer externen Haupt platine realisiert sind Hierbei werden zentrale Funktionen wie die Takt und Ausl se signalerzeugung von selbstentwickelten Komponenten bernommen C Socket Connections Abbildung 6 Die Struktur der Karlsruher Auslesesoftware basiert auf dem Linux Be triebssystem f r welches entsprechende Hardwaretreiber entweder entwickelt I2C oder angepa t FED MIO DIO wurden Aufbauend hierauf sind die Funktionalit ten der unterschiedlichen Komponenten in Bibliotheken abgelegt auf welche die verschiedenen Prozesse zugreifen Die Steuerung erfolgt ber eine graphische Benutzerschnittstelle Lab View welche ber TCP IP angebunden ist VIII Zusammenfassung Entries 641 Mean 21 22 2 x ndf 59 64 72 E y Width 1 752 0 3362 u MP 19 0 4266 430 3 18 57 2 a 5 407 0 5364 2 a a 8 SEL el ON ARAS EN MAA TER a o Dec wo Inv Mode wo Cal Module 30216630300027 a o a Y Y a W S Y a D Ss a a Abbildung 7 SNR im Deconvolution Modus bei einer Verarmungsspannung von 300 V gemessen bei 10 C Die Breite der Verteilung wird vom Rauschen der Elektronik do miniert Peak Modus Decon Mod
67. End cap Modules 23 module type radius active area of strips of of cm cm APVs modules Ring 1 233 0 320 2 768 6 6 144 Ring 2 323 0 411 1 768 6 6 288 Ring 3 392 1 502 7 512 4 640 Ring 4 504 1 619 2 512 4 1008 Ring 5 603 2 750 4 768 6 6 720 Ring 6 727 0 910 9 512 4 1008 Ring 7 888 4 1094 1 512 4 1440 these are rings of double sided modules Table 4 2 The end caps consist of ten module types with four single sided rings and three stereo rings consisting of two different modules mounted back to back i are MA ES EE unu rt AA a 5 O A Figure 4 1 A ring 5 module left and its components right The capton cable is glued to the CF frame for electrical insulation before the silicon sensors and the FEH are glued to it 4 1 1 Mechanics Each of the 15232 modules consists of a carbon fiber CF frame onto which the silicon sensors one or two depending on the module type and the FEH will be glued Between the CF and the silicon sensors a capton cable is glued which contains the electrical lines needed to contact the sensors backside and filter capacities 10 kQ 47 pF 3 kQ Furthermore the capton will isolate the sensors backplane and its applied high voltage see Fig 4 1 The CF frames consist of two legs holding the silicon substrate s and a cross piece supporting the FEH and a pitch adapter According to the number of different module geometries there are also 14 different types of CF fra
68. F cm at depletion voltage The coupling capacity C e of the Al readout strip to the implant is specified to be 1 2pF cm um with the strip length in cm and the Al readout strip width in um These values directly determines the noise contribution of the front end electronics again compare Sect 4 3 on page 37 Hartmann 2003b 4 Silicon Strip Tracker and its End cap Modules 27 Front End Module Front End Driver FED Analog zZ lt 3 5 ADC FPGA DPM a p O lt i TTC RX Ww v c Optical links TTC Front End Controller FEC CONTROL SYSTEM uF vO Control path digital FRONT END 100m BACK END on detector 5 control room Figure 4 5 CMS Silicon Strip Tracker readout and control system Friedl 2001 4 2 Readout and Control Chain In the CMS Silicon Strip Tracker a one directional chain delivers clock and trigger to each detector and a bi directional control chain exchanges control information such as configuration parameters or temperature monitoring data between control room and front end electronics These control data are conveyed purely digital while on the other hand a one way analogue readout chain transmits the measured data from the detectors to the control room Within the experiment all signals are transmitted through about 100 m of optical fibers between front end and control room On either end the light information is converted to el
69. Fitch and R Turlay Evidence for the 2 Pi Decay of the K 2 0 Meson Phys Rev Lett 13 138 140 1964 C Civinini Studies on noisy strips CMS Tracker Module Test Meeting January 2003 http hep fi infn it CMS moduletest wg html FED Manual CMS Front End Driver PMC User Manual Version 5 0 CLRC Ratherford Appleton Laboratory 5 0 edition May 2001 CMS Collaboration The Compact Muon Solenoid Technical Proposal CERN LHCC 94 36 1994 CMS Collaboration The Electromagnetic Calorimeter Project Technical Design Report CERN LHCC 97 33 1997a CMS Collaboration The Hadron Calorimeter Project Technical Design Report CERN LHCC 97 31 1997b CMS Collaboration The Magnet Project Technical Design Report CERN LHCC 97 10 1997c CMS Collaboration The Muon Project Technical Design Report CERN LHCC 97 32 1997d CMS Collaboration The Tracker Project Technical Design Report CERN LHCC 98 6 1998 146 REFERENCES CMS Collaboration Addendum to the tracker TDR CERN LHCC 2000 016 2000 C Coldewey Test of silicon strip detectors for the ZEUS microvertex detector Nucl Instrum Meth A453 149 152 2000 M Della Negra CMS overview Eur Phys J direct C4S1 03 2002 A Dierlamm Silicon sensors in high radiation environment PhD thesis Universit t Karls ruhe TH 2003 to be published G Dirkes F Hartmann L DeMaria and M Mechini Procedures for Module Testing CERN 2002 V Drollinger et al Se
70. GUI The KaRinA applications main function parses given arguments for flags like verbose and reads the configuration files by calling the Config class The main function makes this Config class accessible to all threads by a global pointer variable and starts all threads spec ified in the configuration files Within the configuration files also test setup specific settings are stored reflecting e g the use of different slow control hardware controlled by dedicated threads Besides the Config class pointer variable additional global variables are used to share data between the threads This includes a common verbose flag and a set of pointer variables used to collect data for the data base which has to know about settings and results from all three main threads The synchronisation between all these threads especially between the communication and their corresponding main thread is a delicate task which is done by semaphores Semaphores are special counters for resources shared between threads which support two basic operations increment the counter atomically and wait until the counter is non null and decrement it atomically 6 3 4 Graphical User Interface The Graphical User Interface GUI and therefore the main part of the test systems control is implemented as a LabView program Fahrer 2003 The decision to use LabView for this Semaphores are counters for resources shared between threads The basic operations on semaphores are in
71. Higgs width increases The main background is continuum ZZ production but the signal is very clean this is considered as the golden decay mode With 100fb a signal in excess of six standard deviations is visible over the entire range 200 lt My lt 600GeV c Experimentally an excellent vertex reconstruction capability is required 2 2 3 High mass Higgs As the SM Higgs mass increases further its width and its production rate falls and one must turn to decay channels with a larger branching ratio The first of these is H gt ZZ llvv Here the signal involves a Z decaying to lepton pairs and a larger fraction of missing energy Therefore the signal appears as a peak in the missing energy spectrum There are more potentially important sources of background in this channel than in the 4 final state In addition to the irreducible background from ZZ final state there are Z jets events where the missing energy arises from neutrinos cracks or other detector effects that cause jet energies to be mismeasured 8 2 3 Supersymmetry Substantially larger event samples are available if the decay mode H gt WW lv jets and H ZZ jets can be exploited efficiently which requires enormous reduction of W jets and Z jets background by kinematical cuts which may be possible The LHC at full luminosity will be able to probe the entire range of Higgs masses from lower limit set by LEP up to the value where it is no longer
72. IEKP KA 2003 18 Entwicklung und Umsetzung von Strategien zur Qualit tssicherung von CMS Silizium Mikrostreifenspurdetektormodulen Guido Dirkes Zur Erlangung des akademischen Grades eines DOKTORS DER NATURWISSENSCHAFTEN der Fakult t f r Physik der Universit t Karlsruhe TH genehmigte DISSERTATION von Dipl Physiker Guido Dirkes aus Riesenbeck NRW Karlsruhe 30 Juni 2003 Tag der m ndlichen Pr fung 18 Juli 2003 Referent Prof Dr Th M ller Korreferent Prof Dr W de Boer IEKP KA 2003 18 Development and Implementation of Quality Control Strategies for CMS Silicon Strip Tracker Modules Zur Erlangung des akademischen Grades eines DOKTORS DER NATURWISSENSCHAFTEN der Fakult t f r Physik der Universit t Karlsruhe TH genehmigte DISSERTATION von Dipl Physiker Guido Dirkes aus Riesenbeck NRW Karlsruhe 30 Juni 2003 Tag der m ndlichen Pr fung 18 July 2003 Referent Prof Dr Th M ller Korreferent Prof Dr W de Boer Nothing is too wonderful to be true if it be consistent with the laws of nature Experiment is the best test Michael Faradays Research Notes March 19th 1849 Abstract Abstract The Large Hadron Collider will explore physics at the energy frontier It will allow to address many open questions in particle physics amongst which the following ones are of highest importance search for the Higgs boson and the study of its properties search for new
73. LED system for pinhole identification The detected pinholes must be unbonded To reduce the pinhole sensitivity of the APV25s two changes in the FEH design have been made In the first step the resistors in the HV return line are changed While the prototypes tested were equipped with a return line resistor combination of Rret 22k 2 2k resistors the final version of the FEH will use R c 2kQ 1000 This reduces the potential of the p implant so that the implant potential will match the APV25s virtual ground at leakage 7 Test System Performance and Module Qualification Studies 119 ice al o I ol N o o al o o EE ULLI Mean calibration amplitude ADC 0 2 A 4 50 100 150 200 RN 250 300 350 40 12 net P 0 Leakage current uA Figure 7 52 Pinhole induced gain loss of APV25 The response of the APV25 to a calibration pulse can be used as a measurement of corresponding the APV25 gain Here the mean response of the APV25 in Peak mode to a calibration pulse of one MIP equivalent charge is plotted which is extracted form plots like Fig 7 50 with the pinhole channels excluded For very low leakage currents the APV25 is not effected at all even with a large number of pinholes attached With leakage currents increasing over 50 1A the chip gets sensitive to the pinholes because the preamplifiers of the pinhole channels saturate with negative potential at its output causing the inverter
74. PV25s The difference between them arises only from two missing APV25s and some auxiliary components for them in the middle of the four APV25s version In principle this two chips can be added later which of cause is a delicate task The first general ASIC fault found is a pedestal error Figure 7 48 shows the pedestal of the affected FEH Obviously the two APV25s in the middle have a significantly increased pedestal value This has to be flagged as a fault although the two chips are working proper and both noise and calibration behaviour are inconspicuous 7 Test System Performance and Module Qualification Studies 117 Figure 7 49 Inverter fault of ring a 5 prototype mod Sensor channel number g 700 600 500 400 300 200 100 ule For Peak modes with D PE zz de out calibration the noise is O eee ee ee en ee AA shown With the Inverter on 3 E the third APV25 shows a clear a distortion while the inverter F off mode behaves normal Fur 3b ee ee AOE cee ER EN ther defects are visible beside E l l l l E 3 two shorted channels 2154216 A A a sg the channels 368 amp 376 have 38 a missing bond and around a e e E E channel 480 several strips are E 23 noisy The APV25 edge ef a on a do ee 3 fect is slightly visible while the DB channel number sensor edge noise effect is bold Figure 7 50 Effect of a dozen pinholes on calibration ampli g E tude The pinhole channels
75. Passeri et al 2000 This design choice will move the high edge electric field from the silicon into the much more resistant oxide layer 4 Silicon Strip Tracker and its End cap Modules 25 SiO isolation HV Diode Figure 4 2 Schematic structure of CMS silicon strip sensors Based on a n type substrate one side gets an n layer for a good Ohmic contact to the back plane aluminium while on the other side ptt strips are implanted Applying an electric field across the bulk results in collection of holes on the p implant strips while the electrons drift to the back plane The signal on the implant strips couples through a thin layer of SiOa to the Al readout strips connected to an amplifier Guard ring ba a Bias ring Bias resistors DC pads AC pads i I Pd Figure 4 3 Close up view of a tracker end cap ring 6 detector Hartmann 2003b 26 4 1 Silicon Strip Tracker Modules Sensor Design Baseline J SiO n bulk ve bias voltage Figure 4 4 Schematic of a silicon micro strip sensor Feld 2002 reducing the risk of electrical breakdown The thickness of the aluminium strips is required to be thicker than 1 2 um to reduce noise contributions due to the electrode resistance compare Sect 4 3 on page 37 An array of poly silicon re
76. The new high energy regime also offers a unique opportunity to look for the unexpected physics and new phenomena There is a true multitude with some perhaps less well motivated than others Examples always at the 1 10 TeV scale are e Possible new electroweak gauge bosons with masses below several TeV e New quark or leptons e Extra dimensions with a mass scale for a few TeV The very high cross sections and resulting event rates make the LHC a true factory for the production of particles like the top quark Finally high rate phenomena can be used to measure precisely the properties of Heavy Flavours There is even some place for B physics at low luminosity in particular sensitive studies of CP violation in the B hadron system will be carried out 2 1 Large Hadron Collider The Large Hadron Collider LHC machine is a proton proton collider that will be installed in the 27 km circumference tunnel formerly used by the Large Electron Positron collider LEP at the European Laboratory for Particle Physics CERN Hereby the LHC will extend the accessible energy range by a factor of ten compared to the highest energy collider currently op erating the Tevatron The LHC accelerator will use 1100 superconducting dipole magnets with a magnetic field of 8 4T Given the LEP circumference this implies proton beams should attain an energy of 7TeV The proton bunches in the machine are separated by 25ns with an RMS length of 75mm and intersected
77. The test procedures contain fast functionality tests checking for severe faults arising during one of the production stages reliability tests checking the long term behaviour of the module and finally qualification tests typically performed in connection with the reliability test Besides the electrical tests also the mechanical stability is tested by performing cooling cycles partially with active readout Finally an electrical stress test can be made with artifical leakage currents For the production of ring 5 modules in Karlsruhe test systems are developed and es tablished Absolute calibration of the test systems have been performed using cosmic ray particles The measured noise of the readout chain including front end electronics as well as the back end ADC is in perfect agreement with the calculated values arising from theoret ical description by the noise and signal theory Furthermore the systematic studies of the signal to noise ratio SNR confirmed the predicted module behaviour With the infrared LED system a new analysis tool for silicon strip detectors is developed The pulsed LED illumination can be used for signal generation while a constant illumination stimulates an artifical leakage current The latter reveals an unique signature for AC coupling capacitor short so called pinholes in noise as well as in calibration amplitudes Furthermore the constant illumination reveals a pinhole induced gain loss of the APV25 This le
78. UT LIBRARIES C Karlsruhe Readout libraries The readout software is split into logical units placed in individual libraries which typically contain one or several classes compare Fig 6 13 on page 78 These classes and their key functionalities are discussed on the following pages giving a more detail insight into the KaRinA substructure RALFED library For the FED the RAL has developed a collection of C routines that contain all low level routines needed to run the FED These are collected in a library called libralfed so compare Fig 6 12 on page 75 The RALFED library contains the following functions fedpmc_configure_bridge function configures the PCI bridge local registers fedpmc_config_fpga loads the firmware from the flash memory to the FPGA This operation takes a few seconds and will only be done after a power off reboot fedpmc_init function is the main function of the RALFED library It sets the sampling speed and depth the number of channels sampled the clock delay and the clock and trigger inputs to be used fedpmc_set_apv_sync_thresholds routine controls the thresholds used by the FEDs header detection which detects high and low bits of the digital header by compar ison to a low and a high threshold fedpmc_set_runmode routine changes between the header finding mode a scope mode without header finding and a software trigger mode only as scope mode possible With header finding turned on the FED delivers only the APV
79. a PMC card from CERN is chosen Finally with a MIO card slow control sensors are read out via a slow control multiplexer board 6 2 Hardware Components Based on a small set of key components produced within the CMS collaboration or industries like FED as ADC I C card and slow control multiple input output MIO the basic readout system is based on several dedicated modular components served by a motherboard see Fig 6 2 on the facing page Mainly all PCB design and layout is done with Eagle which is a light weighted design program from cadsoft Eagle On board logic is implemented with programmable logic devices PLDs from Altera mainly of the MAX7000 series Programming and simulation of the PLDs is done using the MaxPlus II package from Altera This program allows as well graphical programming of the PLDs as compiling and loading of Altera hardware description language AHDL written programs The latter option is mainly used MAX PLUS If the program s simulation shows the right timing characteristics it will be loaded into the PLD which can be done in two different ways If the PLD is soldered directly to the PCB you have to implement a joint test action group JTAG interface on the board The JTAG signals need four of the input output pins of the PLD which reduces their number available for I O operations accordingly Without the JTAG interface sockets for the PLD have to be used In this case the logic chips will be program
80. a Veo Vea Veo Vea Veo 0 0 1 0 Figure 2 3 The unitarity triangle The version on the left uses the definition given in Eq 2 8 on the preceding page while the rescaled version on the right uses the definition ofn and p Ellis 2001 The values of individual matrix elements can in principle all be determined from weak decays of the relevant quarks or in some cases from deep inelastic neutrino scattering As far as phenomenological applications are concerned the following parametrisation of the CKM matrix the Wolfenstein parametrisation Wolfenstein 1983 which corresponds to a phenomenological expansion in powers of the small quantity A Vus sindc 0 22 turns out to be very useful E 1 3 A Ad3 p in Verm ZA 4 AX 0 2 9 AM 1 p in A M 1 where A p and 1 are real numbers that were intended to be of the order of unity However in the LHC era the experimental accuracy will be so tremendous that one has to take into account the next to leading order terms of the Wolfenstein expansion Kowalewski 2003 The relations dictated by unitarity allow a convenient geometrical representation of the CKM parameters The product of any row column of the matrix times the complex conjugate of any other row column results in three complex numbers that sum to zero and can be drawn as a triangle in the complex plane There are three such independent triangles Two of the three have one side much shorter t
81. a abstract base class Motherboard for all motherboard ac cesses The pure virtual methods are intended as follows getSlot has to transfers to the motherboards PLD which slot has to be address during the next data transmissions and will block next getSlot calls compare Sec 6 2 1 on page 63 releaseSlot unblocks the motherboard which will be accessible again afterwards for the next getSlot call APPENDIX C KARLSRUHE READOUT LIBRARIES 133 sendData transmits data to the motherboard receiveData receives data from the motherboard The derived classes MotherboardI2C MotherboardDIO and Motherboard1284 override this virtual methods with their specific implementations The use of polymorphism allows applications to access the motherboard independently of the chosen concrete class by using a pointer to its abstract base class In this way the access to the hardware via different links like parallel port I C bus or DIO card is realised Sequencer library hosts everything needed to run the SEQ Therefore it contains the Se quencer class which supports the methods Sequencer constructor call taking a pointer to the Motherboard class and optional an integer giving the slot which hosts the SEQ If no slot number is given it tries a default value sendTrig sendLedTrig and sendCalib are the methods used to send a corresponding trigger signal to the connected hardware setLatency controls the number of clock cycles between the calibration pattern
82. ab have already seen their first events and although the physics potential of these experiments is very promising it may be that the definite answer in the search for new physics in B decays will be left for second generation B experiments at hadron machines Already during the low luminosity phase 102 1013 b events will be produced at the LHC Previously existing doubts regarding the possibility to reconstruct B hadrons in the very complicated topology at hadron machines have been dispelled by the Collider Detector FermiLab CDF collaboration which also demonstrated the key role of tracker and vertex systems for B physics Theoretical background In the SM with SU 2 U 1 as the gauge group of elec troweak interactions both the quarks and leptons are assigned to be left handed doublets and right handed singlets The quark mass eigenstates are not the same as the weak eigenstates and the matrix relating these bases was defined for six quarks given an explicit parametrisa tion by Kobayashi and Maskawa Kobayashi and Maskawa 1973 1973 This generalises the four quark case where the matrix is described by a single parameter the Cabibbo angle 0 Cabibbo 1963 By convention the mixing is often expressed in terms of a 3 x 3 unitary matrix V operating on the charge e 3 quark mass eigenstates d s and b d Vaa Vas Vub d d s Vea Ves Vo s Voxm s 2 8 b Via Vis Veo b b 10 2 5 B Physics pn Va Vas Via Vib Ve
83. ad to a design change of the HV return path on the FEH reducing the risk of pinhole induced gain loss Nevertheless pinhole identification is a major task of the CMS module test strategy A new characteristic of the CMS silicon sensors is the metal overhang technique used to improve the sensors high voltages resistance A draw back of this an increased sensibility of the sensors towards surface scratches is identified A surface scratch can disturb the electrical field configuration inducing effects like shorted channels high leakage currents of the cor responding channels and reduced break down voltages of the AC coupling capacitor causing a pinhole at higher leakage currents For the identification of the latter an artifical leakage current is mandatory Another effect found in Karlsruhe is a module edge channel noise In this case the prob lem arise from a common mode noise coupling into the edge channels This resulted in an additional filter capacitor added on the HV return line on FEH The quality of the prototypes showed some minor problems mainly due to noise on the module edge channels and on the APV25 border channels Applying the strict cuts from the module quality criteria definitions most of these border channels have to be tagged as noisy resulting in a grade B for the modules However these channels are functional and tests have provem that their level of noise is uncritical in terms of their expected signal to noise ratio
84. after 10 years of LHC operation Therefore these channels can be used for the final readout Beside this noisy border channels only a very limited number of defects were found on the prototypes so far and their overall behaviour is astonishingly robust From this point also the early prototypes have to be graded A which is promising for the mass production to come Finally with the developed tool for pinhole detection a stable operation of the Silicon Strip Tracker modules can be guaranteed even after 10 years LHC operation 122 APPENDIX A ACRONYMS AND ABBREVIATIONS 123 A Acronyms and Abbreviations The following acronym and abbreviations are used within this work Acronym AC ADC AHDL APD APSP APV25 ARCS ASIC BC CCU CCUM CDF CERN CES CF CORE CM CMOS CMS COTS CPLD CPU CSC CTE DO DAC DAQ DC DCU DIO DMA DNL DPM DRDC DTS DT DTBX DUT EBW ECAL Definition alternating current analogue digital converter Altera hardware description language avalanche photodiode analogue pulse shape processor see Sec 4 2 1 2 on page 29 analogue pipeline voltage chip see Sec 4 2 1 2 on page 29 Aachen readout and control system application specific integrated circuit Bonding Centre see Sec 5 4 2 on page 53 communication and control unit see Sec 4 2 4 2 on page 36 communication and control module see Sec 4 2 4 2 on page 36 Collider Detector FermiLab experiment at FermiLab European Laboratory
85. age 50 full width at half maximum gate controlled diode ground Graphical User Interface see Sec 6 3 4 on page 77 field programmable gate array Fast Test Station see Sec 6 4 on page 80 hadronic calorimeter highly ionising particle hadronic forward calorimeter high voltage 12C bus protocol International Business Machine integrated circuit interconnect board Imperial College London integral non linearity Irradiation Qualification Centre see Sec 5 2 3 on page 49 joint test action group Karlsruhe Readout is now eAsy see Sec 6 3 3 on page 77 light emitting diode see Sec 6 2 5 on page 67 Large Electron Positron collider Large Hadron Collider LHC Committee linear laser driver see Sec 4 2 2 1 on page 34 Letter of Intent least significant bit low voltage low voltage differential signal Milestone 200 see Sec 5 2 on page 47 module assembly centre see Sec 5 4 1 on page 52 multiple input output card from National Instruments as used within the Karlsruhe readout system CONTINUED ON NEXT PAGE APPENDIX A ACRONYMS AND ABBREVIATIONS 125 Acronym MIP MIS MOS MSB MSGC MSSM APVMUX NIEL NIM NRZI OEC 00 PA pdf PIN PbWO PCI PCB PIC PLD PLL PM PMC PPC PQC QTC RAL RAM RMS R amp D RPC SCL SCSI SDA SEQ SEU SM SNR SSC SST TDR TEC TID CONTINUED FORM PERVIOUS PAGE Definition minimum ionising particle metal insulated semiconductor metal oxide s
86. alised Figure 7 33 on the next page presents the normalised signals from Fig 7 32 Furthermore for tagging of a missing signal the normalised signal amplitudes are compared with the expected signal gained from fitting the individual light cones like shown in Fig 7 31 This removes false tags arising from limited overlap of adjacent LEDs like in Fig 7 33 for channel 128 106 7 6 Infrared LED Studies Sensor channel number 280 260 240 220 200 180 160 140 120 o 10 20 IR LED signal ADC 30 Figure 7 31 Signal of a single infrared LED The light cone 22 of the LED is reduced by a col 50 E g limator block to the shown size ee ad ee O E z of 15 channel at FWHM which A E is driven by the need of suf E E E E E 28 no between nelle B0 pge O Fi bouring LEDs DB channel number Sensor channel number 500 400 300 200 Figure 7 32 The envelope of A A E E the LED signals shows a large fluctuation of the signal am NN ON NN NN plitudes Hereby the sensor is scanned with the LED array and the highest signal for each channel is plotted The small peaks around channel 60 80 arises from limited overlap be tween neighbouring LEDs Al ready in this raw data three de fects can be found with a sim i 00 200 300 200 500 ple cut on the amplitude soe IR LED signal amplitude ADC Module 3020002000063830200020000638a Peak wo Cal wo Inv Mode Sensor channel numbe
87. alues Caner p values Caner 2001 2001 As expected at large 7 the resolution worsens as the tracker lever arm decreases For low pz tracks multiple scattering becomes significant and the y dependence reflects the amount of material traversed by the tracks In Fig 4 16 the spatial resolution along the beam pipe Zimp is shown as a function of pseudo rapidity for single muons The n dependence of the Zimp resolution is due to multiple scattering 4 3 1 Signal Creation and Collection The energy loss of heavy particles in matter can be described by H A BETHE and F BLOCH s formula Hagiwara et al 2002 BEN EB E je ES 6 1 per IN _ p dx A 8 2 I Tmax Equation 4 5 represents the differential energy loss per mass surface density MeV g cm where ze is the charge of the incident particle Na Z and A are Avogadro s number the atomic number and the atomic mass of the material m and r are the electron mass and its classical radius I Tmax is the maximum kinetic energy which can be imparted to a free electron in a single collision given for an incident particle of the mass M by 2mec By u A A 142 1 677 TE I is the mean excitation energy v c y 1 82 71 and y is a correction for the shielding of the particle s electric field by the atomic electrons the density effect caused by Ties 4 6 HANS ALBRECHT BETHE 1906 in Strasbourg Most of the time he worked with the Cornell University interrupt
88. amatsu Photonics K K Hamamatsu City Japan and ST Microelectronics Catania Italy The first one is contracted to produces all 320 um thick sensors and the second one the 500 um sensors Both companies send their ma terial via CERN to the CMS collaboration CERN acts as centralised control and distribution centre for all industrial companies From here the materials are distributed to the specialised test centres within the individual institutes of the collaboration see Fig 5 2 The capability of producing testing and qualifying of the silicon detectors has been checked using a staged schedule with embedded milestones The first part was the milestone Milestone 200 M200 which evaluated on a basis of 400 sensors sufficient to build 200 modules the producers process and material and verified the set ups and quality assurance procedures The second step based on a pre series production of 5 scale of the full production to verify the capability of a fast pace and adequate quality production and testing inside the collaboration as well as from the producers Based on this the full scale production has been launched in spring 2003 5 2 1 Sensor Quality Test Centre The Quality Test Centres QTCs in Karlsruhe Perugia Pisa and Vienna are responsible of the overall silicon quality and will check 100 of the sensor under the microscope while simultaneously measuring their geometrical size The cutting precision is required to be better than 2
89. amplitude in Peak mode reaches a maximum of 84 ADC and in Deconvo lution mode of 62 ADC which is a reduction of 26 In Deconvolution mode the amplitude drops well below 50 within one latency step which is equivalent to the bunch crossing time For Peak mode the signal is still at a level of 85 of the maximum The shorter rise and fall time of the Deconvolution mode allows discrimination of events from consecutive bunch crossings For a SNR estimation the calibration amplitude gives a coarse estimate which is of lim ited precision because of the 40 absolute accuracy of the calibrations amplitude Friedl 2001 Figure 7 13 on the following page shows typical behaviour of the APV25 calibration amplitudes This kind of plots is gained by the superposition of the signals from the eight calibration units compare Sec 4 2 1 2 The calibration amplitude shows fluctuations of up 96 7 4 Signal Performance _ 9 sob Figure 7 12 Calibration pulse shape in g E 4 70 Peak and Deconvolution mode While 2 OE the Peak mode signal spreads over more 5 eo o E than 5 latency steps the Deconvolution t j S aoe signal stays well below 2 latency steps at 3 t 5 25 Latency steps width 3 a e E i 3 FWHM allowing an effective discrimi E 38 a a E 23 nation of events from consecutive bunch 20E 28 E s crossings The loss in amplitude stays aa haitiana 8 below 3
90. and the mean energy loss z dE dx Leo 1994 The Landau distribution resembles a distorted normal distribution with a long upper tail due to rare but highly ionising knock on electrons The tail of an ideal Landau distribution will extend to infinite energies which is unrealistic In practice the measurement range is always limited which leads to a truncated Landau curve As a result of its asymmetry the mean energy loss is higher than the most probable Amp However the latter is much easier to obtain from measured data and therefore usually stated in experimental results The most probable loss Amp increases as function of in a first order approximation as Amp In e 0 198 6 4 8 while the ratio w Amp where w is the full width at half maximum FWHM of the energy loss distribution see Fig 4 17 decreases with increasing absorber thickness x With particle energies far below the MIP range corresponding to thick layers where the energy loss exceeds half of the original energy knock on electrons are improbable the Landau tail vanishes and thus the resulting distribution begins to approximate a Gaussian This intermediate region between thin absorber covered by the Landau theory and the Gaussian limit is treated by Symon and Vavilov compare Leo 1994 For energetic particles and photons the energy required to create an electron hole pair in silicon is J 3 6eV which is larger than the band gap because pho
91. andom noise signals that can be reduced but never eliminated since this noise is due to fundamental properties of the circuits themselves DETECTOR BIAS A Ch aaa Sea PREAMPLIFIER PULSE SHAPER BIAS R RESISTOR A SN tw Ce Rs OUTPUT DETECTOR aE Figure 4 18 Schematic silicon detector front end circuit Hagiwara et al 2002 Typical modern detector front ends follow the schematic shown in Fig 4 18 where the detector is represented by a capacitance Cq which is biased through resistor Ryoj and the signal is coupled to the preamplifier through a coupling capacitor Ce All resistances present in the input signal path like the electrode resistance any input protection network and parasitic resistances in the input transistor are represented by a series resistance Rs The preamplifier provides gain and feeds a pulse shaper which tailors the overall frequency response to optimise signal to noise ratio while limiting the duration of the signal pulse to accommodate the signal pulse rate see Sec 4 2 1 2 on page 29 The equivalent circuit for the noise analysis Fig 4 19 on the next page includes both current and voltage noise sources The leakage current of a semiconductor detector for 42 4 3 Expected Module Performance DETECTOR BIAS SERIES AMPLIFIER RESISTOR RESISTOR PULSE SHAPER Rs ens Beer Ca 3 8 inb ina O Ind l Figure 4 19 Equivalent circuit for noise analysis Hagiwara et al 2002
92. arching for Higgs Bosons in Association with Top Quark Pairs in the H bb Decay Mode CMS Note 054 2001 Eagle Cadsoft web page http www cadsoft de 2003 N Ellis B physics in the LHC experiments Nucl Phys Proc Suppl 93 317 323 2001 P F Ermolov et al A technique for testing the Si tracker modules for the DO collider ex periment FNAL and comparative analysis of the test results Instrum Exp Tech 45 183 193 2002 M B Fahrer 2003 private communication Fairchild Inc SPT7861 10 bit 40 MHz 160 mW A D converter CLRC Ratherford Appleton Laboratory 2001 L Feld Halbleiter Spurdetektoren f r den LHC Physikalisches Kolloquium Universit t Freiburg May 2002 http sct physik uni freiburg de atlas sct talks habil_final ppt L Fiore The role of automation in the construction of the CMS silicon strip detector Nucl Instrum Meth A473 39 43 2001 M Friedl The CMS Silicon Strip Tracker and its Electronic Readout PhD thesis Vienna University of Technology 2001 P Giacomelli The CMS muon detector Nucl Instrum Meth A478 147 152 2002 K Hagiwara et al Review of particle physics Phys Rev D66 010001 2002 http pdg lbl gov G Hall Front end electronics for silicon tracking at LHC IEEE Trans Nucl Sci 41 1086 1090 1994 G Hall et al LHC front end electronics NIM A 453 353 364 2000 F Hartmann Entwicklungsarbeit am Spurendetektor fiir das CDF Experiment am Tevatro
93. ard while the 12C card utilises a PCI9050 bridge PLX 1999 to connect its internal 8 bit bus to the PCI bus on the Midas20 board Based on the documentation of these bridges and the Lynx operation system a device driver was developed The device driver basically has to initialise both bridges correctly and has to ensure that all memory needed by the cards is allocated and mapped correctly With this initialisation given the access to the card functionality is simply realised by writing data and control sequences to the according memory either mapped to the internal registers of the bridges or to the I C cards bus 6 3 1 Device Driver To address new kind of devices like the FED or I C card within a Unix system the operating systems kernel has to know how to address the new devices This is done by device drivers representing a standardised interface to address all kind of devices Furthermore most Unix Trademark of National Inc Tecc GNU project c and c compile Copyriht by the Free Software Foundation Inc edb The GNU Debugger copyright by the Free Software Foundation Inc 6 Karlsruhe Test Stations 75 Socket Connections Shared Memory Figure 6 12 Basic design of the Karlsruhe readout system software Based on Linux as operating system device drivers are developed 12C or adapted FED MIO DIO and basic functionalities are placed in shared libraries The different tasks needed to readout module and contr
94. at each LED trigger pulse singleShot selects a LED Therefore it takes the array number the switch number and the number of the LED as argument compare Sec 6 2 5 on page 67 The order of the LEDs goes from switch led number 0 0 over 0 1 0 7 to 7 7 lvdsDisable disables the LED array specified by its argument The array stays disabled till a following singleShot call selects a LED from it The parallel port specification IEEE1284 motivates the chosen name 134 APPENDIX C KARLSRUHE READOUT LIBRARIES function code connect to motherboards negative power line 0 connect to motherboards positive power line 1 connect to peltiers negative power line 2 connect to peltiers positive power line 3 Load battery 5 Disconnect battery 7 Table C 1 PowerPack modes constIlluminationSensorl and constIlluminationSensor2 control the DACs gen erating the bias voltages for the LEDs and steers so the constant illumination Power Pack library contains the PowerPack class which works internally very similar to the Led Array library due to the fact that both use the same kind of control card Nevertheless the supported methods are different powerSwitch turns the power pack on or off selectMode connects a given battery number 2nd argument to a given function 3rd argument see Tab C 1 for function codes batteryOff is a short cut for the select code call with the disconnect battery function argument PeltierControl library bases again on the sam
95. at four points where experiments are placed Two of these are high luminosity regions housings the A Toroidal LHC Apparatus ATLAS and the CMS detectors The other regions house the A Large Ion Collider Experiment ALICE detector to be used for the study of heavy ion collisions and LHC B LHC B a detector optimised for the study of b flavoured hadrons The beams cross at an angle of 200 urad During the first year LHC will operate as proton proton collider at a centre of mass energy of vs 14TeV with low luminosity 10 cm s7 which subsequently will be increased to the design value of 10 4cm s During the low luminosity phase the large non diffractive inelastic cross section of about 70 mb will already result in an average of 18 minimum bias interactions per bunch crossing in 25ns time intervals The interesting weak physics signals are buried in this enormous background and will have to be disentangled by selective and hierarchical trigger system The weak signatures of new physics can show up in a number of sometimes complex final states of leptons jets and missing energy This 4 2 2 Standard Model Higgs puts extreme requirements on the performance of detectors they must have good particle identification high count rate capability good energy momentum and angular resolutions for charged leptons jets photons and missing transverse energy which especially holds true for the central tracking detectors
96. ation in B decays is BP J pK In this case the decay is dominated by the tree level diagram with internal W cs emission leading to bd TW d c Sd This is not a flavour neutral state at the quark level but becomes so at the hadron level through K mixing For this decay one finds Vin Via Ves Veb Vea Ves Afep ner L Bay 2 19 ER VIVA y mixing VcsV decay VedV x mixing Im Af ner sin2 2 20 with very little theoretical uncertainty Kowalewski 2003 The CP eigenvalue nep is 1 for Bo J bK it is 1 This decay mode is also favourable experimentally The product branching fraction B B J YKY B J Yp gt 7 5 x 107 is well within the reach of B factories and the final state includes a lepton pair enabling excellent background suppression The experimental determination of sin 28 involves three key elements e The reconstruction of the CP eigenstate J b Kg This requires good momentum and energy resolution for charged particles and photons e The determination of the b quark flavour of the recoiling B meson at the time of its decay which requires good particle identification in order to cleanly identify the charged Kaons and leptons that are used to infer the b quark flavour e The determination of the difference between the decay times of the B decay to the CP eigenstate and the recoiling B requiring excellent vertex resolution to extract the spatial distance b
97. ation of the signal is expected as long as the poly resisitor is not completely shorted For the case of a broken poly resistor Rpoly 00 the p implant is floating In this case the capacitance of the readout strip will drop with similar effects to the noise and the calibration amplitude like for a broken bond 5 5 3 7 Noisy Channels Although we have a set of test methods which allows to check for several defect origins we still find channels which simply behave noisy and where the noise source itself can not be identified For example bad poly resistors are easy to tag within the QTC test procedures but they are hard to identify within the module tests later We will find a noisy strip but it can also be caused by the APV25s amplifier The noise reference for a strip is given by the mean and RMS value o for the whole module and all channels with more than 50 above the mean are flagged as noisy 5 6 Module Quality Grades The module quality grade are based on the number of bad channels taking the assumption that no ASIC error was found and that the leakage current is within the specifications given below Using the default APV25 settings given in Jones 2001 a bad bad channel is defined by a percentage cut for pedestal noise and calibration amplitude values which all have to stay with a range of 20 around their mean values These percentage thresholds are given 60 5 6 Module Quality Grades in the Procedures for Module T
98. automatic test systems for module qualification are developed and test strategies are worked out For the electrical tests a complete readout system is developed based on readout mod ules available within the collaboration and extended by home build modules These are based on a modular approach with less complex functional units attached to a motherboard and includes key functionalities like clock and trigger generation and their distribution high and low voltage supply and test signal generation usable with lasers or infrared LEDs The motherboard is connected to a standard PC hosting a fast ADC interface cards to the motherboard and the front end electronics Already during the R amp D phase of this readout system first prototype tests were performed and some weak points of the design were uncovered resulting in changes of the electronics design of the front end hybrids Two test stations are built The first one focuses on a fast functionality test which includes an active thermal cycle with readout at 10 C performed for each individual module The other test station focuses on debugging and repair requirements It disposes of sufficient space for a flexible use of the system including the possibility of additional test options with lasers radioactive sources probes and LEDs For quality control measurements at module level it turned out that LEDs are of good use Besides external signal generation by running them in a pulsed way the
99. bond quality by producing bond samples for pull tests The pull force has to be at least 8g After bonding the modules are imme diately pre qualified concerning the pro cedures given by the module test working group Dirkes et al 2002 A first check for IV behaviour pedestals noise and calibra tion pulse responses with and without in creased leakage current is done Addition ally the power consumption is monitored and the DCU PLL and APVMUX func tionalities are tested Furthermore a thermo cycle is applied to the module with active readout during the cycle if possible The thermal cycles with readout are also declared as Figure 5 10 Karlsruhe automatic bonder from Hesse amp Knips 54 5 5 Module Error Type Detection active thermo cycles while passive ones are without readout during the cycle Independent of passive or active thermo cycle the tests will be repeated afterwards The mechanical stress induced by the thermo cycle is supposed to identify weak bonds which have to be repaired This immediate pre qualification ensures that the bonding procedures are performed with out damaging the modules where again great risks endanger the silicon as the ultrasonic welding may cause damage A full qualification pipeline based will be done on a sample base during a long term test of the modules Therefore a 72h test with active cooling cycles will be done During the commissioning phase all modules will go through this long
100. ccess permissions to the hardware are simply controlled by the permissions given to the corresponding device file 6 3 1 1 I C Driver For the I C card a device driver is developed implementing the con trol sequences needed for the Philips PCF8584 chip Philips 1997 The basic skeleton for addressing the PCI interface is available within all Linux kernels the initialisation of the PCI9050 bridge is well documented thus addressing and configuration of the PCF8584 chip on the I C card turns out to be straight forward Attention has to be paid only for byte ordering which turns out to be different for PC than on most high level platforms like the CES RIO system 6 3 1 2 FED Driver The byte ordering turned out to be one of the bigger changes while porting the FED driver from Lynx to Linux Fortunately a Linux system knows if it is running on a little or big endian platform Within a device driver this can be checked by including lt asm byteorder h gt and checking whether __BIG_ENDIAN or __LITTLE_ENDIAN is defined Additionally access to the FED is implemented in a way unusual for device drivers Dur ing the development of the FED device the engineers at RAL used the memory mapping functionalities of the PCI bus to get access to the internal structure of the FED and placed the code needed to run the FED into a library in user space The key element hereby is the mapping of the physical memory address space attached with the FED hardware and whic
101. controls three different voltages typically used for biasing the LED array which produces a constant illumination and for pulse heights in the case of pulsed operation of a single LED Finally it can switch off the connected LED arrays completely reducing the head load close to the modules Weiler 2002a 6 2 5 Infrared LED System Starting from the idea to use fast infrared LEDs as a replacement of a much more expensive and complicated laser system it was realised that an infrared LED system can also be used to simulate effects of increased leakage currents as they will occur for irradiated sensors by means of constant illumination The system developed consists of an external control card see Fig 6 6 hosted by the motherboard and up to four LED arrays see Fig 6 7 on the following page placed directly on top of the module The separation of control card and front end electronics allows to turn off the front end completely which prevents noise effects probably induced by LED systems electronics Furthermore this reduces the heat load during cooling cycles The control card hosts the obligatory PLD used for communication and can run up to six LED arrays The PLD controls three 8 bit DACs which set the voltages applied for the constant illumination and the pulse height While the pulse hight voltage is shared by all LED arrays two independent voltages for the constant illumination can be used to disentangle the effects on modules with tw
102. crement the counter atomically and wait until the counter is non null and decrement it atomically 78 6 3 Software Layout Ihomelgdi CMS readouk poerver dirke Datei Ansicht Erweitert Archiv Einstellungen Hilfe DU File Name e Status homergdi CMS readout Baltera contig labview layouts Bsc e ANALYSIS APY A COMMUNICATION SDB 2DCU A Drivers BFED BHIGHVOLTAGE A120 LED a MAIN AMOTHERBOARD MULTIPLEXER AMUXPLL SOPTICALHYBRID APELTIERGONTROL 2PLDSEO RPOWERPACK 2 RAL_FED READOUT QREPEATCTAL ASEO BSLOWCONTROL ASTEPPERMOTOR Figure 6 13 Snapshot of the CVS tree The top level hosts directories for PCB layouts and their corresponding ALTERA pro grams layout amp altera folders as well as for the LabView GUl and for configuration files labview amp config folder The readout soft ware is stored in the src directory for which the subdirectory struc ture is shown Each of its subdi rectories contains a library which are described in the surround sub section The other top level di rectories have also a substructure e g corresponding to the different boards for the layout and altera directory h h h i p 4 part is based on its support of external measurement units like oscilloscopes pulser or source measurement units as well as on its excellent graphics support Furthermore the flo
103. cussed and their typical behaviour is shown Hereby four different settings of the bias voltage are studied The modules which has a depletion voltage of 80 V ring 6 is at 400 V significantly over depleted and valid qualification data can be taken at 20 V the field configuration at the p implant is formed but the module is still far from being fully depleted without any bias voltage applied the capacitive load of the readout channels is significantly increased increasing the sensibility for channel defects finally the behaviour under forward bias 9 V is compared While only the data taken under full depletion can be taken as serious qualification data in terms of reproducibility and reliability the other bias schemes can be useful identifying faults 7 7 1 Shorted Strips Shorted strips are a common defect type which occur on the FEH and PA as well as on the silicon sensors For the latter they can be produced by surface scratches like shown in Fig 7 34 The module taken for Fig 7 34 has additional faults and therefore it is not possible to deduce the influence of these shorts on the leakage current The expected shorted strips signature is discussed in Sec 5 5 3 2 on page 57 In general they can easily be detected using the APV25s internal calibration unit Shorted channels will generally show an decreased calibration amplitude like it is shown in Fig 7 35 on the following page Beside the drop in the calibration amplitude which is
104. cy Sensor channel number 380 360 340 320 _ 300 ee a ge je 2 E i Peak Mode 2 ln SE er eee kr i i i 6 120 140 TT 180 200 5E PS 38 E 38 3H 2s E 28 2 Es gs E a Me Se E x 200 33 DB channel number z Figure 7 39 Missing bond signature in noise and calibra tion for Peak wo Inv Mode taken with 400 V bias voltage The APV25 border between channels 255 and 256 is clearly visible in both plots compare Fig 7 13 on page 96 for the calibration amplitude The missing bond at channel 264 results in a higher calibration amplitude upper plot due to the missing capacitance and to an increased noise because of the different CM behaviour Figure 7 40 signature in calibration pulse shape The channel 264 which is disconnected between the Missing bond two sensors has a faster rise and fall time and an increased peak value which both reflects a lower capacitive load of the preamplifier Figure 7 41 Missing bond signature in noise at a bias voltage Vbias 0 V The over all noise of a non depleted sen sor is significantly higher due to its increased capacitance The missing bond gives a clear tag for both Peak upper plot and Deconvolution lower plot modes independent for the in verter used or not 112 7 7 Module Fault Detection Studies Sensor channel number 500 400 300 200 5 T gt T T T E T T I Y T 8 9 E Figure 7 42 Missing bond 425 flags arising
105. de and start with quality control on sensors at QTC followed by module assembly bonding and testing to the final integration of modules into larger structures and finally via sub detector integration into the tracker assembly Over 20 different institutes and companies are involved in the production chain Della Negra 2002 All IBM wafers are delivered first to CERN There are two sets of CMS tracker wafers those for ASICs on the detector modules which constitute the largest number of wafers and whose production must at least match the schedule for module assembly and remaining ASICs mainly for the opto electronic and control systems whose delivery schedule is slightly less critical The second series of ASICs are grouped together in a multi chip wafer design and are the responsibility of CERN for design and test The remaining CMS tracker wafers which contain APV25s APVMUX and PLL chips plus standard CMS test structures will be delivered by CERN to the United Kingdom for testing at Rutherford Appleton Labora tory RAL and Imperial College London ICL Following probe station testing the wafers will be shipped in groups to a designated company for cutting and packing into carriers for transportation and then good chips will be sent to the CMS group in Strasbourg who have taken responsibility for FEH assembly 5 Quality Control at CMS Tracker Modules 47 5 2 Sensor Quality Test Strategy CMS uses two producers for silicon sensors Ham
106. dung 1 Compact Muon Solenoid Detektordimensionen und Layout Della Negra 2002 3 Der Spurdetektor Der zentrale Silizium Spurdetektor welcher im Inneren der Kalorimeter installiert wird besteht aus einem inneren Pixel Spurdetektor der vom Silizium Mikrostreifenspur detektor SST siehe Abb 2 umschlossen ist Dieser wiederum ist unterteilt in einen Zylinderbereich und zwei Endkappen welche den Kollisionspunkt mit 10 bzw 9 Lagen umschlie en 3 1 Die Detektormodule Die Gesamtfl che welche von den 15232 SST Detektormodulen aufgespannt wird umfa t 210 m was erst durch die Verwendung von 6 Wafertechnologie m glich wird Die n tigen Signal zu Rausch Verh ltnisse werden erhalten indem Auslesechips mit ex trem geringen Rauschanteilen entwickelt wurden Weiterhin werden in den u eren SST Bereichen in denen die Streifenl ngen bis zu 20 6 cm erreichen Sensoren mit einer Dicke von 500 um verbaut w hrend die inneren Lagen 320 um starke Sensoren verwenden Um den enormen Strahlenbelastungen von bis zu 1 6 x 1014 1 MeV equivalent n cm des 10 j hrigen LHC Betriebes standhalten zu k nnen wird der SST bei 10 C betrieben Dies reduziert sowohl die Dunkelstr me der Sensoren als auch das Reverse Annealing welches hilft die Verarmungsspannung der Sensoren nach der Bestrahlung klein zu hal ten Aus diesem Grund werden auch nur Sensoren mit einer lt 100 gt Kristallstruktur und hoher Resistivit t verwendet In d
107. dures and implies a lot of additional handling due to testing which also arises serious risks Reliability and performance tests play a particular role during the first phase of the pro duction where production and test procedures have to be established and inherent sources of faults have to be identified and eliminated Later these reliability and performance test can be strongly reduced and will typically be performed at the end of the production chain Furthermore present manpower time schedule and existing equipment have to be taken into account All three kinds of quality tests are performed during the production and construction phase of the CMS tracker and an extensive quality control program has been launched by the collaboration Similar approaches have already been chosen by collaborations like CDF DO DO Ermolov et al 2002 5 1 Silicon Strip Tracker Production Scheme The general structure of the SST production scheme is shown in Fig 5 1 on the following page Starting point for all components used inside of CMS is the CERN which acts as control and distribution centre for external industrial companies But nevertheless some of the components like FEH pitch adapters or frames are in the responsibility of institutes outside of CERN Missing in Fig 5 1 on the next page is the line the ASICs go before their integration e g into FE Hs Access to the 0 25 ym CMOS process which is to be used for CMS Tracker ASICs is based
108. e 134 Sensor and database numbering scheme nen 136 REFERENCES 145 References MAX PLUS MAX PLUS II Getting started Altera 2002 http www altera com M Apollonio et al Test results from a prototype lead tungstate crystal calorimeter with vacuum phototriode readout for the CMS experiment Nucl Instrum Meth A484 287 298 2002 S Asai Supersymmetry at LHC Eur Phys J direct C4S1 17 2002 M Axer et al A Test Setup for Quality Assurence of Front End Hybrids CMS Note 2001 046 2001 F Bei el ARC_FE III Physikalisches Institut RWTH Aachen 2001 T Bergauer et al Process control strategy of the silicon sensors production for the CMS tracker NIM A 494 218 222 2002 N Bingefors et al A Novel technique for fast pulse shaping using a slow amplifier at LHC Nucl Instrum Meth A326 112 119 1993 S Braibant et al Investigation of design parameters and choice of substrate resistivity and crystal orientation for the CMS mircostrip detectors CMS Note 11 2000 N Cabibbo Unitary Symmetry and Leptonic Decays Phys Rev Lett 10 531 532 1963 A Caner Performance of the all silicon CMS tracker NIM A 462 270 277 2001 P Cederqvist et al Version Management with CVS 2002 http www cvshome org G Cervelli A Marchioro P Moreira and F Vasey A Radiation Tolerant Laser Driver Array for Optical Transmission in the LHC Experiments LEB 2001 J H Christenson J W Cronin V L
109. e 40 08 MHz LHC clock and a trigger line Table 4 3 shows patterns transmitted via this line and the corresponding meanings The digital functionality of the chip can be compromised patteri Zen by single event upsets SEUs Effects are also expected in the 100 trigger oe E 3 q analogue circuit but they cannot cause logic errors which af 110 calibration 2 fect chip operation The areas which can be upset consist of 101 soft reset pipeline control logic I C registers first in first out FIFO memory which stores addresses of pipeline columns awaiting Table 4 3 Symbols trans readout and the main control logic block which handles exter mitted over the trigger line nal communication and controls readout sequencing Jones 2001 When a trigger occurs the pipeline cell at the current loca tion of the trigger pointer is reserved for readout and its address is loaded into the FIFO The write pointer will subsequently skip over columns marked for readout until they have been transmitted The FIFO has a depth of 31 cells and further address will cause a FIFO error signaled via the error bit in the data output header The fact that Deconvolution and Multi mode both need three consecutive sample restricts the number of stored trigger to 10 For Peak mode only one address is stored per trigger and thus the limit is 31 This and the time needed for data transmission data frame length limits the L1 trigger frequency 4 2 1 3 APVMUX Savi
110. e 5 3 Picture of a standard test structure From left to right the TS Cap Sheet GCD Cap TS AC mini sensor Cap TS DC Diode and two MOS structures can be seen for details see text Bergauer et al 2002 Figure 5 4 Proton irradiation beam line in the Forschungszentrum Karlsruhe with thermal insulation box and plot of the ac tivity of a scanned Ni foil by autoradiog raphy A piece was cut out for dosimetry white area The inner frame marks the important sensor area the outer dashed frame marks the scanned area which is more than one beam radius larger than the sensor area itself Dierlamm 2003 level The last three resistances are made of aluminium with the same design of the p implanted strips to extract the metal lines resistivity The most left structure TS Cap is made of 26 AC coupled strips without any poly silicon resistor used to test the coupling capacitances and coupling insulator breakdown applying up to 200 V across the insulator Bergauer et al 2002 5 2 3 Irradiation Qualification Centre To verify the radiation hardness of the silicon detectors during production the Irradiation Qualification Centres IQCs in Louvain and Karlsruhe irradiate 1 of sensors and 5 of stan dard test structures The Quality Test Centre QTC in Louvain will check for bulk damages using neutron irradiation while the QTC in Karlsruhe see Figs 5 4 and 5 5 on the following page will use proton irradiation to study bulk and addit
111. e 8 F es 5 P A ARA 3 e 3 Rae 3 4 S phate ES x BE AA g E gps H A 30 i A F 2 20f 10 ar yi EA Ki 0 100 200 300 400 500 0 100 200 300 400 500 voltage V voltage V Figure 7 10 Module Leakage current versus bias voltage at 10 C left and 22 C right The temperature is measured a the modules transport frame close to the FEH which is the main heat source of the module Therefore the silicons temperature will be much closer to the temperature of the heat sink which is typically 5 to 7 C cooler more or less irrelevant since the calibration modes are only used for testing not for data taking 7 3 Module Leakage Current The module leakage current Ijeg is an important qualification value Therefore scans of the module leakage current as function of the bias voltage current voltage characteristic IV curves are performed For a given device the leakage current Ij ar is at first order a function of the thickness of the depletion zone xq and of the intrinsic charge carrier concentration ni The depletion depth xg of the silicon scales with the square root of the applied bias voltage Voias B 2 till the full depletion of the device is reached Therefore the modules leakage current leak is expected scale with the square root of the bias voltage until the depletion voltage is reached The intrinsic charge carrier concentration n depends strongly on temperature and the leakage current will be strong
112. e Boltzmann constant and T the temperature Typical amplifier noise parameters ena and ing are of order nV V Hz and pA y Hz Besides thermal and shot noise a third kind called Flicker noise is found in all active devices as well as in carbon resistors but it occurs only if a direct current DC is flow ing Flicker noise usually arises due to traps in the semiconductors where charge carriers that would normally constitute DC current flow are held for some time period and than released Flicker noise is commonly refered as pink noise or 1 f noise since its spectral power densities is well modelled as A ee ae 4 10 where is in the order of 0 8 to 1 3 and with the device specific noise coefficient Ay which is of the order 10 10 10712 V2 A fraction of the noise current flows through the detector capacitance resulting in a frequency dependent noise voltage 2 wC which is added to the noise voltage in the input circuit Since the individual noise contributions are random and uncorrelated they add in quadrature The total noise at the output of the pulse shaper is obtained by integrating over the full bandwidth of the system Superimposed on repetitive detector signal pulses of constant magnitude purely random noise produces a Gaussian signal distribution Shot noise is like rain on a tin roof Horowitz and Hill 1989 4 Silicon Strip Tracker and its End cap Modules 43 Since radiation detectors typically conve
113. e Plate Cathode Strip Chambers esc Chambers DT Chambers RPC Resistive Plate Chambers RPC Figure 3 1 Compact Muon Solenoid detector dimensions and layout Della Negra 2002 3 3 Central Tracker The innermost part of the CMS detector is the Central Tracker which is divided into the Pixel and Strip Tracker see Fig 3 2 Initially it was planed to use MSGCs for the outer part of the Strip Tracker but in 2000 the design baseline was changed to an all silicon solution although the MSGCs had passed all milestones and showed their viability of stable track ing within a LHC experiment Main argument towards an all silicon solu tions was the chance to concentrate all available effort to a common type of detectors the reduced price for sil icon as well as an easier replacement of detector elements or upgrade The Central Tracker is described by detail in Chapter 4 3 3 1 Pixel Vertex Detector Figure 3 2 Central Tracker dimensions and layout The overall length is 5 4m Hartmann 2002 The most interesting events at the LHC are likely to contain several b jets originating from the decay of heavy particles In order to allow efficient tagging of these jets as well as of other objects c r the tracking must extend most closely towards the reaction vertex The high 3 Compact Muon Solenoid 17 Figure 3 4 Construction principle of a pixel detector with bump bonded readout electronics CMS Collabora tion 19
114. e Va mode only a a e Vint Vem Yom Vem Vom gt Vour gt gt gt Figure 4 10 Small signal model of the APV25 CM suppression Raymond 2001b From the small signal model compare Fig 4 10 results for the sum currents into node vp UR R Im Vin VOM UR 127 gm vcm UR 4 1 UR Vin 128 VCM ImR gt Vin 128 vVCM ImR In UCM Y VCM 4 2 1 1289m R 1289m R 128 with the amplification factor gm While the currents down the left hand branch of Fig 4 10 give 9m vouT Im VIN VOM UR 4 3 But if vr vom then for the output voltage results VOUT VIN 4 4 which shows an effective CM suppression Raymond 2001b Inverter stage With two mutual switches either the direct or the inverted output of the APV25 preamplifier can be sent to the shaper stage The unity gain inverter that follows the preamplifier if shown in Fig 4 9 While the direct output is intended for use with n bulk detectors the inverted output is intended for p bulk detectors which produce current pulses of opposite polarity Since the dynamic range of the shaper is limited its working point is not centered between the supply voltages The optional inverter between preamplifier and shaper thus allows linear operation with input signals of either polarity 32 4 2 Readout and Control Chain Calibration unit Another useful feature of the APV25 is its internal calibration circuit
115. e kind of control card like the LedArray and is designed to share one together with the PowerPack class This is realised on the same way like the control of several LED arrays This library contains the Peltier Control class which drives the Power Packs peltier switch network It contains the methods allOff allSeriell seriellSeriell parallelSeriell and allParallel which re sults in an increasing cooling power applied to the peltier elements Furthermore the method APVCooling powers only the peltier elements connected to the FEH com pare Sec 6 4 on page 80 For safety reasons the cooling power is reset each 30sec by the Power Pack which forces the slow control software to check and adjust the cooling power with the corresponding frequency Data base library contains a collection of classes used to store the results of a module qualification to a root file which is used as local data base This root file can be parsed to generate a xml file for the central production data base The basic layout of the local data base file is a common development of the CMS collaboration which has been extended to additional needs of the Karlsruhe test system Readout library collects the functionalities needed to readout a FEH This includes all the ASICs related libraries for the APV25s the PLL the APVMUX and the DCU as well as the libraries needed to run the readout hardware like the FED and SEQ Furthermore 1t receives and sorts the data coming from the
116. e motherboard and allows reading and writing sequences to the RAM and the sending of the different trigger sequences The other PLD controls the feedback loop and enables external trigger input It also blocks read write access via a busy flag during active sequences The SEQ has five output lines of which one can be configured as external input line One output line is fixed to LVDS levels while the others are connected with converter PCBs to the output connector The PLD transmits and accepts only transistor transistor logic TTL levels while outside electronics may use a variety of different standards Transmission over longer lines will typically be done as LVDS signals while external nuclear instrumentation module NIM logics uses NIM signals The converter PCBs allow simple signal adaptation 6 Karlsruhe Test Stations 65 40MHz Clock Seq lt APV LVDS TTL Cik1 sequences lt 7 RAM 8 Bit Data g Pie Se PLD PLD 2 FED LVDS TTL D Clk2 MT55L128L18F1 cle Saa 3 a a s 2 Seq3 gles Strobe LED Array LVDS TTL Clk3 o ALTERA RAO 15 ALTERA gt gt MAXTOS4AETCI00 rs lt i AB MAX7OS4AETCIOO Acknowiedge lt 5 rn Ext Trigger gt NIM gt TTL switch A QS3390 gd Select AO A15 Figure 6 4 Karlsruhe sequencer card picture and block diagram Like all cards of the Karls ruhe s
117. e readout path the analogue data coming from two analogue pipeline voltage chip APV25 front end amplifiers are multiplexed onto a single line by the multiplexer APVMUX chip and sent over the analogue optical link Hall et al 2000 The data is then digitised and pre processed by the front end driver FED 4 2 1 Front End Hybrid The detector module front end hybrid FEH supports several ASICs needed to operate and control the module together with power supply decoupling capacitors In detail discussed below are the APV25 the APVMUX the PLL and DCU ASICs The physical layout of the FEH enables efficient transfer of heat between the tracker cooling system and the front end electronics of which the APV25 dissipates the largest fraction 4 2 1 1 PLL Inthe CMS central tracker system the LHC reference clock 40 08 MHz and the first level trigger decisions L1 are transmitted from the counting house to the detector using a single optical fiber To achieve this both the clock and trigger signals are encoded as a single signal as schematically illustrated in Fig 4 6 The coded signal is identical to the LHC clock except when a trigger decision is signaled Then the coded signal stays at logic level 0 for the duration of a LHC clock cycle I i LHC clock 40 08MHz 1 i i i 1 1 1 ji First Level Trigger i 1 Clock amp Trigger Figure 4 6 Combined cod f i i 4 ji i i
118. e signals from the two photomultipliers PMs are processed by standard NIM logic modules After two independent constant fraction discriminators the 98 7 4 Signal Performance 700 Entries 600 500 400 300 200 100 Peak wo Inv Mode wo Cal Module 30216630300027 A 0 5 128 5 256 5 384 5 512 5 Channel Figure 7 16 Hit map of Sr detection shows a clear peak indicating the source position coincidence of the two PM signals is checked and the resulting signal is send to the SEQ The signal output of the coincidence unit is not processed further thus any coincidence detected during a clock cycle will cause a readout trigger after the given latency adapting signal propagation and processing times Thus the net timing jitter of the trigger concerning the particle crossing is 25 ns The effect of timing jitter can be estimated using the pulse shapes as shown in Fig 7 12 on page 96 Taking the average amplitude of 0 5 latency steps around the maximum in com parison to the peak value will give the net reduction due to the timing jitter While this reduction stays small for the Peak mode less than 1 5 the effect in Deconvolution mode is 5 times larger and leads to an average signal loss of 7 5 The effect even gets worse if the timing jitter interval is not centred around the peak A displacement by 5ns will increase a reduction to 2 for Peak mode and to 18 for Deconvolution While for the calculations of SNR valu
119. e to the CM suppression of the APV25 compare Sec 4 2 1 2 on page 30 the missing bond shows up as noisy channel Figure 7 40 on the next page shows the effect of a missing bond in the calibration pulse shape Due to the decreased capacitive load the calibration pulse has a faster rise and fall time as well as an increased peak value A missing bond can also be easily detected in the noise for non depleted modules which have a larger capacitive load resulting in higher noise for all channels connected to the silicon see Fig 7 41 on the facing page Beside the plain missing bond information its position is needed for a repair Although a missing bond can usually be identified optically there are also missing bonds that are optically inconspicuous Therefore the test systems have to be able to deduce the missing 111 7 Test System Performance and Module Qualification Studies o o T npp p pn mpm Calibration amplitude ADC a a o a Noise ADC N Cal Signal ADC counts Noise ADC Noise ADC Fa up no ua aa DD yn oom Gana oo a 250 m i Sensor channel number 240 235 Peak wo Inv Mode w Cal Module 30200020000638 DB channel number 3 88 o 8 a i A iZ ly 128 129 130 131 132 133 88 Laten
120. ecays they are of particular interest for hadron machines like the Tevatron at low statistics possible and LHC Wilkinson 2000 Experimentally the By J w channel has a clean signature and very big samples will be available The SM prediction for the CP asymmetry in this channel is very small A 2X n O few where A and y are the Wolfenstein parameters of the CKM matrix A large observed asymmetry would therefore be a sign of new physics Good proper time resolution is essential given the rapid oscillations of the asymmetry Ellis 2001 2 6 Heavy Ion Collisions Heavy ion beams at LHC will provide collision energy densities well above the threshold for formation of quark gluon plasma QGP ALICE as a dedicated experiment for heavy ion physics is part of the LHC program and ATLAS as well as CMS have good capabilities for a heavy ion program Wrochna 2002 3 Compact Muon Solenoid 15 3 Compact Muon Solenoid The concept of a compact detector for the LHC based on a solenoid the CMS detector was presented for the first time during the LHC Workshop in Aachen in October 1990 Basic ideas are an optimised muon detection system reached through a very strong magnetic field and a compact detector design including tracking and calorimetry The choice of a superconducting solenoid generating a magnetic field of 4 Tesla strength is a pre requisite for good momentum resolution even for high momentum muon tracks 1 TeV up to pseudo ra
121. ectrical signals and vice versa see Fig 4 5 CMS decided to use an analogue readout system for several reasons The first argument is the possible gain in overall performance The material budget is strongly coupled to the front end electronics The number of needed connections as well as the power dissipation influence the cooling requirements Lower power simply results in less material and services needed The second argument towards an analogue readout is an expected gain in reliability using an analogue system which gives direct access to pulse height informations Experience from former experiments shows that systems are easier to debug and operate if pulse height information are available 28 4 2 Readout and Control Chain Local receiver boards TTCrx provide the signals from the timing trigger control TTC system for the modules The front end controller FEC takes this information adds specific control signals and sends these data to the front end control module using a digital optical link The communication and control unit CCU interprets the received information and passes it to the front end electronics Temperature voltages and currents of the detector module are monitored by the detector control unit DCU The CCUs collect these data and transmit them back to the FEC To synchronise all distributed front end electronics special phase locked loop PLL delay chips are used to adjusting clock phase and trigger delay In th
122. ed Within this reliability test several active thermo cycles will be performed 5 5 Module Error Type Detection Section 5 4 2 on the preceding page already addressed the detection of module faults A lot of experience has already been gained in former experiments how to detect all types of different errors that may occur during a large production 5 5 1 Non electrical Errors Although most of the module qualification is based on electrical test we also perform non electrical tests like optical inspections and measurements on the mechanical and thermal stress behaviour 5 5 2 General ASICs Error Detection All error types of the general ASICs errors have to be stated as serious Typically these defects affect the functionality of the module in general and thus imply that the module will be unusable 5 Quality Control at CMS Tracker Modules 55 To the general ASIC errors belong e all 12C connection problems to one or several of the FEHs ASICs e a high leakage current outside of the specifications e APV25 header faults or trigger problems e APVMUX and PLL failures e increased or decreased low voltage power consumption of the FEH Nearly all these defect types can be detected by simple test procedures 5 5 2 1 I C Connection Problems will be detected by scanning the 12C bus and ap plying several read write read cycles to the registers of all ASICs connected Although I C is not a very fast protocol by means of frequencies used
123. ed by sabbaticals leading him to CERN and other research centres For his contributions to the theory of nuclear reactions he was awarded the Nobel Prize in 1967 FELIx BLOCH 1905 in Zurich 11983 in Zurich He was working with several universities and research centres like Stanford and CERN The Nobel Prize was awarded to him in 1952 for nuclear magnetic precision measurements 4 Silicon Strip Tracker and its End cap Modules 39 atomic polarisation In thin layers the deposited energy is reduced because a fraction of the lost energy is carried off by energetic knock on electrons also known as 6 electrons These considerations lead to restricted energy loss which is expressed by an additional term C Z in the Bethe Bloch equation where Tupper inf Teut Tmax with Teut depending on the material and the incident particle momentum The mean energy loss dE dx per unit absorber thickness by charged particles in matter as given by the Bethe Bloch formula Eq 4 5 on the facing page is more or less useless in describing the behaviour of a single particle in a thin absorber because of the stochastic nature of the energy loss process The probability distribution function pdf f describing the distri bution of energy loss A in absorber thickness x is usually called the Landau distribution f A af exp ulnu uA sin tu du 4 7 wih Es z ee ent a eee Ine In G 6 P 2mc 8 and the Euler s constant yg 0 577
124. edener Umgebungsvariablen wie die beiden Versorgungsspannungen auf dem Hybri den den Gesamtleckstrom des Moduls oder die Temperatur eines Thermistor auf dem Auslesehybriden APV25 multiplexer detector control unit Zusammenfassung 3 3 Leistungsverhalten der Module Die zu erwartenden Signalh hen werden durch eine Landau Verteilung beschrieben deren mittlerer Energieverlust durch die Bethe Bloch Formel f r die verwendeten Teil chenenergien und das Absorbermaterial Silizium gegeben ist F r minimalionisierende Teilchen MIPs erwarten wir die Erzeugung von 23500 Elektron Loch Paaren in 320 um starken Sensoren und 36 700 Elektron Loch Paare in 500 um starken Sensoren Entschei dend f r den Betrieb des Spurdetektors ist ein hinreichendes Signal zu Rausch Verh ltnis SNR auch nach 10 Jahren Betrieb Da die Ladungssammlungseffizienz mit der Be strahlung leicht abnimmt und der Leckstrom sowie der zugeh rige Rauschanteil um mehre re Gr enordnungen zunimmt wird sich das SNR im Laufe des Betriebes deutlich ver schlechtern Im Wesentlichen tragen drei verschiedene Quellen zum Rauschen bei Diese sind das sog Shot Rauschen welches z B in Halbleitern durch die statische Fluktuation der Elektronemission verursacht wird das sog Thermische Rauschen hervorgerufen durch thermische Geschwindigkeitsfluktuationen der Ladungstr ger und das Flicker Rauschen wof r Trapping Detrapping Prozesse in Halbleitern bei DC Str
125. edl 2001 Dec wo Inv Mode wo Cal Module 30216630300027 Peak wo Inv Mode wo Cal Module 30216630300027 SNR SNR 7 Test System Performance and Module Qualification Studies 103 E x E En i zZ End E a o 201 i 20 i E L T E e F E e T aO 16 p 16 BEE les E J 14 14 E Eo 38 aa oo i Pola 28 H 12 en 3 3 12 By BE E 10 a ee Beis Pei ee een 10 Fa ner Meee ee 50 100 10 200 250 300 30 2 100 200 300 400 500 Bias voltage V Bias voltage V Figure 7 25 SNR in Deconvolution mode E 8 Figure 7 26 SNR in Deconvolution mode as function of bias voltage at room tem 3 E 90 as function of bias voltage at 10 C perature taken with a Sr source on the i s taken with cosmic rays on the FTS DTS 20 57 20 5 L z I F o E 20 20 E a H 19 5F i 19 5 E En ds orale ET i E Se 3 a e a 18 5 18 5 oi E E 4 E E C i l i 18 388 18 T 17 5 EJES 17 5 I ES 28 F I E San H i i 17E E ala ea ela A A AA 0 200 400 600 800 1000 0 200 400 600 800 1000 Leakage current A Leakage Current pA Figure 7 27 SNR in Deconvolution wo inverter mode as function of leakage current taken with a Sr source at a voltage of 300 V at the DTS Both modules tested show a deterioration of the same level of about 10 Performing the same measurement with a Sr source results in a significantly later settle ment of the SNR to
126. el arises from its decay to a bb pair in the ttH channel Drollinger et al 2001 In the channel ttH ggbbbb where the Higgs Boson decays to bb one top quark decays hadronically and the second one leptonically DS The events are expected to show one isolated lepton missing transverse energy Ey and six jets four b jets and two non b jets but initial and final state radiation are sources of additional jets So the number of jets per event is typically higher than six On the other hand not all six quarks of the hard process can be always recognised as individual jets in the detector in which case it is impossible to reconstruct the event correctly even if there are six or more jets For the reconstruction of resonances it is necessary to assign the jets of an event to the corresponding quarks of the hard process The nominal mass of the leptonically decaying W boson together with E and the lepton four momentum is used to calculate two solutions of the longitudinal momentum of the neutrino pz v which is needed for the mass reconstruction of the leptonically decaying top Good mass resolution and the identification of b jets is essential to reduce the number of wrong combinations in the event reconstruction A good mass resolution can be obtained when the energy and direction of each reconstructed jet agree as closely as possible with the quantities of the corresponding parent quark The identification of b jets is even more
127. ell as connectivity between modules and optical hybrids The voltages are split up into three groups of low and six groups of high voltage to limit the voltage drops across the individual low voltage groups and the currents on the high voltage lines 4 1 2 Silicon Micro Strip Sensors The necessity to instrument over 210 m of active area is one of the reasons behind the choice of single sided p on n silicon micro strip devices Beside the fact that these allow very high voltage operation and can be manufactured using industrial 6 wafer lines compatible with standard micro electronics technology leading to a reliable and cost effective production there is another important reason why p on n silicon is chosen During irradiation additional acceptor defects are produced which will change the effective doping concentration For n type substrate material the effective doping concentration decreases till the type inversion is reached and the material becomes effectively a p type and afterwards the effective doping concentration increases As the depletion voltage scales with the effective doping concentration compare App B III on page 128 n type substrates will result in lower depletion voltage after irradiation then p type Efficient charge collection of irradiated sensors can be achieved provided that the sensor can be biased significantly above its depletion voltage This requirement drives the choice of substrate resistivity as well as the s
128. em n dotierten Bulk sind p Implantatstreifen einge lassen welche ber Polysilicon Widerst nde nominal 1 8 MQ auf den Biasring gef hrt werden Die Auslese erfolgt ber kapazitiv gekoppelte Aluminiumstreifen welche in der Breite ber die p Implantatstreifen hinausragen und somit eine h here Spannungsfes tigkeit ergeben Die R ckseite der Sensoren ist aluminisiert eine n Schicht garantiert einen guten Ohmschen Kontakt Die Sensoren werden auf U f rmigen Kohlefaserrahmen montiert auf deren Beinen Kaptonkabel zur Spannungsversorgung und Isolation aufgeklebt sind Die Beine werden an einem Ende von einem Querst ck zusammengehalten auf dem der Auslesehybrid und der Pitchadapter Streifenabstandswandler aufgebracht sind Zusammenfassung Abbildung 2 Dimensionen und Layout des zentralen Spurdetektors Die L nge ber alles betr gt 5 4m Hartmann 2002 3 2 Auslese und Steuerung Im CMS Spurdetektor findet ein uni direktionales Takt und Ausl sesignal Netzwerk Verwendung w hrend die Steuersignale wie z B Konfigurationsparameter ber ein bi direktionales Netzwerk bertragen werden Diese beiden Netze sind als digital optische Netzwerke realisiert im Gegensatz zu dem Auslesepfad welcher eine analog optische Uber tragung nutzt Auslesehybrid Die Auslesehybride vereinen die verschiedenen Chips welche f r den Betrieb und die Steuerung der Module ben tigt werden PLL Der PLL Chip regeneriert den Takt und das Ausl
129. emiconductor most significant bit micro strip gas chamber Minimal Super Symmetric Model multiplexer see Sec 4 2 1 3 on page 32 nonionising energy loss nuclear instrumentation module non return to zero with invert 1 on change data coding scheme optical electrical converter object oriented pitch adapter probability distribution function positive intrinsic negative lead tungstate peripheral component interconnect Industrial bus system printed circuit board Petal Integration Centre see Sec 5 4 3 on page 54 programmable logic device phase locked loop see Sec 4 2 1 1 on page 28 photomultiplier PCI Mezzanine card parallel plate chamber Process Qualification Centre see Sec 5 2 2 on page 48 Quality Test Centre see Sec 5 2 1 on page 47 Rutherford Appleton Laboratory random access memory root mean square Research amp Development resistive plate chamber serial clock small computer system interface serial data sequencer see Sec 6 2 2 on page 64 single event upset Standard Model signal to noise ratio Superconduction Super Collider Silicon Strip Tracker see Sec 4 on page 22 technical design report tracker end cap see Sec 4 on page 22 tracker inner disks CONTINUED ON NEXT PAGE 126 Acronym TOB TTC TTCrx TTL VCC VDD VEE VHDL VPT VME VSS VUTRI WLDC WLS APPENDIX A ACRONYMS AND ABBREVIATIONS CONTINUED FORM PERVIOUS PAGE Definition tracker out
130. en at the FTS The modules ambient temperature was kept at 10 C during the one day measurement and a bias voltage of 300 V exceeding the depletion voltage of approximately 77 V strongly which ensures full SNR performance compare Sec 7 5 2 on page 102 The fit functions are in perfect agreement with the measurements and do not depend on fit ranges The Gaussian component for both Peak and Deconvolution mode are significantly different although their absolute values nearly match While the Deconvolution mode SNR distribution gets strongly blurred by the timing jitter the Peak mode is only minor affected The loss in 7 Test System Performance and Module Qualification Studies 101 signal amplitude due to the timing jitter as stated in Sec 7 4 2 2 on page 97 also causes a broadening of approximately the same order of the measured SNR distribution Table 7 2 shows the results of the cosmic rays calibration for the FTS For both readout modes the measured noise is larger than the predicted one compare Tab 4 5 on page 44 While the Peak mode differs only on a scale of 6 25 the Deconvolution mode result differs by 15 8 But this deviation for the Deconvolution mode may arise from a slight timing offset of 5ns Furthermore the Deconvolution mode is stronger affected by signal truncation caused by the cluster algorithm than the Peak mode because of its smaller overall SNR value Peak mode Deconvolution mode Deconvolution mode strong timing j
131. en eines einzelnen Detektorkanals h ngt prim r von der an den Vorverst r ker angeschlossenen Kapazit t ab und ist damit sensitiv auf fehlende Bondverbindungen und Kurzschl sse zwischen mehreren Kan len Allerdings zeigen sich auch andere Streifen fehler im Rauschen wie z B Streifen mit erh htem Leckstrom oder mit fehlerhaftem Vor widerstand Um alle diese Fehlertypen eindeutig identifizieren zu k nnen ist insbesondere die interne Kalibrationseinheit sehr hilfreich Kurzschl sse der kapazitiven Koppelung der Auslesestreifen sog Pinholes lassen sich allerdings nur teilweise im Rauschen und der Kalibration identifizieren Andererseits k nnen diese Kurzschl sse nach Bestrahlung einen ganzen APV25 Chip mit seinen 128 Kan len in Mitleidenschaft geziehen 5 Die Karlsruher Teststationen In Karlsruhe wurden zwei Stationen speziell f r die Qualit tskontrolle von Spurdetek tormodulen konzipiert entwickelt und gebaut Die erste Station ist auf die Anforderungen einer schnellen Qualifizierung der Detek tormodule nach dem Bonden optimiert Obwohl hierbei ein voller thermischer Zyklus von Raumtemperatur auf 10 C durchlaufen wird ist es m glich mit der Rate des Bonders zu testen da die Station ber eine entsprechende thermische Isolation und ein automatisch steuerbares K hlsystem verf gt Der gesamte Aufbau dieser Station ist sehr kompakt gehalten Die zweite Station ist auf tiefergehende Analysen von Modulfehlern ausgelegt
132. ensor design characteristics which improve the high voltage performance of the detectors Furthermore surface damages during irradiation may affect the capacitive coupling between adjacent strips this effect can be kept under control by choosing lt 100 gt silicon material Braibant et al 2000 The sensors are single sided with p strips on a n type phosphorus doped substrate of a resistivity of 1 5 4kQ cm thin sensors and 4 8kQ cm thick sensors The schematic structure of the sensors used within CMS is shown in Fig 4 2 on the facing page A picture of a tracker end cap TEC sensor is shown in Fig 4 3 on the next page and a schematic of the sensor design in Fig 4 4 on page 26 inside a large border area the guard and bias rings are visible together with bias resistors and bonding test pads located at the beginning of the strips The relative low resistivity material used in the inner region will help to keep the depletion voltage low when the bulk will be type inverted after irradiation while the high resistivity substrates used in the outer parts where the irradiation level will be lower reduces the naturally higher initial depletion voltage of the thicker sensors Viep d p My 2001 The width of the implant strips depends on its pitch A constant width pitch ratio of 0 25 is used within CMS Aluminium readout strips coupled capacitive over the p implants will be wider than the implant underneath metal overhang
133. epresent a 0 The idle symbol which is sent when no other data are pending is 11111 resulting in a square wave output of half the clock frequency 4 2 4 4 TTC The timing trigger control TTC system as a common development for all LHC experiments will distribute timing pulses from the LHC master clock using high power lasers operating at 1310nm by means of optical fibers with small dispersion and a passive hierarchical network of optical couplers This type of system is well advanced and achieves splitting ratios of up to 1 1000 and timing jitters on clock edges of less than 50 ps 4 3 Expected Module Performance The efficiency to reconstruct single isolated muon tracks is expected to be 100 over most of the pseudo rapidity coverage First results of dedicated studies of performance at the detectors large n boundaries are in progress The precision of the track reconstruction performance are evaluated studying single muon tracks at several p values As an example the precision of the track curvature is shown in Fig 4 15 on the next page in terms of o p p as a function of pseudo rapidity 38 4 3 Expected Module Performance o 1 3 10 30 10 100 a 0 30 go 300 1000 GeV et OZ ip M1 o p p 107 10 Figure 4 15 Transverse momentum Figure 4 16 Zimp resolution as a resolution as a function of pseudo function of pseudo rapidity for sin rapidity for single muons for several gle muons for several p v
134. er barrel timing trigger control see Sec 4 2 4 4 on page 37 timing trigger control receiver ASIC transistor transistor logic voltage at common collector positive electrical connection voltage drain drain voltage emitter emitter VHSIC very high speed integrated circuit hardware description lan guage vacuum photo triode Versa Module Eurocard IEEE 1014 industrial bus system voltage source source very ultimate tracker readout interface wall less drift tube wavelength shifting Table A 1 Acronyms used within this work APPENDIX A ACRONYMS AND ABBREVIATIONS Symbol a PB C Definition fine structure constant e 4rreghc speed in units of c speed of light mean excitation energy Avogardro s number Atomic number Atomic mass Electron mass Classical electron radius e Areomec Stefan Boltzmann constant Incident particle energy yM e Kinetic energy mean energy loss Euler s constant Radiation length Moli re Radius leakage or dark current of sensors leakage current driven by a single strip leakage current over dielectric poly resistor dielectric resistance inter strip resistance return line resistor on the FEH bias voltage coupling capacitance inter strip capacitance strip to backplane capacitance strip total capacitance amplifiers total capacitive load potential of p implant strip 127 Unit or Value 1 137 035 999 76 50 299 792 458ms 1 eV 6 022 141 99 47
135. es only the smaller correction of central timing jitter will be applied the larger correction gives an estimate for the precision being much better for Peak mode 7 4 2 3 Clustering Algorithm In contrast to the calibration pulses a physical signal will typically spread over more than one readout channel either due to its inclination which should stay small within a detector or simply due to charge sharing of tracks penetrating the silicon between two strips Therefore a clustering algorithm has to be used to collect all the signal generated by a passing particle The clustering algorithm used within the analysis software searches for 1 a channel with a signal of at least five sigma significance 2 and collecting all neighbouring channels with signals of more than three sigma signifi cance Requiring at least a five sigma signal for the central channel reduces background significantly while the three sigma cut on neighbouring channels prevents noise but may also result in a truncation of the signal for larger clusters The cluster charge is the sum of all contributing channels To compute the cluster noises several methods may be used Either one adds the noise of the individual cluster channels like it is done for the signal or one takes the average noise of the involved channels or calculates a weighted sum or uses the maximal channel noise within the cluster Finally a conservative method was chosen The cluster noise is computed
136. est document of the Module Test Working Group Dirkes et al 2002 and will be revised by the final quality control document The current version also includes cut definitions for backplane pulsing and light test which have to be reviewed and are not cited here To reach the design goal of less than 2 bad channels in the SST this number is reflected by the definition of the module quality grades given in Tab 5 1 Grade Number of Module leakage current bad strips one sensor two sensors at 450 V increase 450 550V at 450V increase 450 550 V less than 1 lt 10pA less than 104A lt 20uA less than 20 yA less than 2 lt 10pA less than 104A lt 20uA less than 20 yA more than 2 bad strips or leakage current out of specifications Q We Table 5 1 Module quality grade definitions 6 Karlsruhe Test Stations 61 6 Karlsruhe Test Stations Besides the silicon sensor based activities of being a QTC and IQC centre the Karlsruhe CMS group has the responsibility of building 1600 ring 5 modules within the two years production period This leads to an average work load of 4 modules per day calculating 200 working days a year and thus implies the need of well adapted test and quality control systems Taking the needs arising from our production engagement a profile for a test system can be sketched as follows e It has to serve the FEH with stabilised and low noise low voltage power e High voltage has to be supplied with low noise contributions as
137. eturn line resistor Rye on the FEH The voltage drops over the two resistors define the potential Vimp of the pt implant strip Vimp Rpoly Istrip Rret Teak 5 16 The APV25 on the other side tries to keep its amplifier inputs at the constant level of its virtual ground which is for the APV25s at about 0 75V As long as the potential of the p implant strips stays smaller as the APV25 virtual ground a current will flow out of the APV25 which drives the amplifier into saturation V V250 When the leakage currents increases and the potential Vimp matches the virtual ground of 0 75 V the amplifier works normal without any distortion If the affected strip has a normal leakage current behaviour Eq 5 16 evaluated with Rret 22kN 2kO Rpoiy 1 8 MQ and Istrip Teak 512 shows that the potential matching will occur at about 30 4A With further increasing leakage current the potential will increase above the virtual ground and the pinhole drives a current into the APV25 Again the amplifier saturates but this time with the opposite sign V VSS which turns the inverter transistor on and the latter will draw a current The discussion above handles about ideal pinholes but experience shows that a pinhole as a defect in the insulation layer is typically coupled to further defects The ZEUS detector 5 Quality Control at CMS Tracker Modules 59
138. etween the decay points of the two B mesons The BaBar and Belle experiments first observed non zero CP violation in 2001 and have now measured sin 28 with good precision sin28 0 719 0 074 0 035 Belle 2 21 sin28 0 741 40 067 0 034 BaBar 2 22 The systematic errors are dominated by the statistics of the auxiliary samples used to evaluate them and should continue to fall with increasing luminosity The combined precision of the 14 2 6 Heavy Ion Collisions experiments will be impressive at the start of LHC but nonetheless the huge statistics of the LHC can bring valuable new insights to this important channel There is presently substantial effort on determination of the angle a This angle is ac cessible in b u transitions leading to uudd final states e g B ata In contrast to the golden mode however there are both tree and penguin decay amplitudes that con tribute appreciably to these decays rendering the precise determination of a is more difficult The experimental situation is also less favourable due to the very small branching frac tions B B mtr 5 x 1070 higher backgrounds and the difficulty in distinguishing B str decays from the more numerous BP Ktr decays The e e B factories BaBar and Belle operating at the Y 4 resonance will not be in a position to explore the Bg system Since it is moreover very desirable to have large data samples available to study Bs d
139. for Particle Physics Creative Electronics Systems carbon fiber cost review committee common mode complementary MOS Compact Muon Solenoid commercial of the shelf complex programmable logic device central processing unit cathode strip chamber coefficient of thermal expansion DO experiment at FermiLab digital analogue converter data acquisition system direct current detector control unit see Sec 4 2 1 4 on page 33 digital input output direct memory access differential non linearity dual port memory Detector R amp D Committee Diagnostic Test Station see Sec 6 5 on page 83 drift tube drift tube with bunch crossing identification device under test electron beam welding electron calorimeter CONTINUED ON NEXT PAGE 124 Acronym EDR ENC ESD FEC FED FEH FPGA FIFO FHIT FWHM GCD GND GUI FPGA FTS HCAL HIP HF HV PC IBM IC ICB ICL INL IQC JTAG KaRinA LED LEP LHC LHCC LLD Lol LSB LV LVDS M200 MAC MIO APPENDIX A ACRONYMS AND ABBREVIATIONS CONTINUED FORM PERVIOUS PAGE Definition engineering design report equivalent noise current typical stated as RMS electrons producing a SNR of one electro static discharge front end controller see Sec 4 2 4 1 on page 35 front end driver see Sec 4 2 3 on page 35 front end hybrid see Sec 4 2 1 on page 28 field programmable gate array first in first out FEH Industrial Tester see Sec 5 3 1 on p
140. ftware developed A description of the two different test systems and their scope of application concludes this section In Sec 7 the last section before the conclusion an overview of the experimental results and the module behaviour is given After an analysis of the noise performance the absolute calibration of the test system using cosmic ray particles is shown Based on this systematic studies of the modules performance in terms of signal to noise ratio SNR are presented proving that the modules follow the expected performance The last part of this section deals with fault detection and shows the experimental signatures of the different module faults This results in the presentation of the discovered pinhole behaviour and its influence on the final module design 2 High Energy Physics at the Large Hadron Collider 3 2 High Energy Physics at the Large Hadron Collider Among currently approved projects in high energy physics the LHC has the unique poten tial sufficient energy and luminosity to probe in detail the TeV energy scale relevant to electroweak symmetry breaking to uncover and explore the physics behind it This study involves the following steps e Discover or exclude the single Higgs boson of the Standard Model SM and or the multiple Higgs bosons of supersymmetry e Discover or exclude supersymmetry in essentially the full theoretically allowed mass range e Discover or exclude new dynamics at the TeV scale
141. g 10 contingency Rings 1 amp 2 4 4 amp 7 5275 sensors Rings 34 5 4 6 6150 sensors QTC Vienna QTC Karisruhe uo 625 sensors 4650 sensors 6150 sensors x MAC Vienna MAC Lyon MAC Brussels Ring 2 625 modules Ring 14447 3050 modules Ring 345463425 modules Yv Ring 2 625 mod Ring 4 1125 mod Ring 7 1625 mod Ring 1 amp 3 1000 mod Ring 5 1600 mod Ring 6 1125 mod BC Vienna BC Zurich BC Strasbourg 8C Kartsruhe BC Aachen Petal Integration each BC to each PIC 38 transportways 41 Petals 41 Petals 41 Petals 41 Petals 41 Petals 41 Petals 41 Petals A PIC Aachen Pic Hamburg PIC Karlsruhe PIC Lyon PIC Strasbourg SM1 Aachen SM2 Ka gt Lyon Figure 5 8 Module logistics for tracker end caps Simonis 2003 5 4 Module Quality Test Strategy The real construction of a module starts with the mechanical assembly within the module assembly centres MACs From here the modules go to the Bonding and Module Quality Assurance Centres which finalise the construction by wire bonding of the silicon sensors to the PA and sensor sensor Figure 5 8 shows the logistics chain of the tracker end cap TEC production with respect to the flow of sensors and modules 5 4 1 Module Assembly Centre Figure 5 9 on the next page shows a photograph of an automatic assembly robot showing two working platforms an assembly platform in the front and a supply platform in the back as well as a vacuum
142. ger control is implemented by a complex programmable logic device CPLD Trigger and clock LVDS signals are brought in on the front panel or can be taken from the PC clock together with internally generated software triggers The fine adjustment of the clock phase with respect to the data can be set under software control in order to obtain the optimum sampling point at the ADCs For the FED prototype this clock phase adjustment is implemented only once for all channels while for the final FED each channel can be adjusted independently The heart of the FED design is an field programmable gate array FPGA which permits a large fraction of the cards functionality to be re configurable in firmware and thereby maintains a flexible hardware architecture programmable in VHSIC very high speed integrated circuit hardware description language VHDL The current firmware design version 2 45 configures the FED to provide raw data capture together with APV25 header finding algorithms and clock and trigger counters of 32 bit depth During normal operation the FPGA is loaded on power up under software control from an on board Flash memory FED Manual Unfortunately the FEDs logics turns out to be very heat sensitive which especially makes problems in a heavily populated PCs PCI bus where the FED is mounted with a PMC PCI adapter card causing a limited heat transfer To prevent ghost triggers header detection errors or even worser problems one has to ens
143. gh a switch and a 4009 resistor to a reference voltage pad typically ground GND This allows to variate the termination in steps between 4002 and 502 to adjust the dynamic range of the output differential voltage 4 2 1 4 DCU The detector control unit DCU is a special ASIC to be used mainly in the Silicon Strip Tracker for the monitoring of embedded parameters like supply voltages currents and temperatures on the FEHs Based on an internal band gap reference voltage the DCU measures up to seven external channels and one internal temperature sensor The DCU provides a 10 uA and a 20 A con stant current source that can be used to drive external thermistors for temperature measure ments These measurements are performed using a 12 bit successive approximation analogue digital converter analogue digital converters ADCs with the absolute value of the integral non linearity INL and the differential non linearity DNL smaller than the least significant bit LSB for the complete range of operation temperature from 50 to 50 C The conversa tion time will be always below 25 us and with a RMS noise of less than 0 5 LSB Furthermore the power consumption could be kept below 40mW Readout and control is done using a 12C standard protocol with 2 5 V CMOS levels Magazzu et al 2001 The connections of the DCU as they are used on CMS end cap modules are shown in Fig 4 12 on the following page 34 4 2 Readout and Control Chain 4 vDD
144. goes through a minimum where the current and the voltage noise contributions are equal At short shaping times the voltage noise 44 4 3 Expected Module Performance Noise source Type ENC RMS e expression at T 10 C Deconvolution a a parallel y 108 Vin pA T ms XDA n e Bias resis e KIT a Ts Ns tance Ryoly parallel N Boag x 22 5 Ron Ma x0 45 Metal stri kT Rs ES Rs 2 resistance Bi RETES ge Ctoty T 13 Crosl PF Ts 08 EE en series 246 36 Ctor PF 396 59 x Otor PF Table 4 4 Noise sources types and relative ENC evaluation CMS Collaboration 1998 2000 dominates whereas at long shaping times the current noise takes over The noise minimum is flattened by the presence of 1 f noise Increasing the detector capacitance will increase the voltage noise and shift the noise minimum to longer shaping times Table 4 4 summarises the main noise sources and their corresponding dependencies as well as expressions for ENC evaluation for Peak and Deconvolution mode at 10 C as they are listed in the CMS Collaboration 1998 and CMS Collaboration 2000 Beside the contributions from the electronics and the sensors to the noise the pitch adapter PA has also to be taken into consideration A contribution of the PA arises from its additional resistance which increases the series resistance R Another contribution arises from PAs strip capacities which increase the effective strip capacity C
145. h is directly accessible only from the kernel to the virtual memory seen from the library and thus from the application side Memory mapping itself is a complex topic which requires some knowledge about memory management in operating systems and is done in slightly different ways by Lynx and Linux Furthermore the PCI9080 bridge supports DMA access which is a hardware mechanism that allows peripheral components to transfer their data directly to and from the main memory without the need of the systems CPU to be involved in the transfer A draw back of DMA access is that it needs a consecutive physical memory while the memory management is organised in pages of the order of a few kB size Thus if the module needs a larger block of consecutive physical memory and the FED driver needs 2MB it is getting complicated to allocate it during run time Fortunately there is a way of allocation memory during boot Multi byte values are stored on PCs LSB first little endian while most high level platforms work the other way big endian 6 Karlsruhe Test Stations 77 time before the memory management itself starts by passing a mem argument to the kernel But this way of allocating memory has the disadvantage that it can not be returned to the systems memory management and is lost for other applications 6 3 2 Libraries The way the FED is accessed by memory mapping moves code from the driver level to the library level A library in general
146. han the others i e have one side that is proportional to a higher power of A than are the others but the remaining triangle formed by multiplying the first column by the complex conjugate of the third column has all sides of order A It is this triangle that is usually discussed when considering the impact of experimental measurements on the parameters of the CKM matrix These unitarity relations need to be verified experimentally a violation of unitarity would point to new physics e g a fourth generation in which case the 3 x 3 submatrix need not be unitary The unitarity triangle of interest namely Vua V Vea Vo Via Va 0 is usually rescaled by dividing through by V gV3 giving a triangle whose base has unit length and lies along the x axis and whose apex in terms of the parameters A A p and 7 and using the definition p 1 p and 7 1 7 is at 9 7 up to corrections of order A By dividing out V we largely eliminate dependence on the parameter A which is in any case relatively well known A 0 85 0 04 The angles of the unitarity triangle are Via Vi Vea Vo Vud Vat Ho e u 2 1 a arg Via Vo j b arg VaV y arg Va Vs 2 10 compare Fig 2 3 2 High Energy Physics at the Large Hadron Collider 11 so eo HI wr da Wo y d Figure 2 4 B B transition diagrams An invariant measure of the size of the CP violation in the CKM matrix is given by the Ja
147. hang technique used compare Fig 5 11 This can influence the local breakdown voltage significantly resulting typically in a localised breakdown and a strong increase of the strips leakage current that also can be visible in the modules V characteristic In such a case the increased strip leak age current will cause an significant increase of noise for the corresponding channel compare Tab 4 4 If the scratch goes deeper into the material and damages also the SiO2 Si3 N4 oxide layer a pinhole may be created compare Sec 5 5 3 5 Furthermore a scratch will typically affect several strips which also can be shorted 5 Quality Control at CMS Tracker Modules 57 on Al short SIO SIN N J Figure 5 12 Schematic of short fault in silicon strip de tectors which can occur on the surface readout strips as well as in the substrate p ayer n bulk implants p implant E field 5 5 3 2 Shorted Strips A shorted strip as schematically shown in Fig 5 12 will cause a higher capacitive load to the APV25 amplifier Furthermore the shorted strips will show the same CM behaviour which leads to an increased noise in the shorted channels compared to what is expected from the noise analysis refer to 4 3 and Tab 4 4 on page 44 Furthermore shorts will reduce the resolution due to charge sharing between the strips Shorted strips can be detected easily using the APV25s internal calibration circuit com pare Sec 4 2 1 2
148. hannel noise effect is still unclear But it has to be stress again this APV25 border channel noise effect occurs only in Peak modes while the Deconvolution modes only see the sensor edge effect The sensor edge effect can be explained by the coupling of the first and last channel to the sensors bias ring The distance between the bias ring and these two strips depends on the 116 7 7 Module Fault Detection Studies P a Pedestal ADC N o ITETTETTTI Figure 7 48 Pedestal fault of 130 i eerie ee aring 5 prototype FEH This is E z the same kind of plot as given 100 aa Z o in Fig 7 1 The two APV25s a ss te the middle have an sig 50 a ee a E S nificantly increased pedestal E 83 which does not further disturb ooo BENET 40050060000 F the APV25 functionality Abs sensor type but is typically in the order if the sensors pitch The resulting impedances are of the order of 104Q at a frequency of 1 MHz which is a hundred times smaller than the poly resistors Therefore a reduction of the noise on the bias ring results in a reduced sensor edge noise effect Civinini 2003 Consequently an additional filter capacitor is added on the HV return path on the FEH compare Fig F 2 on page 139 reducing the CM contributions to the edge channels noise significantly Finally an increased noise effect will be found for the channels 256 amp 257 and 5128512 for FEHs with six APV25s if the timi
149. he APV25 input channel Section 5 5 3 5 on page 58 discussed the expected pinhole behaviour The given configuration of the prototypes as shown in Fig 7 43 utilises two serial resistors of 22kN and 2kQ in the detectors HV return line Rret 24kQ which are used for leakage current measurements by the DCU compare Sec 4 2 1 4 on page 33 The final version of the FEH replaces this resistors with smaller ones with values of 2kQ and 1kQ which is motivated by results that will be shown in Sec 7 8 on page 117 Together with a poly silicon resistor Ryoly of about 1 8 MQ these define the pt implant potential Vimp according to Eq 5 16 on page 58 For an irradiated prototype module showing a leakage current of 500 yA the implant potential results to Vimp 12V For the final FEH this value drops to Vimp 2 4V But these calculations consider a homogeneous generation of the module leakage current by the individual strips and we have serious hints that this assumption is not true for pinholes Former experiments have shown that pinholes have higher leakage currents Coldewey 2000 and present QTC results show a clear relation between pinholes and an increased strip leakage currents Thus assuming an increased strip leakage current strip by the order of one magnitude for a pinhole seems to be much more realistic and results in a p implant potential for the final FEH design of 10 3 V Even the optimistic value of 2 4V for the implant potential it wi
150. he FED Weiler 2002b 6 2 7 Multiplexer Device Another development for the CMS collaboration is the multiplexer device see Fig 6 9 allowing the measurement of several modules or even a complete petal with a reduced test system Basic idea is to use only one FED with its 8 input channels and to test in serial during a long term test The multiplexer device is based on the motherboard and its communication and powering scheme On the motherboard up to eight multiplexer cards are mounted performing each a 10 1 multiplexing Therefore ten switches of type OMRON G6H 2 are used which are miniature surface mount relays A contact resistance of less than 60mQ and an insulation resistance of more than 1000 MQ ensure minimal damping and crosstalk Weiler 2002b 6 2 3 FED The used FED is a prototype of the final CMS silicon micro strip tracker FED It is imple mented as an 8 channel ADC on a PMC format The choice of the PMC format which has 6 Karlsruhe Test Stations 71 un FPGA gt Ported E34 nz i Memory Xilinx Channel7 29 XC 4036 gt screen N 4 address LVDS en receiver Trigger Clock Figure 6 10 Block diagram and picture of FED shows the building blocks of the FED Based on a large FPGA here covered by three small cooling blocks a DPM is used to store the output of the8 ADC The connection to the PC is done via PLX 9080 bridge supporting a full PCI interface i
151. he kernel and to check the availability of hardware For the PCI bus components this is directly supported by the PCI bus specifications that requires a vendor and device ID and which can be checked via the PCI subsystem of the Linux kernel 76 6 3 Software Layout Open Close routines will make the drivers functionality available to application programs Furthermore these device driver functions control multiple access to the same hardware by blocking unblocking or attaching different sub devices if available and supported Read Write functions send or receive data from or to the connected hardware Ioctl gives control over the hardware specific behaviour These control operations are usu ally not available through the read write functions abstraction For example setting the slave address of the I C bus is implemented by ioctl calls While the open close read write and ioctl function are accessible by the user through the corresponding system calls the install and uninstall functions are hidden in application space and it is in the kernels responsibility to execute them For a Linux system the details of writing device drivers can be found in Rubini 1998 Access to the device is given by creating a corresponding device file in the device directory dev This has a device number assigned matched by the device driver itself and by the module autoloader which looks up the corresponding driver name in the etc module conf file The a
152. he power consumption is monitored and interlocked to prevent damages due to electric faults 5 5 3 Strip Error Detection Strip errors are principally much less critical than electrical errors on the FEH because they typically will not affect larger groups Nevertheless there are serious defects based on strip errors The most problematic ones are pinholes which may drive a complete APV25 into saturation Unfortunately the CM suppression of the APV25 complicates the understanding of single strip faults behaviour Effectively the CM suppression reduces bias line voltage noise to which all strips are sensitive compare Sec 4 2 1 2 on page 30 But only bonded channels transfer this signal to the APV25 This causes that unbonded channel seem to be noisy ON sCralCh high E field Figure 5 11 Schematic of SIO SIN N scratch fault in silicon strip de a tectors Due to the metal overhang technique used a surface scratch may have dra matic consequences for the field configuration near the p E field implant resulting in the worst case in a localised breakdown ayer p implant 5 5 3 1 Scratches on the top side of the sensor strongly influence the detectors perfor mance As soon as the scratch penetrates the passivation layer on top the readout strips the latter can be damaged If an edge on the metal readout strip is produced a distortion of the electric field inside the sensor may result because of the metal over
153. hwankungen Hierbei kommt es zu einem durchschnittlichen Verlust von 7 5 bei optimalen Einstel lungen der bei einer Abweichung von nur 5ns sich auf 18 vergr ert In Tabelle 2 werden die gemessenen SNR Werte mit den vorhergesagten Werten aus Tabelle 1 verglichen Die Abweichung von der Vorhersage liegt im Peak Modus bei weniger als 7 wohingegen die Abweichung im Deconvolution Modus stark von der Korrektur abh ngt Bei minimaler Korrektur ergibt sich eine Abweichung von 16 wohingegen bei einer mittleren Korrektur 5 ns Fehler die Abweichung auf ein 5 Level abf llt W hrend normale Streifenfehler Fehlende Bondverbindungen Kurzschl sse unter brochene Auslesestreifen zu Ausf llen eines einzelnen Kanals f hren k nnen Pinholes Zusammenfassung SENSOR APV _ ee Test input Rinv 1 000 a Vimp O fav strip Rpinhole Test v125 8 x i i N this point common for F gt all 128 chans BR yee gt I uf H N p det Rpoly 1 8 MQ X bias gt DCU 75V Rret 24kQ VSS base line restorer Abbildung 8 Pinholes und der APV25 Schaltplan Calibration amplitude ADC 0 50 100 150 200 250 300 350 400 Leakage current yA Peak w Cal wo Inv Mode Module 30200020000638 Abbildung 9 Die Kalibrationsamplituden f r normale Kan le sind unabh ngig vom Leckstrom Ande
154. iff ADC Raw noise ADC Raw noise ADC Raw noise diff ADC 0 5 Sensor channel number 100 Inverter off Ber 100 Be nee ni ss i 200 300 400 500 DB channel number Sensor channel number 100 w Calib w Inverter on wo Calib w Inverter on 400 Difference 3 i 100 i 200 i 300 i i 400 500 DB channel number Peak w amp wo Inv Mode wo Cal Module 30200020000503 30200020000503 1 Peak w amp wo Inv Mode w Cal Module 30200020000503 30200020000503 1 93 Figure 7 8 The in verter stage influence on raw noise for Peak mode without calibration cir cuit powered The upper plot shows the raw noise with inverter off and the plot in the middle with inverter on Significant differences arise between inverter off and on mode are obvious Figure 7 9 Inverter stage influence on raw noise for Peak mode with calibration circuit used The upper plot shows the raw noise without usage of the stage but with powered calibration circuit while the plot in the middle shows inverter and cali The differ ence lower plot shows together with Fig 7 8 on page 93 that the calibra tion circuit has a clear influence on the noise be haviour of a module inverter bration on 94 7 3 Module Leakage Current T 60F H a H H a El 5 Bet nn a L i AA E S 50 a E i
155. ifferent persons Cederqvist et al 2002 Figure 6 12 on the facing page shows the basic structure of the readout and control soft ware The level of all hardware accesses are realised by device drivers First experience using the FED was gained in 2000 based on a Creative Electronics Systems CES RIO system with a Unix like real time operation system called Lynx This platform utilises a PMC Power PC central processing unit CPU mounted on a VME card On the card a second PMC slot is available and was used for the FED For this configuration a basic application and the corresponding drive driver was available from the CMS test beam group For communication and control of the FEH a special prototype of the FEC was used This so called FECO was in principle able to generate the needed communication commands as well as clock and trigger sequences but it had serious faults and was only partially functional This FECO was replaced by a first prototype of the SEQ based on a VME development board and an 12C card While the first SEQ prototype was easy to implement this was more difficult for the 12C card see Sec 6 2 9 on the page before due to the fact that all PMC slots on the CPU board where occupied Therefore Midas20 adapter board is used to run the 12C card in the VME bus system and a corresponding device driver for this combination was written The Midas20 board uses a Universell bridge Universell to connect the VME bus to a PCI bus on bo
156. ika und Qualit tsstudien 0 0c eee eee eee VI 7 8chlu folgerungen vt 2 ae a da an un X 1 Physik am LHC Die Skala auf der die elektroschwache SU 2 x U 1 Wechselwirkung auf die schwache unf elektromagnetische gebrochen wird wird allgemein mit der Higgsmasse identifiziert Obwohl das Standardmodell exakte Vorhersagen ber die Produktions und Zerfallskan le macht erlaubt es keine genaue Vorhersage f r die Masse des Higgsbosons Allerdings wer den mit zunehmender Higgsmasse die Selbstkopplungen und die Kopplungen mit den W und Z Bosonen gr er welches auf der Energieskala des LHC entweder die Entdeckung des Higgsbosons oder die von neuen Strukturen im dynamischen Verhalten der W W und ZZ Wechselwirkungen garantiert Weiterhin bietet der LHC die M glichkeit ver schiedene supersymmetrische Erweiterungen des Standardmodells zu testen und nach Er weiterungen im elektroschwachen Sektor zu suchen die sich z B in der Pr senz von Z W manifestieren Hypothetisch aber sehr spektakul r ist die Aussicht auf ein Hinweis auf kleinere Strukturen und h here r umliche Dimensionen Aufgrund seiner Schwerpunktenergie von ys 14TeV und Luminost t von 10 cm s7 erlaubt der LHC die Wechselwirkungen und Eigenschaften schwerer Quarks pr zise zu vermessen Die enorme Anzahl an top Quarks welche mit den beiden gro en Mehrzweckdetektoren ATLAS und CMS untersucht werden k nnen erlaubt selbst die Vermessung seltener
157. important for efficient background suppression 2 High Energy Physics at the Large Hadron Collider 7 Branching Ratio Figure 2 2 Branching ratios for the main decay modes of SM Higgs bosons Hagiwara et al 2002 l 500 m GeV 2 2 2 Intermediate and high mass Higgs A second high precision discovery channel offered by the SM is ZZ decays into four charged leptons electrons and muons This channel can be used for searching on a broad mass range of My 130 600 GeV c Below 2Mz the event rate is small and the background reduction more difficult as one or both of the Z bosons are off shell In this mass region the Higgs width is small lt 1GeV c and so lepton energy and momentum resolution are of great importance in maximising the significance of a signal Below the ZZ threshold the main background arises from t Zbb and continuum ZZ Zy production The tf background can be reduced by lepton isolation and pair invariant mass cuts which surely requires excellent vertex reconstruction The Zbb background can be sup pressed by isolation requirements The ZZ process is an irreducible background The H ZZ 4 channel is sensitive over a wide range of SM Higgs masses from 2M z upwards to about 400 GeV c with 10fb and about 600 GeV c with 100fb For lower SM Higgs masses the width is quite small and precision lepton energy and momentum reso lution are helpful for larger masses the natural
158. including nested loops 80 6 4 Fast Test Station T H Sensors Figure 6 15 The Karlsruhe Fast Test Station FTS hardware setup is based on an Karlsruhe readout system with additional functionalities like fast cooling added Within the readout PC the FED and the PC card are placed which are connected via the ARCS repeater card to the FEH Furthermore the PC hosts a MIO card which is used for slow control purposes Communication with the motherboard is realised via the PCs parallel port The motherboard carries the Karlsruhe readout components An additional control card is implemented to run the peltier based cooling system which is connected to the power pack in between Last but not least scintillators with photomultipliers and NIM logic allows comics usage 6 4 Fast Test Station The Fast Test Station FTS is the quality control station for detector modules at the bonding centre Functionality tests are performed here as well as complete module qualifications Therefore the design of the FTS focuses on module qualification which naturally includes the functionality test abilities needed The chosen qualification procedure includes an active cooling cycle which has to follow strict timing requirements arising from the given output flow of the bonding process These points drive the mechanical and electrical layout of the FTS based on the aspects collected in Sec 6 on page 61 By chosen reflects that the CMS testing procedures re
159. ing ratio into common decay channels 2 5 2 CP Violation in the B System Similar to the Kaon system the B and B are mixtures of CP eigenstates Bi B BYN and By B B V2 of eigenvalues 1 respectively 1 B pB qB B eB y 1 lel _ 1 B pB qB B eB y 1 lel p 1 e yMa ali Mi iT ce a BRE Mae y Mi 3032 Mio r Indirect CP violation in B0 B mixing arises if Pp Pr ezo 2 15 q which results from the fact that the physical flavour eigenstates Bp and B are different from the CP eigenstates B and Ba Taking the time depended decay rate B t gt f N f B t of an initially tagged B into some final state f where N f is time independent normalisation factor leads to the decay amplitudes 2 F 2 16 2 50 Ar BOP A UB where Ar corresponds to the decay rate of B If the final state f isaCP eigenstate with its CP conjugate state f Ar and Ar are defined in the corresponding way Finally the quantity q Af a tes 2 17 DA plays a central role in CP asymmetries and other observables in B mixing The three quantities y P are independent of phase conventions and the only complex quantity here is Af Its phase is a physical observable If any of these quantities is not equal to 1 or 1 for A then CP is violated in the particular decay ie oe 2 18 CP violation in mixing g pl 4 1 If p
160. into a B or vice versa Measurements give an average Amp 3 07 0 12 x 10 eV which is a hundred times larger than the corresponding Amx in the Kaon system Once the neutral B mesons are produced in pairs their semileptonic decays inclusive or exclusive provide an excellent method to measure the B op mixing From their respective quark contents B decays into a positive charged lepton while B goes into a negative If B and B do not mix the produced pair BY B would have a distinctive signature of a dilepton with opposite signs 0 Therefore a fully reconstructed u 1 same sign event would unambiguously demonstrate the conversion of a B into a BY requiring a sufficient tracking precision to reconstruct the secondary and primary vertices unambiguously For the neutral BO and B system the off diagonal term T12 is extremely small since the overlap in the decay products of B and B is rare Indeed the B decays mostly into anti charmed and unflavoured particles described by b d u while the Be decays into 12 2 5 B Physics charm from b c d u These final decay products are completely distinct There exist only a few common channels into both B and B decay There are BP D D or nt r B coming from b gt c d or b gt u d u However their rates are suppressed by V Veal or VubVua Experimentally only an upper bound lt 107 is known for the branch
161. inzu kommt dass auch mit leichten Kratzern auf den Sensoren weitere Pinholes erzeugt werden k nnen Hierbei kommt zum Tragen dass bei den CMS Sensoren der Aluminium Auslesestreifen breiter als das unterliegende Implantat ist Hierdurch tr gt der Auslesestreifen pr gnant zur Feldformung bei welches zu einer h heren Spannungsfestigkeit der Sensoren f hrt Wird nun allerdings durch einen Kratzer eine Spitze erzeugt so ndert sich auch die Feldkonfiguration Die resultierende Feldspitze kann insbesondere falls auch die Oxidschicht zwischen dem Implantat und dem Auslesestreifen verletzt wurde zu einem Durchbruch der isolierenden Kopplungskapazit t f hren welches dann in einem Pinhole endet 7 Schlu folgerungen Die ersten Erfahrungen mit Prototypen des CMS Spurdetektors verifizieren das ro buste Design der Module Die Zahl der rauschenden oder defekten Kan le ist deutlich unter der geforderten Quote von 2 Verbesserungen des Moduldesign ergeben sich aus einem Randkanalrauschen welches durch eine Filterkapazit t deutlich reduziert werden kann und aus der Sensitivit t der APV25 auf DC Kopplungen via Pinholes Im letzteren Fall wird die Spannungsversorgung modifiziert Hierdurch wird die Empfindlichkeit auf Pinholes der Auslesechip reduziert allerdings besteht weiterhin die Notwendigkeit in der Qualit tskontrolle alle Pinholes zu identifizieren und von der Auslese zu trennen Beide nderungen erfolgen aufgrund der in Karlsruhe ausgef h
162. ion DTS results in a small angular acceptance limiting the inclination of tracks to angles below 3 4 Thus as first order approximation the effective thickness of the sensor seen by cosmic ray tracks can be taken as the perpendicular sensor thickness Figure 7 15 shows the distribution of cosmic ray tracks over a module as measured within the DTS The flatness of this distribution is sensitive to the angular acceptance and reflects the chosen configuration Within the DTS radioactive sources can also be used for signal studies and for calibration Usage of low energy rays requires the placement of the scintillators directly below the module which results in a huge angular acceptance Therefore the limiting factor for the tracks inclinations is the source collimation Unfortunately source collimations are typical poor and the Sr source used is packaged in a plexiglass cover of 5mm thickness which has a drilled collimator hole of 1mm diameter This results in an opening angle of 11 3 Using as approximation a Gaussian distribution to describe the emitted particle angles results in an increased average sensor thickness seen by the particles of 2 4 Figure 7 16 on the next page shows a cluster position distribution seen with the Sr source The FWHM of 34 2 channels corresponds to a geometrical width of 6 2mm which fits to the used sensor to source distance of approximately 2cm and with the sources opening angle 7 4 2 2 Timing Jitter Th
163. ion mode differs slightly but the overall picture stays the same for both The level of deterioration is of the order of 10 while the Peak modes deterioration turns out to be three times larger 7 6 Infrared LED Studies One option implemented in the Karlsruhe test station is the use of infrared LEDs for controlled pulsed illumination and for artifical leakage current generation see also Sec 6 2 5 on page 67 The wavelength of the LED used determines the penetration depth into the silicon Within the FTS LEDs of 950 nm wavelength are available which have a penetration depth of about 80 um while within the DTS another additional type of LEDs with a wavelength of 1050 nm is available The latter kind of LEDs penetrates the sensors completely Both types of LEDs have a rise and fall time of 10ns comparable to the signal collection time in silicon 7 6 1 High Leakage Current Behaviour of Modules The discussion of the different noise contributions of Sec 4 3 3 on page 41 predicts an increase in the module noise due to the reverse bias current In Tab 4 5 on page 44 are the different contributions from the reverse bias current scaling from 0 054 A to 1mA at 10 C to the noise given While that contribution can be neglected for low leakage currents the increased current dominates the noise for Peak mode which rises by about 50 Figure 7 29 on the facing page shows the measured noise for different leakage currents induced by infrared LEDs
164. ional surface damage induced by ion ising particles Both IQCs will irradiate with a fluence of 1 6x10 1 MeV equivalent n em as expected for 320 um sensors Before irradiation the device under test DUT is qualified at room temperature These measurements extract bulk parameters like leakage current Ileak and full depletion voltage Viepi using the IV and CV characteristics Furthermore measurements of inter strip resistance Rint inter strip capacitance Cint poly silicon resistance Rpoly coupling capacities Ce and strip leakage currents styip are per 50 5 3 Quality Test Strategy for the Front End Electronics Figure 5 5 Stacker for irradiation with five standard test structures front and two full size sensors back Dierlamm 2003 formed before and after irradiation at operation conditions of CMS Vyas 400V and T 10 C Irradiation itself will be performed on biased detectors in a dry environment at temperature of 10 C However due to the large amount of charge carriers during irradiation heat dissipation forbids operation at bias voltages of 400 V Important for surface damage is the electric field over the oxide during operation The voltage drop over the oxide will be about 1 V at the end of the experiment due to an increased leakage current of about 1 uA per strip at the bias resistance of 1 MQ Therefore a voltage of 1V should be applied to the oxide layer during irradiation which can be realised due to
165. ionality tests an internal calibration logic is implemented Heier 2001 192 storage cells x 128 channels shapers clock de skewing mechanism that allows to phase shift the clock signal up to a maximum of 25 ns in steps of 1 04ns A trigger coarse skew compensation function is also implemented allowing to delay the L1 trigger signal up to a maximum of 15 LHC clock cycles Finally an auto calibration logic element sets the PLL to optimum bias conditions at start up This auto calibration mechanism is transparent to the user and can be controlled by an I C bus protocol I C interface for testing purposes Placidi et al 2000 The I C interface uses four consecutive I C addresses to access the internal registers controlling the clock and trigger delays and storing status information of the auto calibration circuit 4 2 1 2 APV25 The analogue pipeline voltage chip APV25 is a 128 channel readout chip built in a 0 25 micron complementary MOS CMOS process Each channel consists of a low noise charge preamplifier followed by an inverter stage see below The preamplifiers output inverter is supplied through a common resistor of all channels which results in an on chip common mode CM suppression see below The inverter stage is followed by a CR RC shaper with a peaking time of 50ns Taking minimum ionising particles MIPs generating about 25000 electron hole pairs in 300 um silicon the total voltage output of the charge preamplifie
166. ionality of laser diodes with very high threshold currents as a conse quence of radiation damage The laser diode biasing scheme current sink is compatible with the use of common anode laser diode arrays A combination of linearisation methods allows achieving good analogue performance 8 bit equivalent dynamic range with 250 MHz band width while maintaining wide input common mode range 350 mV and power dissipation of 10mW channel Input signals are transmitted to the laser driver using 100Q matched transmission lines The driver is optimised for analogue operation in terms of exhibiting good linearity and low noise However input voltage levels are compatible with the digital low voltage differential signal LVDS standard 400 mV over 1000 The gain can be chosen from 4 preset values Gain control provides an extra degree of freedom for optimally equalising the CMS Tracker readout chain The integrated circuit IC modularity is 3 matching FEH with 6 APV25s Cervelli et al 2001 4 Silicon Strip Tracker and its End cap Modules 35 4 2 2 2 Analogue Opto hybrid The analogue opto hybrid circuits are required to carry the electro optical components situated in close proximity to the detector FEH Due to limited space and material budget the whole circuit is placed on a printed circuit board PCB of 23 x 30mm size Apart from a linear laser driver LLD chip the number of laser diodes mounted matches the number of used APVMUX outpu
167. iple all nodes on the network are equivalent the network monitoring and ini tialisation functions are generated only by the FEC One of the main functions of the control chain are clock and trigger distribution The FEC receives timing and trigger informations 36 Electrical connection 4 2 Readout and Control Chain T1 Control Module O E and E O Conversion Figure 4 13 The FEC and CCU token ring architecture CMS Collaboration 1998 SCL 2 gt 3 gt 16 x 12C Buses JTAG Master __ Trigger CLKI A Ir 29 Decoder CLKI B Trigger Counter CLKO A amp other timing logic DO A DI A CLKO B DO B Temes DI Memory Bus Interface Ext Reset CET Aa D 0 7 A 0 15 R W CS Local Bus PA 0 7 PB 0 7 PC 0 7 PD O 7 Figure 4 14 The CCU block diagram Marchioro et al 2002 from the TTC system and encodes it in a way suitable for the CCUs using a special timing trigger control receiver ASIC TTCrx The current prototype of the FEC is based on a PMC and able to control one digital link available electrically as well as optically The final FEC version will be a 9U VME board housing control for several control rings The data links and the embedded electronics connected are synchronised to the LHC clock frequency and the data links have a raw capacity of 40 Mbit s An important function which will be implemented in the final FECs is the emulation of the APV25s front end pipeline logic t
168. irrored wavelength shifting WLS fiber inserted The other ends of the fibers are fused to clear fibers The light is detected by photodetectors that can prove gain and operate in high axial magnetic fields proximity focused hybrid photodiodes 3 4 3 Hadronic Forward Calorimeter The very forward region from y 2 75 to n 5 25 will be covered by the hadronic forward calorimeter HF allowing both the measurement of missing transverse energy and forward jet tagging Operation at such high pseudo rapidity requires the use of a technique that is radiation resistant to giga rads faster than bunch crossing time and insensitive to high am bient radioactivity These requirements can be achieved with quartz optical fibers embedded 3 Compact Muon Solenoid 21 absorber material Shower particles produce light in the fiber by the Cherenkov effect gener ating a signal at the speed of light 3ns duration The quartz HF meets the challenge of operation in the extremely hostile radiation environment with event rates up to 16 per bunch crossing The quartz fibers are inserted into grooves in a copper matrix The quartz HF is only sensitive to relativistic charged particles Hence it does not see low energy neutrons which will traverse it in large numbers Furthermore high purity quartz is one of the most radiation hard substances known Hence the detector is largely insensitive to the effects of induced radioactivity 3 5 Muon System The CMS mu
169. irst silicon sensor e between the two sensors Modules of rings 5 7 of the end caps and the outer barrel modules While the position of the missing bond should be deducible from the noise figure in principle it turns out that this is practically impossible The required careful fine tuning of the tagging cuts which are dependent on the module type and the test systems CM contribution cause a significant mistagging rate Using the LED system compare Sec 6 2 5 on page 67 a direct observable for a proper connection is available if the arrays are placed properly above the different sensors Fig ure 7 42 shows the result of a LED test indicating clearly the position of the missing bond Table 7 4 summarised the signature for missing bonds 7 Test System Performance and Module Qualification Studies 113 SENSOR APV a Test input Rinv 1000 Vimp O fav Istrip Rpinhole Test V125 8 IN Sen Y YAZ J o T cap C a N this point LA e des L N common for o all 128 chans det Rpoly 1 8 MQ R a A bias gt DCU 75V Rret 24kQ VSS base line restorer Figure 7 43 Pinholes and the APV25 circuit 7 7 4 Pinholes As pinholes replace the AC coupling of the APV25 by a DC coupling the readout chips have to face a constant current which is dependent on the potential attached to t
170. is a collection of compiled code that can be linked either at the end of a compilation or during start up of a application depending on static or a dynamic linking The use of dynamic libraries has several advantages and is preferred in general For example during development of big application a bug fix in a shared library only requires the recompilation of the library and not of the complete program On the other side there are limitations for dynamic libraries like restrictions for semaphore usage Figure 6 13 on the following page shows a snapshot of the CVS tree holding all the different libraries in their own folder A more detailed description of the different libraries is given in App C on page 130 6 3 3 Threads The level above the libraries is a multi threaded application called KaRinA The applications main function creates all its subprocesses as separated threads It contains three main threads the readout thread which contains all functionalities to control and readout data from the module the analysis thread which is a client of the readout thread shared memory with its block structure and finally the SlowControl thread responsible for all slow control related tasks like the temperature control or the interlock systems compare Fig 6 12 on page 75 Each of them runs independently and they are controlled by the GUI Therefore each of the main threads has a communication thread attached which performs the communication with the
171. is connected to the cooling system of the FTS It utilises peltier elements whose cooling power can be controlled by the voltage applied to them This means one either has to use a low voltage power supply with an individual channel for each peltier element and which is fully controllable by a PC or one builds a switching device which can change the connection scheme of the peltier elements for all in parallel to all serial in several steps The latter option is chosen Weiler 2003 The power pack hosts up to 5 rechargeable lead acid batteries of the type Panasonic LC R0612P which provide 6 V and 12 Ah For the cooling ramps external power supplies are used see Fig 6 8 The peltiers are connected on the front side of the power pack This turned out to be problematic due to a mechanical workshop next door 70 6 2 Hardware Components E Hag K ka me hi ERDE aa KOCMS Muitiplexer Q ihc hehehe A eS E S621 way FED j O 8 Bit Data a o E Select PLD 5 switch switch switch Select MRON G6H 2F JOMRON G6H 2F OMRON G6H 2F 8 Dik JK J Strobe D ES E O OEC ALTERA gt gt MAX7064 SLC44 6 Acknowledge g Figure 6 9 Karlsruhe multiplexer device is used for module long term test as well as for petal tests From the left the pig tails of the OEC are plugged to the multiplexer which connects one input to the output on the right towards t
172. issing Bonds are very similar to broken Al readout strips expect for the leak age current The capacitive load of the APV25 amplifier will be reduced There are up to three positions where a missing bond connection can occur e between the two sensors e between the sensor and the PA e between the PA and the APV25 In the Bonding Centres only the first two should be be found Missing bonds between PA to APV25 should be found and repaired by the FEH Bonding Centres pinhole Al SIOJ SIN N Figure 5 14 l Schematic of p implant pinhole fault in silicon strip detectors which replaces the capacitive coupling of the Al E field readout strip by an Ohmic con nection n layer n bulk 5 5 3 5 Pinholes are low ohmic lt 1 MQ contacts between the pt implant and its cor responding Al readout strip as sketched in Fig 5 14 To understand the impact of pinholes dielectric shorts or C shorts what they are also called sometimes one has to study the be haviour of am APV25 channel DC coupled to the p implant connected via the metal readout strip In this case the APV25s amplifier input sees through the pinholes resistance Rpinhole the potential of the p implant strip which depends on the leakage current drawn by the individual strip I strip The circuit diagram is shown in Fig 5 15 The current has to pass the individual poly silicon resistor Rpoly and as part of the total module leakage current Teak it has also to pass the r
173. itance of the lines Furthermore the APV25 itself adds a channel dependent resistance to the resistive load of its field effect transistor FET of the order of 13 to 70Q Turchetta et al 2001 These resistive loads results in an expected change for the noise calculated with the expression given in Tab 4 4 on page 44 of the order of 0 06 ADC and fits well with the results from Migliore 2002 and Raymond 2001b Both of them found a derivation of up to 4 by this effect Nevertheless the increase of noise towards the edges of the module as shown in Fig 7 5 is stronger The CM corrected noise as shown in the middle plot of Fig 7 5 on the preceding page is more or less flat and its noise RMS results typically in the order of 0 06 0 15 ADC see Fig 7 6 in good agreement with the statistical expectation 7 2 3 Readout Mode Dependence of Module Noise The readout mode itself has an impact on the modules raw noise compare Figs 7 7 to 7 9 on pages 92 93 Beside electronic differences in the signal processing arising from the APSP filter and the inverter stage additional noise effects occur if the calibration circuit is used In section 4 3 3 the effect of the APSP Filter with respect to noise has been quantised as difference between Peak and Deconvolution mode see Tab 4 5 on page 44 The expectation results in an increased noise for the Deconvolution mode in the order of approximately 50 Figure 7 7 on the following page shows the raw noi
174. itter cor SNRmp 34 53 0 37 19 00 0 43 19 00 0 43 Timing jitter cor 1 015 1 075 1 180 SNReorr 35 0 0 4 20 4 0 5 22 4 0 5 ENC meas 986 12 1691 45 1540 45 ENC reale 928 1460 1460 Noisemean ADC 1 97 0 13 2 72 0 17 2 72 0 17 ADC 501 33 621 39 566 35 Table 7 2 SNR values measured for Peak and Deconvolution mode The MIP signal in a 500 um thick sensor 470 um effective creates 34500 electron hole pairs The measured most probable SNRyp value and the timing jitter corrected SNR orr are given For Deconvolution mode a second timing jitter scenario is also shown Furthermore the measured ENC and the predicted ones are given and an ADC calibration is calculated 7 5 Signal to Noise Ratio Studies For a full understanding of the detector response the dependencies of the SNR concerning bias Voltage leakage currents and temperature are studied For parts of the SNR scans a radioactive source Sr is used which results in minor systematical problems comparing the source data with the cosmic rays While the particles emitted by a source are typically no real MIPs their geometrical acceptance significantly differs from that of cosmic rays which results in different cluster size distribution and different systematics mainly truncations arising from the clustering algorithm 7 5 1 Signal to Noise Ratio vs Temperature The signal dependence on the temperature arises from changes in the
175. ks under each components The assembly platform containing the four assembled modules can be removed for glue curing and another assembly platform brought in to start the construction of the next modules The placement accuracy is within 5 um and the assembly rate amounts to approximately 100 modules week Fiore 2001 The assembly of a module is naturally connected to several risks The automated han dling of sensors as well as of the highly integrated FEH has to be performed accordingly Furthermore glueing and mechanical stress can in principle harm the silicon eg producing microcracks Therefore during the commissioning phase of module production and later on sample base an V curve for the individual sensors and a fast FEH functionality test will be performed The V characteristic measurement requires the use of a micromanipulation probe connecting the bias ring to GND before bonding the module 5 4 2 Bonding and Module Quality Assurance Centre The modules are electrically finalised during the bonding procedure in the Bonding Centres BCs For the TEC community there are five bonding centres located at Aachen Karlsruhe Strasbourg Vienna and Zurich Each of these centres takes care of one or two rings compare Fig 5 8 on the facing page Taking the advantages of industrial au tomatic bonding machines see Fig 5 10 each of the BCs can build up to four mod ules per day To the standard program of the BC belongs frequent checks of
176. ll cause a saturation of the APV25 preamplifier including that the inverter transistor is turned on V VSS In Figs 7 44 and 7 45 on the next page the pinhole behaviour in the calibration amplitude and noise are shown The pinhole vanishes in both cases at a leakage current of 30 uA as it is expected for a prototype module where the pt implant potential matches the APV25s virtual ground of 0 75 V in this current region While artifical pinholes produced by a bond connection between the AC and DC pad will have a normal strip leakage current it is expected that real pinholes will vanish at significantly lower leakage currents Particular an inconspicuous behaviour of a pinhole is possible and even likely in the normal leakage current region of 0 4 10 A The dependence of the pinhole behaviour on the leakage current is a unique feature of pinholes and can especially for the calibration amplitude Fig 7 44 be used for tagging resulting in a unambiguous fault identification Another possible strategy for pinhole detection is based on forward biasing Applying a forward bias voltage of 9 V results in a leakage current of the order of 250 yA and a potential 114 7 7 Module Fault Detection Studies Figure 7 44 Pinhole search with artifical leakage current and calibration signals re veals a unique pinhole signa a ture For different leakage F currents the calibration ampli
177. ly reduced if the detector is operated at lower temperatures approximately with a factor of 15 from room temperature to 10 C Figure 7 10 shows the IV behaviour at 10 C and at room temperature 22 C The full depletion voltage Vaepl of the sensors used are around 70 V and a change in the IV curves slope is clearly visible Beside the dependence on temperature and voltage additional dependencies occur on relative humidity and previous treatment of the device e g handling in direct sunlight will increase the leakage current temporary Observing the leakage current as a function of time it will decrease strongly At least two different time scales with exponential decreasing behaviour are observed a fast and strong decrease on a several minutes timescale and a slow decrease on a time scale of hours Therefore IV curves results are always an upper limit for the ideal IV behaviour 7 Test System Performance and Module Qualification Studies 95 For modules tested so far typical leakage currents at room temperature scale in the range from 0 4 4A to 10 uA Allowed by the sensor specification are leakage currents for modules with two sensors of up to 20 A below 450V with an additional increase of less than 20 yA in the region from 450 V to 550 V compare Sec 5 2 1 on page 47 and Sec 5 6 on page 59 The leakage current is also sensitive to surface defects like scratches arising from mishandling compare Sec 7 7 4 on page 113 and Fig 7 53
178. m perfect clear to still tag able Visible reflects that a distortion can be seen but its structure does not reflect necessarily the short The experience gained from the limited number of modules tested so far shows that shorts are not very frequent and most of them are introduced already before bonding of the module on the PA level 7 7 2 Broken Strips Although broken strips have been reported from several collaborations as a relevant fault type Hartmann 2000 and they are also found by the QTCs no module with this kind of defect has been tested so far Broken metal readout strips may be created later by handling which is discussed in Sec 7 9 on page 119 The metal overhang technique used for the CMS silicon sensors will cause that most of the broken strips may result in a noisy behaviour which is similar to scratched channels compare Sec 7 9 For a discussion of the expected signature for broken strips see Sec 5 5 3 3 on page 57 7 7 3 Missing Bonds A missing bond does not mean literally the absence of the complete bond but reflects the missing connection typically due to a failed bond foot Figure 7 38 on the following page shows a missing bond where one side did not hold Missing bonds are closely related to the bonding quality The automatic bonding machines are very sensitive to problems during the 110 7 7 Module Fault Detection Studies QM AU Dunes Figure 7 38 Picture of a missing bond The remnan
179. med externally Eagle is an acronym for Easily Applicable Graphical Layout Editor 6 Karlsruhe Test Stations 63 Power supply HV card Multiplexer Sequencer LED controler Motherboard Interface card Figure 6 2 Karlsruhe motherboard with all cards mounted 6 2 1 Motherboard The Motherboard has in total 10 slots of which two are reserved for special purposes compare Fig 6 3 on the next page The first one is specialised for input output connections to e g a PC via an interface card Separating the interface from the motherboard itself allows usage of different connection types like I C parallel port or digital input output DIO cards The last slot is reserved for a power regulator card supplying the motherboard with stabilised 5 V 3 3V and 5V All application slots 0 7 are served with all power lines a data strobe DS an acknowledge ACK a reset RESET and the eight data lines D 0 7 On the motherboard inversed logic levels are used which are compatible to the standard used for the parallel port of PCs From the specialised interface slot additional lines are connected to a PLD used to se lect deselect the application slots by their SEL lines Therefore the PLD is connected with the interface slot by additional select SELECT data strobe DS acknowledge ACK and reset lines If the select line SELECT is set high by the communications counterpart the PLD will check if the four MSB 7 4 are eq
180. men verantwortlich sind DETECTOR BIAS SERIES AMPLIFIER RESISTOR RESISTOR PULSE SHAPER Ry ens MW O ena N Rp Ny DO Ge gt ind Abbildung 4 Equivalenter Schaltkreis zur Rauschanalyse Hagiwara et al 2002 Alle diese Rauschquellen finden sich auch auf den Modulen wieder vgl Abb 4 Hier bei unterscheidet man im Allgemeinen zwischen Rauschquellen welche parallel und seriell zum Vorverst rkereingang geschaltet sind Parallel geschaltete Rauschquellen verhalten sich wie Stromgeneratoren w hrend seriell geschaltete Rauschquellen als Spannungsgen eratoren modelliert werden In Tabelle 1 sind diese Rauschanteile f r SST Module des Endkappenrings 6 bei 10 C angegeben Zusammen mit dem Signal von MIPs ergibt sich ein SNR von 29 9 im Peak Modus und 19 2 im Deconvolution Modus Rauschquelle Art Peak Modus Deconvolution Modus Leckstrom parallel 8 1068 3 480 Vorspannungswiderstand Ryory parallel 165 74 Auslesestreifenwiderstand R seriell 425 627 Ausleseelektronik seriell 808 1316 Summe 928 1415 1460 1537 skaliert von 0 1 A f r nichtbestrahlte Module auf bis zu 1mA nach 10 Jahren LHC Betrieb Tabelle 1 Rauschanteile f r Ring 6 Module in quivalenter Elektronladung Berechnet f r T 50 ns inb 0 05 uA 1mA homogen verteilt ber alle 512 Auslesekan le und beide Sensoren Rpoly 1 85 MQ Rs 220 Q Crot 15 6pF 4 Qualit tskontrolle f r CMS Spurdetektor
181. mes Automated procedures based upon high precision robotic positioning machines gantry are used for module assembly These gantries allow a placement accuracy of all components on the CF frame within 5 um Surrow 2001 However the production of the first express line proto types showed a root mean square RMS accuracy of 50 um Schwerdtfeger 2002 which has been improved by a new calibration of the gantry systems 24 4 1 Silicon Strip Tracker Modules For the tracker end caps the integration into larger structures is done in two steps As a first step radial sections of a disk called petal are integrated Each disk consists of 16 petals of which eight are mounted on the front side and the other eight on the disk s back side The individual modules will overlap in both r and amp to guarantee a hermetic sensitive detection area The petals basic honeycomb structure has cooling pipes embedded which are connected via inlets to the electronic components hosted on the petal modules FEH optical hybrids and CCUMs These inlets have to be placed with high precision because they are also used as fixation points for the modules providing at the same time a good thermal contact between cooling pipes and heat sources The total thermal load of a single petal will scale from 44W before irradiation up to 88W after 10 years of LHC operation Mounted on each petal is an interconnect board ICB which serves low and high voltages to modules as w
182. minated by gluon fusion where the superscript denotes the electric charge Thus one gets a valid perturbative field theory in term of the redefined field H It represents the physical Higgs boson while the three other components of the complex doublet field are absorbed to give mass and hence longitudinal components to the gauge bosons Depending on the Higgs mass different decay channels are optimal for signal detection 2 2 1 Low mass Higgs For SM Higgs masses just above the LEP limit of My gt 114 5 GeV c and below 2Myy the dominant decay mode is into bb compare Fig 2 2 on the next page which is rather unrealistic to extract because there is no trigger for this process to disentangle it for the enormous background of Quantum Chromodynamics QCD bb production For this reason the decay into yy is the most promising in this mass region Also the branching ratio is very small and there is large background from pair production of photons via 47 YY gg yy and bremsstrahlung processes Excellent photon energy resolution is required to observe this signal which drives the very high quality electron calorimeter ECAL development for ATLAS and CMS For SM Higgs masses up to 140 GeV c this channel seems to be the most promising A full coverage of this mass region requires a run at high luminosity and detector performances as designed or better Kunszt 1997 If the Higgs boson is lighter than 130GeV c another promising chann
183. module Produktion und Bau der SST Module werden sich ber einen Zeitraum von ca zweiein halb Jahren ziehen wobei sich die Arbeiten auf ber zwanzig verschiedene Institute weltweit verteilen Hierbei ist eine vollst ndige Qualit tskontrolle der gesamten Produk tionskette essentiell um Fehlerquellen fr hzeitig zu erkennen und zu eliminieren Hierzu VI Zusammenfassung wird ein umfangreiches Qualit tssicherungsprogramm eingesetzt welches sowohl Funk tionalit tspr fungen f r alle elektronischen Komponenten beinhaltet als auch umfan greiche Tests auf Sensorbasis Letztere umfassen sowohl Qualifizierungstests wie auch Proze und Bestrahlungskontrollen Die Detektormodule werden auf automatischen Montagerobotern mechanisch zusam mengebaut bevor sie in den Bondingzentren mit Industriebondern elektronisch vervoll st ndigt werden Da diese Module die kleinste unabh ngige Detektoreinheit darstellen ist deren Qualit tspr fung f r die Gesamtqualit t des sp teren Spurdetektors mitentschei dend Im Rahmen der Detektormodulqualifizierung werden elektrische und mechanische Tests durchgef hrt W hrend Letztere sich auf mechanischen Stre innerhalb eines K hl zyklus von Raumtemperatur auf 10 C beschr nken umfassen die elektrischen Tests eine Vielzahl von Einzelmessungen Hierbei spielen insbesondere Rauschmessungen die interne Kalibrationseinheit und Messungen mit externem Licht von IR LED eine gro e Rolle Das Rausch
184. mperature increase of 11 C the leakage current is also a strong function of temperature B II Charge carrier velocities While for low electric field the drift velocity of charge carrier within silicon is proportional to the applied field This holds no longer if the drift velocities approaches the thermal velocity of 10 cm s for room temperature For high electric field the experimental results can be approximated by the empirical expression for holes 10 cm s 4 1 2x10 Lh B 4 Vp valid for Si at room temperature B III Depletion Voltage and effective Doping Concentration The depletion voltage of a silicon detector depends upon the effective doping concentration of the substrate material i ed Vai y Nest B 5 ESi where d is the depth of the diode The value of Nerf is determined by the concentration of space charge in the depletion region In non irradiated devices the space charge arises predominantly for the phosphorus dopant APPENDIX B SILICON STRIP DETECTOR CHARACTERISTIC 129 Irradiation results in an accumulation of negative space charge in the depletion region due to the induction of acceptor defects which have energy levels deep within the forbidden gap Therefore n type detectors will go through a type inversion during irradiation and becomes an effective p type where the inversion fluence depend strongly on the initial resistivity of the substrate material 130 APPENDIX C KARLSRUHE READO
185. n PhD thesis Universitat Karlsruhe TH 2000 F Hartmann The CMS all silicon tracker strategies to ensure a high quality and radiation hard silicon detector Nucl Instrum Meth A478 285 287 2002 F Hartmann 2003a private communication REFERENCES 147 F Hartmann The silicon sensors for the compact muon solenoid Design and qualification procedure submitted to elsevier science 2003b S Heier Entwickung einer Teststation f r CMS Mircostreifendetektoren Diplomarbeit Uni versit t Karlsruhe TH Institut f r Experimentelle Kernphysik 2001 S Heier 2003 private communication A Herve et al Status of the CMS magnet 2001 DAPNIA STCM 01 10 A Honma Technical specifications and procedures for the assembly bonding and testing of the front end hybrids CMS Tracker Hybrid Assembly Group 2003 P Horowitz and W Hill The Art of Electronics Cambridge University Press 1989 FHIT FHIT Quick reference guide Hybrid Working Group 2002 iseg BPx 10 305 12 HV module iseg http www iseg hv de 2002 D D Johns and K Marin Analog Integrated Circuit Design John Wiley amp Sons Inc 1997 L Jones APV25S1 User Guide Version 2 2 RAL July 2001 L L Jones et al The APV25 deep submicron readout chip for CMS detectors 1999 Prepared for 5th Workshop on Electronics for the LHC Experiments LEB 99 Snowmass Colorado 20 24 Sep 1999 D I Kazakov Beyond the standard model in search of supersymmet
186. n a decrease of induced damages is observable This effect is called anneal ing although true annealing in which the crystal becomes perfect again does not exist But in many cases the defects induced by irradiation are not stable and may just be transformed into another more stable defect during the annealing phase As new defect complexes are produced the effect of annealing may not always be beneficial for the detector performance which is called reverse annealing Measurements show an exponential behaviour of the beneficial annealing effect with time while reverse annealing slowly increases with time and saturates much slower After some time annealing effect goes through a minimum and increases slightly again afterwards The relevant time scales for both beneficial and reverse annealing are dependent on the temperature and decrease with increasing temperature Due to this irradiated sensors are kept under cold conditions to prevent reverse annealing 4 3 3 Noise Analysis For signal detection noise has to be taken into account as well as signal creation Hereby two different types of noise have to be distinguished First we may get interference noise as a result of unwanted interaction between the circuit and outside world like power supply noise on ground wires or electromagnetic interference This kind of noise can be significantly reduced by careful circuit wiring and layout On the other side we have inherent noise which refers to r
187. n about details of the silicon behaviour and their quality control Most of the work presented in this thesis would have been impossible without the team work of the Karlsruhe readout group Therefore and for the good time we had I have to thank the PhD students Stefan Heier Thomas Weiler and Manuel Fahrer as well as my diploma students Wolfgang Schwerdtfeger and Markus Waldschmitt I have to thank Alexander Dierlamm and Alexander Furgeri for all our discussions about silicon detector behaviour and the details on QTC and IQC measurements Tobias Barvich was always a great help concerning mechanical topics and I learned a lot thought our discussions of the ideas and designs incorporated in the test station For her guidance through the bureaucratic jungle I have to thank Edeltraut Haas and also I have to thank Diana Fellner Thedens for her helping hand concering travel arrangements Among all the collaborators special thanks go to Dr Gigi Rolandi Dr Ariella Catai Dr Marco Meschini Dr Lino DeMaria Dr Patrice Siegrist and Prof Dr Wolfgang Braunschweig
188. n by counting the moved steps Schwerdtfeger 2002 6 5 1 1 An Optical Microscope may be attached to the gate systems head to perform visual inspections The available microscope allows a magnification of up 40 which is sufficient to detected bad wire bonds and to estimate the width and depth of scratches 6 Karlsruhe Test Stations 85 6 5 1 2 The Laser System available at the DTS uses a 1050 nm laser with rise and fall times of Ins The amount of light can be variated from an one MIP equivalent signal up to highly ionising particle HIP equivalent ones The laser is triggered by the SEQ while the control has to be done manually 6 5 1 3 A Radioactive Source can also be used within the DTS Therefore two scintil lating fiber arrays are placed below the module which can detect particles by coincidence A Sr is used because of its high energy decay electrons which can penetrate the module as well as the fiber arrays 86 7 Test System Performance and Module Qualification Studies Essential for detailed studies of the SST modules behaviour and performance is the under standing of the test systems characteristics especially for its noise contributions After the reduction of parasitic noise sources from the readout system to a negligible level the equiv alent noise current ENC of the system can be calibrated using a known signal for which typically MIPs of the cosmic radiation are used Based on the ENC calibration the module
189. ncluding DMA FED Manual an interface electrically compatible to the popular PCI bus allows the FED to be used as well on a wide variety of commercial off the shelf VME carrier boards as in desktop PCs The PCI interface itself is implemented by using a PCI9080 bridge from PLX PLX 2000 Figure 6 10 shows a block diagram indicating the basic functional units of the FED Each of the 8 input channels utilises a commercial ADC type Fairchild SPT7861 Fairchild Inc 2001 and is capable of digitising 10 bits of which the 9 most significant bit MSB are stored at clock speeds between 2 and 40 MHz The captured data is stored in a dual port memory DPM This DPM provides each ADC channel with a 64K sample deep buffer and thus is capable of holding the raw data from up to 256 APV25 data frames at any given time The FED is read out in parallel to the ADC capture over the PCI bus The connecting PCI bridge also supports a direct memory access DMA engine for an optional high speed readout mode Furthermore the FED provides a FIFO storage for event timing information 72 6 2 Hardware Components To PMC PCI9050 PCF8584 EP 680 1118 3004 Los debia TO ae Mer N serial 3 EEPROM PCI bus ann interface 2 N m o 2 SCL Opto Coupler SCL 12C Bus Sa Controler PLX zZ SDA Opto Coupler SDA PRR PCI 9050 5 O A Figure 6 11 Block diagram and picture of PC card Murer 2000 Clock and trig
190. next token The read write bit limits the number of devices with unique addresses on one bus to 128 After each token the slave device has to generate an acknowledge by pulling the SDA line down Finally the master stops the transmission by releasing the SDA line while the SCL is high Thus a complete transmission consists of a start signal a 8 bit token addressing the slave device a slave acknowledge a sequence of read or write tokens connected with acknowledges and finally a stop signal for more details see the I C specification by Philips Philips 1992 F Front End Hybrid schematic In Fig F 1 on the following page the schematic of the front end hybrid FEH is given which is cited several times within the thesis 138 APPENDIX F FRONT END HYBRID SCHEMATIC c17 4 70 T STANL6V gt c16 4 70 STAN16V V gt R3 V125INV MUXOUT MUXOUT TRGI TRGO RST Se 24 ADDO RAL FEB 2000 SCLI ALLOW_CONNECT TRUE SDA c4 c7 c8 c5 c6 c9 C2 100N C3 100N 100N 100N 100N 100N 100N 100N 0
191. ng companies FHIT ibration pulses FHIT 5 3 2 FEH Bonding and Quality Test Centre The FEH Bonding and Quality Test Centre performs the mounting of the PA onto the ceramic hybrid carrier In most cases this includes also the glueing of one or two spacers under the PA compare Figs 5 7 This is followed by wire bonding from the APV25s inputs to the PA After reception the FEH are visually inspected and tested using a FHIT Directly after the glueing and bonding steps a series of tests are made to ensure the continued correct functioning of the FEH and that the PA has been faultless bonded This includes a thermal cycle to the operating temperature of 20 C with continuous readout Furthermore an input pulser device made of an insulated metal tape which can be placed next to the PA surface can inject a charge into all input channels by means of capacitive coupling The injected pulse is seen as a large signal in every channel and missing signals indicate either a bad readout channel a failed bond or an interrupted trace on the PA During the thermal cycle the FEH will stay powered However the FEH will be turned off when the operating temperature of 20 C is reached After a while it will be powered on again to check the power on condition at low temperature Honma 2003 soft glue DC3140 hard glue araldite 52 5 4 Module Quality Test Strategy Logistic for the Tracker End Cap 11425 sensors 7100 modules includin
192. ng of the FED does not match with the signal saturation Hereby the increased noise issue form a cross talk of the digital header of the APV25 to the first or last channel transmitted in the data frame Obviously this is not a fault corresponding to the module but to the test station used and can easily be remedied 7 7 7 General ASICs Faults Beside the strip errors discussed in the previous sections also ASIC related errors are detected Within the collaboration at least one FEH has been found with a malfunctioning APVMUX chip which did not switch between the two APV25s attached This can easily be detected compare Sec 5 5 2 on page 54 and was not seen so far in our lab Problems due to malfunctioning PLL APVMUX or DCU have not be found on the proto types series But problems with the I C bus are more or less common These arise mainly from bad contacts of the FEHs connector to the readout interface After the frequent mounting and unmounting during the prototype testing the limited number of contact cycles 100 of the FEH connectors results in this kind of problem especially for the readout interface connector which is obviously has the same type of connector General ASIC problems have only been found twice so far In both cases the affected FEH was a special prototypes with six APV25s where the fifth and sixth APV25 had been added later From the FEH design there is no electrical difference between the version with four and a six A
193. ngs in link numbers are achieved by multiplexing data from two adjacent APV25s onto one optical channel with 256 detector channels connected to it Mul tiplexing is done by interleaving data from pairs of APV25s transmitting at 20MHz into a 40 MHz stream compare Fig 4 11 on the next page Each multiplexer APVMUX chip contains 8 APV25 inputs and four differential output channels for 300 um silicon 4 Silicon Strip Tracker and its End cap Modules 33 ODD APV EVEN APV 8 Bit Address e Bit Header ae E ke La LE ia ET LI LA Lon ODD Switch 20 MHz off Amplitude ii LI LE LI Li Li on EVEN Switch 20 MHz off Tine COMPOSITE DATA h 50ns 20 MHz gt ie 25 ns 40 MHz Figure 4 11 Switching between two APV25s by the APVMUX The output of two APV25s is skewed by 25 ns relative to each other to get maximum settling of the analogue levels Murray 2000 To control the switching the clock and trigger signals used by the APV25s are also received on the APVMUX For optimum performance the data stream from each APV25 is shifted by 25ns if its local address is an odd number with even and odd APV25s providing the two inputs so that maximum settling of the analogue levels at the APVMUX inputs is achieved Murray 2000 Furthermore the APVMUX converts the differential current output of the APV25s into voltages by internal resistors Each differential input line is connected eightfold and parallel throu
194. nic calorimeter in the central region 7 lt 3 is a copper brass scintillator sampling calorimeter surrounding the ECAL Like the ECAL it consists of a barrel part supported on rails by the vacuum tank of the magnet system and end caps supported by the magnet end cap return yoke compare Fig 3 1 on page 16 The transition region barrel end cap has been optimised to avoid dead regions and pointing cracks CMS Collaboration 1997b Copper brass has been chosen because the absorber material is required to have a short interaction length since space is limited A low Z number is needed to avoid degrading the muon momentum resolution The material must be non magnetic and should cause only low costs Furthermore with copper it is relatively easy to construct mechanical structures by electron beam welding EBW The barrel calorimeter has 15 scintillator and copper layers inside the magnet coil corre sponding to 6 6 interaction length including the electron calorimeter at y 0 Two more scintillator layers behind the magnet coil increase this interaction length by 4 more units These scintillator segments outside the coil are known as the tail catcher The scintillators are 4mm thick mega tiles of various width The individual tiles are separated by 3 75 mm deep grooves and filled with white glue The back sides of this grooves are painted black for optical isolation For readout each tile is grooved in a keyhole shape near the periphery and a m
195. nitor also implemented in the ISEG module can not be used because it measures the current on the HV return line while for the CMS modules the bias voltage goes via the FEH to GND This makes current measurements on the ISEG modules return line useless in our application Therefore a current measurement is implemented directly on the HV line using an isolation amplifier delivering a voltage proportional to the voltage drop over a measurement resistor in the HV line Both monitor signals for voltage and current can be read out with a MIO card giving a resolution according to a 16 bit or 150nA and 0 2V The HV output voltage is controlled via a 8 bit DAC which results in an output resolution of approximately 4 0 V Furthermore an interlock system is implemented on the HV card us ing acomparator between the current monitor and an adjustable threshold realised by another 8 bit DAC If the current monitor signal exceeds the references voltage a kill signal is gener ated by the PLD which also can be requested by the PC and read out via the MIO card 6 Karlsruhe Test Stations 67 gt 13 a AS WI i O 8 Bit Data Oo PLD 2 e Select oO os JE 2 Strobe E o ALTERA o MAX7064 SLC44 6 Tan S 8 Bit Data Acknowledge gt Strobe lt rr z Acknowledge Figure 6 6 Karlsruhe IR LED array controller card extents the bus system to external cards like the IR LED arrays Furthermore it
196. non excitation is required for momentum conservation For MIPs the most probable charge deposition in a 320 um thick silicon detector is about 23500 electrons 3 7fC and in a 500 um thick sensor about 36 700 electrons 5 8 fC Hereby differs the most probable value Amp for a typical silicon strip Lev Davipovicu LANDAU 1908 in Baku Azerbaijan 11968 in Moscow The work of the Soviet physicist covers all branches of theoretical physics In 1962 the Nobel Prize was awarded to him for his pioneering theories on condensed matter especially liquid helium TLEONARD EULER 1707 in Basel 11783 St Petersburg The famous scientist contributed important parts to modern calculus and to physics like the Euler gyroscope equations hydrodynamics and fluidics as well as works on the principle of least action and variational analysis CARL FRIEDRICH Gauss 1777 in Braunschweig 11855 in G ttingen German mathematician who worked in the fields of number theory geometry astronomy and geodesy He also introduced the bell shaped curve known by his name which is fundamental in the description of statistical data 40 4 3 Expected Module Performance A x MeV gl cm 0 50 1 00 1 50 2 00 2 50 AAA 10F 500 MeV pion in silicon 7 i N 640 um 149 mg cm ost AA ee 320 um 74 7 mg cm H N 160 um 37 4 mg cm E EN 80 um 18 7 mg cm 30 67 iil i A 7 x F Lw L m wN d i 0 4 ER Es 4 Figure 4 17 St
197. nowadays This test consumes only a few seconds 5 5 2 2 Leakage Current Failures are the most critical and difficult detectable ones Although measuring the leakage current is simple there are several parameters to be taken into account First the silicon is very hygroscopic and the leakage currents are strongly affected by this Furthermore the silicon shows a memory effect due to the population of charge traps which also increase leakage current Principally both dependencies can be controlled by keeping the device in a dry atmosphere biased for some time Obviously this is in conflict with the needs of fast functionality tests during production Additionally the modules should not be handled in a too dry atmosphere due to electro static discharge ESD which calls for a relative humidity at a 50 level 5 5 2 3 APV25 Header Problems can be classified in two groups trigger related errors and control related ones The first one will raise the error flag the last bit of the digital header compare Sec 4 2 1 2 on page 29 This indicates a problem within the read write pointer logic of the APV25s pipeline either produced internally or through noise on the trigger line The latter possibility is closely related to PLL problems and should in principle affect all APV25s on the FEH The other class of header problems reflects faults in the internal circuit of the APV25 If the pedestals do not react on 12C controlled bias changes or if the signal
198. o classes to store the corresponding data which include pedestal noise with and without common mode correction common mode histograms calibration profiles and calibration pulse shapes as well as LED and particle run based plots Slow Control library contains a collection of classes connected to slow control tasks These include the MIOread class which controls the access to the MIO the TempMux class controlling the slow control multiplexer the TempControl class used to steer the Peltier Control class and the Hygro class including the hygrometer control The TempControl class builds the heart of the slow control library It incorporates the other slow control classes and steers the cooling system via the PeltierControl and Pow erPack classes This is done by the use of the two methods setTemp and holdPower which changes the target temperature and which evaluates and sets the cooling power needed to reach it Communication library contains everything needed for TCP IP communication with the GUI embedded in the SocketClient class Another class stored in the Communication library is the Config class The Config class is the central object responsible for loading and parsing of configuration files All config files are stored centralised in the config directory on the top level of the CVS tree They define hardware related setting like slot numbers of the cards on the motherboard as well as online analysis settings like thresholds of error tagging
199. o ensure that its output buffer FIFO do not overflow Sec 4 2 1 2 on page 29 by inhibiting the trigger for the small fraction of events that would cause an overflow 4 Silicon Strip Tracker and its End cap Modules 37 4 2 4 2 CCUM The building blocks of the CCU are shown in Fig 4 14 on the facing page The CCU receives the combined clock and trigger signal and passes it to PLL Moreover it implements a 16 channel I C bus master used for the slow control and readout in the front end It also generates the hard reset signal for other chips and includes memory and I O local bus interfaces which will not be used in the CMS tracker Each CCUM will include a PLL chip for clock and trigger separation These signals are sent to an LVDS fanout ASIC Groups of 1 C clock trigger and reset signals are distributed to each front end module through interconnect boards CCUs are typically mounted on Control Modules CCUM housing the necessary auxiliary electronics such as line drivers receivers and level translators In difference to the simplified schematic from Fig 4 5 on page 27 the FECs are connected to several CCUM in a token ring network topology Fig 4 13 on the facing page In the ring a token is initially sent out by the FEC and passed on from one station CCUM to the next A node which wants to send data replaces the token by a data frame which is forwarded until received by the FEC where the packet is modified and passed on until it ret
200. o sensors connected Furthermore the PLD controls a switch 68 6 2 Hardware Components trigger lg Es r N PLD _ CMOS switch 8 1 CMOS switch 8 1 A A Sx lt ON 8x 8x 8 Bit Data e Vbias ALTERA MAX7064 SLC44 6 Strobe _ ARK HKS N E Acknowledge Figure 6 7 As front end of Karlsruhe IR LED system the LED array hosts 64 IR LEDs A PLD controls via eight 8 1 switches which of the infrared LEDs will be pulsed at the next trigger Weiler 2002a which turns on or off the trigger input The trigger itself is received from the SEQ as a LVDS signal and feed through directly to the LED arrays Finally the control cards PLD can switch the front ends power on and off On the LED array front end a second PLD controls a network of fast switches connected to the 64 LEDs them self This network consists of eight 8 1 switches with low ohmic contacts lt 5Q and fast switching time lt 7ns The infrared GaAs LEDs SFH 4301 from Infinion use a wavelength of 950 nm and have a rising time of less than 10 ns Their light cone opening angle of 7 can be reduced by usage of an collimator block Depending on the size of the collimator ten to twenty strips are illuminated by each LED Weiler 2002a The LEDs wavelength of 950 nm is sufficient to induce signals in the silicon sensors but does not penetrate them deeply 80 um Due to this
201. ode correction 2 2 2 2 2 mn 91 Raw noise comparison between Peak and Deconvolution mode 92 The inverter stage influence on raw noise for Peak mode 93 Inverter stage influence on raw noise for Peak mode with calibration circuit used 93 Module leakage current versus bias voltage at 10 C and 22 C 94 IV curve for a single strip breakdown 2 2 2 2 on Emm 95 Calibration pulse shape in Peak and Deconvolution mode 96 Calibration amplitude in Peak without inverter mode 2 2 2222220 96 Calibration amplitude and cross talk es 96 Hit map of cosmic ray detection 2 2 o En nn 97 Hit map of or detection ns ta en de Bis 98 Cluster size of cosmic ray measurement 2 2 2 2 Emm 99 Cluster size of Sr measurement ea 3 eek ins aan a 99 SNR in Peak mode at 10 C 2 Con nn 100 SNR in Deconvolution mode at 10 C 2 LE CE En 100 SNR vs temperature in Peak mode cosmic rays o 102 SNR vs temperature in Deconvolution mode cosmic rays 102 SNR in Peak mode as function of bias voltage cosmic rays 102 SNR in Peak mode as function of bias voltage Sr 2 22 2222 102 SNR in Deconvolution mode as function of bias voltage at 22 C Sr 103 SNR in Deconvolution mode as function of bias voltage at 10 C cosmic rays 103 SNR in Deconvolution mode as function of leakage current Sr 103 SNR in Peak mode as functi
202. ogenous process quality the Process Qualification Centres PQCs in Flo rence Strasbourg and Vienna perform several measurements on the standard test structures delivered with each sensor Figure 5 3 on the next page shows this standard test structure Starting from the right hand side two metal oxide semiconductor MOS structures are implemented using the same insulator as the sensors These metal insulated semiconductors MISs allow extraction of the flat band voltage by measuring the capacitance versus the gate voltage The next component is a photodiode with the same area as the two MISs structures 24 mm It is used to check the silicon resistivity and the wafer thickness via capacitance versus reverse bias voltage relationship The next device Cap TS DC dedicated to the study of the inter strip resistance is made of a series of nine strips without poly silicon resistors coupling The resistance between the central strip and its two first neighbours is measured while a reverse bias voltage of 20V is applied The largest substructure in the middle is a mini sensor It is a small equivalent in design of the large sensor The measurement of leakage current versus reverse bias voltages of up to 700 V allows to check the behaviour of this curve and also to find the breakdown voltage Again a test structure Cap TS AC build out of a series of nine strips is implemented to study the inter strip capacitance Cint The strips arrangement is the
203. ol system uses a PCI 6035E MIO card from National Instruments as de scribed in Sec 6 2 11 on page 73 to control the slow control multiplexer card and measure the connected temperature and humidity sensors Furthermore it controls the interlock system To keep cooling cycles sufficiently short several steps are taken first the FTSs active volume is reduced as much as possible which also reduces the possibilities of additional manipulations like placing probes on the module or usage of sources for SNR measurements 82 6 4 Fast Test Station OutsideAir Sa 34 07 32 0 45min en 30min i BoxCopperframe 37 7 j Su f A BoxTransportframe 37 a f on Foot2Peltier Down 57 nr A j i Foot1 Peltier Up gt 22 0 j Foot2Watercooling 5 20 0 Foot PeltierDown 7 18 0 i Foot Peltier Down 16 0 er ae 14 0 iu Foot Peltier Down 12 0 SPRAIN 10 0 j Foot Feltier Down 5 8 0 Foi Patron 3 eo l an j Foot PeltierDown 2 0 Fatt Eater Down E oo 2 0 a 6 0 8 0 1003 12 0 14 0 16 0 BI eee eet Cen eee eine et ee E nt ee ey OE et et fat ne eet et a atte et en et TA 46 55 36 47 15 00 47 30 00 47 45 48 00 00 15 00 48 30 00 48 56 46 Figure 6 17 Slow control front end showing two cooling cycles The red curve gives the transport frame temperature showing that a complete cooling cycle can be performed within 90 min The green curve gives the temperature of the co
204. ol the environment were first realised in separated applications connected by shared memory segments and synchronised with semaphores now merged into the multi threaded program The user interaction is realised by a graphical user interface which is connected to the different threads by TCP IP sockets kernels support a modular driver scheme with loading of additional drivers during run time resulting in a much more flexible operation There are mainly three different ways a Unix systems can address devices These are char modules block modules and network interface modules The char or character devices can be accessed like a file and the driver is in charge of implementing its behaviour Block devices usually handle larger blocks of data and access can only be done in multiples of a block Nevertheless this two types are very similar from the application point of view although their internal structure may differ very much The third kind of network interface drivers have to deal with sending and receiving data packages without even knowing the individual transmission specific structure of the underlying protocol Within the Karlsruhe Readout is now eAsy KaRinA software only character devices are implemented although the FED can also be implemented as a block device Main functions of a character device drivers interface are Install Uninstall routines which are needed to register or unregister the device driver and its capabilities to t
205. oling sink and the 4 C difference indicates that the thermal contact between the transport frame and the cooper structure of the cooling system still can be optimised As a second step the cooling volume is strongly insulated by layers of Polyurethane with a width of up to 10 cm which is an excellent thermal insulator The goal of the insulation design is to reduce the heat flow through the insulation down to the level of the heat dissipated from the FEH which has a heating power of approximately 2W The cooling system itself consists of four cooling towers with two staggered layers of peltier elements each The inner layer cools the active volume and the DUT while the second reduces the temperature step on the peltiers elements increasing the efficiency of the peltier cooling significantly The towers are connected to the in house cooling system keeping their feet at a constant temperature of approximately 12 C To keep the temperature gradient small each cooling tower consists of copper and the towers have a square size of 65 x 65 mm On the other side the copper material limits the cooling speed due to its heat capacity Nevertheless copper is chosen because it has a very good ratio of heat conductance to heat capacity only gold performs better The time needed for a cooling cycle is dominated by the heat capacity of the cooling systems copper skeleton This again has its pro and con While the con is mainly the increased time needed for a
206. on a contract between CERN on behalf of CERN users and International Business Machine IBM to which there are strict conditions attached which all users of components are required to observe The contract defines acceptance criteria and recourse in case of delivery of non compliant wafers Briefly IBM undertakes the commitment not to deliver wafers with components outside the electrical specifications of the process which is verified by IBM in testing their own test structures All IBM provides CERN with is a foundry service from which wafers are delivered untested which means that circuits on the wafer are the responsibility of the customer both for design and validation There is no formal radiation qualification or guarantee and therefore this must be based like most other components used at LHC on measurements made by the CMS community 46 5 1 Silicon Strip Tracker Production Scheme Frames Hybrid hybrid Brussels Strasbourg CF carrier A EN sensor s QTC module A Ld assembly N wien 4 yyy bonding amp testing a Y y ZE pao off ROD INTEGRATION TIB TID INTEGRATION PETALS INTEGRATION modules into mechanics sub detector TOB ass embly a TIB ID assembly TEC assembly EC asse Tracker assembly Figure 5 1 Tracker logistics from industrial producers to tracker assembly based mainly on the silicon sensor flow The individual steps of the production are stated on the left si
207. on detector is geometrically made of a cylindrical barrel closed at both ends by two end caps It is integrated in the return yoke of the superconducting magnet which has a magnetic field of 1 8 T This allow an independent measurement of the muons impulse which will be used for fast trigger decision The CMS muon detection system consists of three different detector technologies drift tube DT in the barrel region 0 0 lt n lt 1 3 cathode strip chambers CSCs in the end cap region 0 9 lt n lt 2 4 and resistive plate chambers RPCs in both barrel and end cap regions 0 0 lt n lt 2 1 Just for its sheer size about 2000m for the barrel detector and 1500 m for the end cap detector this muon detection system will be one of the largest and most complex wire chamber systems ever built The position resolution ranges between 50 and 200 um across the covered pseudo rapidity range and ensures a standalone transverse momentum p resolution of about Ap p 10 The muon detector will provide excellent muon detection with at least 16 radiation lengths of material in the pseudo rapidity range 0 lt n lt 2 4 The muon detector is made up of four stations in all its geometrical acceptance and provides at least 3 muon track segments along every muon track The muon trigger identifies muon track candidates with a transverse momentum threshold that can be adjusted to keep the trigger rate under control At the same time the trigger
208. on mode 5 Quality Control at CMS Tracker Modules 45 5 Quality Control at CMS Tracker Modules We have to distinguish two basic but entirely different types of quality control aims The first one checks for basic functionality and is typically applied after each production step The second type verifies long term reliability and is typically applied after major steps and at the end of a production chain Frequently reliability tests are also called burn in test although burn in is also an idiom for thermal stress tests Reliability tests are usually performed as long term operation test which will determine infant mortality Furthermore within CMS we specify a third kind of test for performance evaluation which is called qualification test This test is mainly needed due to the fact that we use an analogue readout which makes performance measurements much more complicated compared to digital readout systems Despite the large number of laboratories and institutes involved in the production a fast feedback and traceability of potential sources of faults is needed in order to guarantee a high efficiency in the construction of highest quality components This can be achieved with functionality tests after each production step Furthermore these tests are mandatory due to the fact that repairs done late in a production chain are known to be very expensive On the other hand it is obvious that this implies a lot of redundancy in the testing proce
209. on of leakage current Sr 2 2 222 220 104 Module noise as function of artifical leakage current 105 LIST OF FIGURES 143 7 30 7 31 7 32 7 33 7 34 7 35 7 36 7 37 7 38 7 39 7 40 7 41 7 42 7 43 7 44 7 45 7 46 7 47 7 48 7 49 7 50 7 51 7 52 7 53 F 1 F 2 G 1 Noise profile as function of artifical leakage current 105 Signal of a single infrared LED 2 2 on Emm 106 Envelope of LED signals 2 2 CC Enno 106 Envelope of normalised LED signals 2 2 LLC m rn nn 106 Scratch induced short of four strips o o nn 107 Shorted strips signature in calibration amplitude 108 Shorted strips signature in calibration signals 02 108 Shorted strips signature in noise 1 e 109 Picture of a missing bond 2 CE nn nn 110 Missing bond signature in noise and calibration 111 Missing bond signature in calibration pulse shape 111 Missing bond signature in noise at Vyas OV 2 ee 111 Missing bond flags arising from LED signals 112 Pinhole effect on calibration signals of the APV25 113 Pinhole search with artifical leakage current and calibration signals 114 Pinhole search in the noise with artifical leakage current 114 Forward biased module noise 2 a 115 APV25 edge noise in Peak and Deconvolution mode 115 Pedestals
210. ot be used in the barrel as they can not be operated in transverse magnetic fields 20 3 4 Calorimeter A pre shower detector of two layers of lead absorber intercepted with silicon sensors will be installed in front of the ECAL end caps to reduce the background for photons coming from 7 decays The absorber has a total thickness of 2 8X 9 interleaved with two orthogonal planes of silicon strip detectors The excellent spatial resolution of silicon detectors improves discrimination between single photons and closely spaced pairs of photons from the 7 decays The energy resolution of the ECAL in the energy range of 150 GeV can be parametrised 0 a Y n E VE p E ce where 0 is a contribution from electronics noise a is a stochastic term arising from photo electron statistics and the amplification noise in the VPTs E in GeV The constant term arises mainly from non uniformity in the light collection along the crystal and inert material in the structure At higher energies the constant term becomes relatively more important and eventually dominates the resolution Measured average values for a and c are 4 10 1 and 0 29 0 06 Apollonio et al 2002 Thus typical energy resolutions of the ECAL detector are 1 05 at 50 GeV and 0 47 at 180 GeV 3 4 2 Hadronic Calorimeter The coil radius is large enough to install essentially all the hadronic calorimeter inside and hence a coil calorimeter interference can be avoided The hadro
211. otential matching the vanish for a small leakage current interval 25 45 yA in the noise Taking the result form the calibration amplitude and the noise it is obvious that the APV25 gain is reduced by the current drawn into the APV25 Taking a look on the APV25 front end circuit Fig 7 43 on page 113 and recalling the arguments from Sec 5 5 3 5 on page 58 the effect of several pinholes gets clear Each individual pinhole will drive its preamplifier into saturation which turns on the corresponding inverter transistor The currents drawn from the inverter transistors cause a voltage drop at the external resistor Riny which is common to all channels of the APV25 Therefore the supply voltage of the inverter stages of all channels decreases causing a reduced gain of the complete chip Figure 7 52 on the facing page shows the dependence of the gain loss on the number of pinholes and on the module leakage current which must be taken as an upper limit due to the strip leakage currents contribution to the p implant potential compare Sec 7 7 4 on page 113 As long as the number of pinholes is not larger than three the APV25 suffers no gain loss but as soon as this threshold is reached a gain loss will be caused by leakage currents larger than 50 yA As a consequence of the DC current sensitivity of the APV25s pinhole detection is a major goal of the test strategy within SST Therefore all test stations will be equipped with an
212. ow control multiplexer is used to attach a large number of temperature sensors to the MIO card which serves only 12 analogue inputs Therefore a 32 1 multiplexer is build which also splits up the small computer system interface SCSI cable attached to the MIO card to easier configurable connectors for the individual sensors Furthermore it allows cross calibrations by the usage of a reference voltage 74 6 3 Software Layout 6 3 Software Layout The same idea of modularity that drives the hardware development is applied for the software design The choice to use an object oriented OO approach which forces a modular and encapsulated design is taken and C is chosen as programming language Nevertheless as a second programming language LabView is used mainly for it easy programmable visualisation capabilities and due to its good support of external devices like oscilloscopes pulsers source measurement units and so on As operating system Linux is chosen mainly due to personal experience with this operating system It also appears to be easier to write device drivers and to get direct hardware accesses under the Unix like environment The used compiler is the gec compiler in its actual version 3 2 and for debugging purposes the gdb debugger is taken Furthermore the concurrent versions system CVS is used during the software development to keep logs of who when and why changes occurred and for synchronising the developments done by d
213. oxide layer will be reduced In combination with an edge of the metal strip causing a localised increase of the electrical field this will result in an additional pinhole Depending on the individual configuration of the defect the break through voltage of the damaged oxide will only be reached of higher leakage currents when the p implant potential grow compare Sec 7 7 4 on page 113 about pinholes This defects can only be found with an increased leakage current like they are produced by the infrared LED systems developed in Karlsruhe Therefore these devices are mandatory for all test systems While scratches typically affect several strips the risk of producing a larger number of pinholes by a scratch is serious Figure 7 53 shows a sample of scratches that induced pinholes Therefore the estimations given in Tab 7 5 are realistic even with the improvements as discussed in the last section on the FEH in mind 8 Conclusion 121 8 Conclusion For the production phase of the CMS Silicon Strip Tracker SST the needed strategy and procedures for quality assurance of the new type of detector modules build from high voltage resistant silicon sensors are developed and tested During the phase of the first prototypes the behaviour of the SST modules is studied and the empirical basis for the final quality cut definitions is available now This will result in corresponding changes of the Procedures for Module Test Dirkes et al 2002
214. pact Muon Solenoid CMS detector design and its sub detectors Thereafter in Sec 4 the Silicon Strip Tracker SST is discussed in more detail A short description of the mechanics of SST modules and an introduction to the silicon sensors is followed by a discussion of the readout chain Hereby special emphasis is given to the front end hybrid FEH and its embedded chips especially to the analogue pipeline voltage chip APV25 The different readout modes and the signal processing as well as other specific characteristics like the common mode CM suppression are introduced The section is closed by a discussion of the module performance dealing with the signal creation as well as with the noise behaviour Section 5 shows the different steps of the SST module production and the corresponding quality control specifications The latter are strongly influenced by the test results of the first prototypes performed in Karlsruhe Those test results finally lead to the definition of the official test procedures given in Dirkes et al 2002 culminting in an unique pinhole tagging method which utilised artifical leakage currents induced by infrared LEDs Finally this section concludes with a detailed discussion of module faults starting from general chip failures and ending with an analysis of pinhole effects to the readout The following Sec 6 is devoted to the test stations built for the module testing in Karlsruhe and describes the hard and so
215. peaking of Deconvolution or Peak mode refers to inverter off modes On the other side the plural form like the Peak modes includes both Peak mode with and without w amp wo inverter 7 2 Noise Studies Beside the intrinsic noise of the module itself as discussed in Sec 4 3 1 on page 38 external noise sources have to be identified and reduced as much as possible Main observables con nected to external noise sources are the common mode CM distribution and the RMS of the noise Crucial for noise performance is the grounding scheme applied to the module For both test stations several iterations of grounding scheme optimisations lead to a significantly re duced system noise spectrum Key strategies hereby are the selection of a good earth ground point for the entire system which has to be close to the DUT and a ground loop free cabling scheme Spieler 2001 Attention has also to be payed on shielding and cabling of the power supply lines Another important point is the cable length between the FEH and FED The design of the APV25 does not foresee the driving of longer lines which asks for the usage of a repeater card But even with the use of a repeater card the length of the connecting line causes attenuation of the signals amplitude and an increase of both rise and fall time The latter disturbs the signal settling at the ADC Finally there will be a noise pickup on the cable 7 Test System Performance and Module Qualification Studies 87
216. per structure of the inverters impact on the pedestal is revealed by Fig 7 2 Taking the difference of the pedestals in Peak mode with and without inverter shows a typical structure which is more prominent for Peak than for Deconvolution mode Switching from inverter on to inverter off results in an decreased pedestal of the order of 3 5 ADC The shape is almost identical for all APV25s on the FEH and the differences which can be easily explained by statistical fluctuations stay below 0 2 ADC Figures 7 3 on the facing page shows the raw noise of the different readout modes for module 30200020000503 Based on the noise analysis done in Sec 4 3 3 on page 41 one expects an increased noise for the Deconvolution modes at a level of roughly 40 which is true for the inverter off modes while with the inverter stage on the increase is slightly smaller and reaches only a level of 20 The RMS of the raw noise distributions are typically of the order of 0 1 ADC Details of the noise dependencies on the readout mode are discussed in Sec 7 2 3 on page 91 7 Test System Performance and Module Qualification Studies 89 Sensor channel number Figure 7 3 Raw noise for the different readout modes shows significant differences As ex pected from the noise analysis the Deconvolution modes have an increased noise compared to the Peak modes Furthermore usage of the inverter stage de Raw noise ADC Peak wo Cal wo Inv Mode Peak
217. performance is studied with particular em phasis on fault behaviour especially pinholes A dedicated analysis method investigating the effect of artifical leakage current on the fault signals has been developed and is presented 7 1 Readout Modes In Sec 4 2 1 2 on page 29 the APV25 and its major readout modes the Peak and Deconvo lution modes were discussed Their different behaviour due to the signal processing by the APSP filter in Deconvolution mode has to be distinguished Both Peak and Deconvolution mode can be used either with the inverter stage on or off Thus the APSP filter and the inverter stage define the four readout modes For completeness the Multi mode has to be mentioned which is useful for debugging purposes but is of no importance for quality control aspects The four readout modes can further be operated together with the calibration unit There fore specifying a readout mode one has to indicate the kind of signal processing used Peak Deconvolution the status of the inverter stage on off and finally if the calibra tion circuit was used or not Thus typical readout mode specification looks like Dec wo Inv Mode wo Cal which reads as Deconvolution without inverter mode without calibration used All figures give this information in the lower right edge together with the ID of the module used Within the text some shortcuts will be used By default the calibration circuit is not used if not mentioned otherwise and s
218. pidities of 2 5 Following on the formation of a proto collaboration for CMS in May 1991 the Letter of Intent Lol based on a conceptual design of the complete detector was signed and submit ted to the LHC Committee LHCC in October 1992 With the Technical Proposal CMS Collaboration 1994 1994 the design phase was finished and detailed Research amp Develop ment R amp D studies during the following years led to the publication of the Technical Design Reports of the individual sub detectors CMS Collaboration 1997a b c d 1998 In 2000 there was a major design change for the Tracker Project The micro strip gas chambers MSGCs in the outer part of the tracker were replaced by an all silicon solution 3 1 Detector Design Goals The design baseline of CMS was driven by e a very good and redundant muon system superimposed on a solenoid e best possible ECAL consistent with the muon system e a high quality central tracking to achieve both points above e and of cause to be a financially affordable detector The large radius of 6 0m of the chosen solenoid allows a full calorimetry located inside the solenoid From the LEP experiments it is well known that very precise electromagnetic calorimetry is essential for separation of photons from decays for example from low mass Higgs bosons Such a precision calorimetry fits naturally in the CMS design Figure 3 1 on the following page shows an overview of CMS The individual sub detectors are discus
219. polarity can not be changed with the inverter this indicates a bad APV25 chip and the module can not be used This type of fault is not expected to be common due to the fact that the ASICs and the FEH are tested before but we are nevertheless ready to identify them if present Typically they will be induced by handling for instance through an electro static discharge 5 5 2 4 APVMUX and PLL Failures are also easy to identify The APVMUX will simply be fed with two APV25 in different operation modes one biased on the other off This has to be reflected by the APVMUX output For the PLL it is in principle not possible to check directly if the trigger line reconstruction is functional This can only be checked indirectly via the APV25s output For the PLL we check the clock and trigger delay functionality The clock delay can be checked by looking at the phase change relative to the FEDs sampling point For this the APV25s tick marks are analysed For the trigger delay check we can not use the APV25s calibration circuit because the request of a calibration pulse is transmitted over the same line Therefore we have to use 56 5 5 Module Error Type Detection external signals so that we can corrected changes in the latency by delaying the trigger with the PLL 5 5 2 5 Low Voltage Power Consumption distribution of the FEH has a very small RMS and reflects the correct functionality of the FEH The main power consumption is caused by the APV25s T
220. pt at 10 C This will drastically reduce the sensor dark current and consequently the risk of thermal runaway of detector modules Furthermore keeping the irradiated sensors at low temperature substantially reduces the reverse annealing effect Moll 1999 and helps keeping the depletion voltage below a reasonable level Crucial for track finding is mainly the number of detectors intersected on the one side and the overall material budget on the other side because all material inserted will cause multiple scattering which reduces the track quality see Figs 3 6 and 3 7 on the next page Critical regions are the gaps between barrel and end caps and the high pseudo rapidity regions However the average hit multiplicity within the Silicon Strip Tracker is at least eight for tracks of infinite momentum and n lt 2 4 where four of these come from double sided detectors refered as 3D in Fig 3 6 on the facing page 3 Compact Muon Solenoid 19 All Tracker N points Beam Pipe A Pixel B Silicon MD Outer Tracker E Common D Outside HARE HA Inner 3D n Figure 3 6 Average number of de Figure 3 7 Contributions of the CMS tectors intersected by infinite momen tracker subsystem to the interaction tum tracks as function of pseudo length The label outside refers to rapidity for the silicon strip tracker all material installed beyond to active Caner 2001 volume of the tracker Caner 2001 3 4 Calorimeter The
221. quire only a passive cooling cycle which means that no readout has to be done during the cycle while we decided to go for an active cycle with full readout capability 6 Karlsruhe Test Stations 81 Figure 6 16 View on the Karlsruhe Fast Test Station FTS The cover can be lifted on rails The counterweights for the cover run inside the pipes which can be seen in the pictures background Fur thermore the onion like struc ture of the thermal insulation layer surrounding the small inner test volume is visible On top partly visible and below the test station scin tillators with photomultipli ers are assembled which can be used for system calibra tion and SNR measurements with cosmic rays The elec tronics motherboard and its hosted cards are place behind the stations body and below the station PC power sup plies power pack and NIM log ics For ESD protection sev eral GND connection points are available for the operator e g cable visible at the lower front corner The readout system embedded in the FTS uses the components described in the previous chapter 6 1 on page 61 Figure 6 15 on the facing page shows the schematic setup and Fig 6 16 a photography of the FTS Besides the readout components collected on the moth erboard additional parts are included like a complete slow control system or scintillators with photomultipliers and NIM logic for cosmic ray detection Waldschmitt 2002 The slow contr
222. r 300 200 iy Figure 7 33 Envelope of nor malised LED signals like they are needed of comparison of the data of different LED ar rays For a missing bond tag this envelope of normalised sig nal has to be compared with an envelope of fitted LEDs cones which removes false tags aris ing from limited overlap of ad jacent LEDs like channel 128 0 10 200 300 400 cr in this plot DB channel number e a Normalised IR LED signal amplitude o o gt gt N Module 3020002000063830200020000638a Peak wo Cal wo Inv Mode 7 Test System Performance and Module Qualification Studies 107 Figure 7 34 Scratch induced short of strips and in this case only the four central strips are shorted first ring 7 prototype module 30200020000632 mu e 7 7 Module Fault Detection Studies Nearly all types of defects can be found by their corresponding noise behaviour but the distinction of different defects types is complicated Beside the noise analysis inherent to a all readout systems additional tools are available such as the APV25s calibration circuit or the infrared LED system Together with external parameters like the applied bias voltage ambient temperature or humidity they provide a vast set of observables from which clear signatures have to be extracted In close relation to Sec 5 5 3 on page 56 dealing this the expected fault behaviour the different defect types are dis
223. r Experimentelle Kernphysik 2002 T Weiler Infrared LED Array for Silicon Strip Detector Tests datasheet Institut f r Ex perimentelle Kernphysik Universit t Karlsruhe TH 2002a T Weiler Multiplexer datasheet Institut f r Experimentelle Kernphysik Universit t Karls ruhe TH 2002b T Weiler 2003 private communication G Wilkinson CP violation in the B sector at the LHC Nucl Phys Proc Suppl 86 491 498 2000 L Wolfenstein Parametrization of the Kobayashi Maskawa Matrix Phys Rev Lett 51 1945 1983 J Womersley The LHC physics program FERMILAB Conf 350 1997 To be published in the proceedings of 20th International Workshop on Fundamental Problems of High Energy Physics and Field Theory Protvino Russia 24 26 Jun 1997 G Wrochna Heavy ion physics with CMS detector Eur Phys J direct C451 23 2002 Acknowledgements The work in modern experimental particle physics as it is faced in a collaborations of the size like CMS require beside excellent physical understanding above all the virtues of communi cation and team work Within the Karlsruhe CMS group I faced a cooperative and collegial atmosphere within a group of different characters with different particular interests and experience Nevertheless or better even because of this constellation we all and I in personal gained very much from this group I have to thank Prof Dr Thomas M ller for giving me the opportunity to do thi
224. r at half system speed 20 04 MHz The latter option enables usage of an external 2 1 multiplexer stage constructing a 40 04 MHz output stream which is done within CMS Thus within CMS the readout of all 128 channels plus digital header information takes 280 clock cycles data frame length 7 us If no output is pending a 1 bit tick mark is send every 70 system clock cycles in order to keep the synchronisation with the DAQ CM Suppression An important feature of the APV25 is its internal common mode CM suppression Figure 4 9 on the facing page shows the working principle The CM signal appears on the external resistor supplying the preamplifiers output inverter stage which results in an effective subtraction of the common mode CM from the signal Hereby the external 100 Q resistor has been introduced for stability reasons after the first FEH prototypes 4 Silicon Strip Tracker and its End cap Modules 31 external resistor on hybrid APV V250 wre 1 per APV chip Rinv 1000 v125 a Yom this point R ee pi common for i 4 IN CM all 128 chans Pa re Vout Vi 75V 8 8 base line restorer Figure 4 9 Working principle of the APV25 CM suppression Raymond 2001b external resistor on hybrid V250 Pa 1 per APV chip one channel Rinv 1009 127 channels with signal with common and common mod
225. r of a single pinhole in the calibration amplitude where no effect on the other channels is visible for an isolated pinhole This picture changes dramatically when the number of pinholes connected to an APV25 increases 118 7 8 High Leakage Current Behaviour of Pinholes Figure 7 51 The effect of a dozen pinholes on the noise shows a similar behaviour as for the calibration amplitudes While the pinholes are noisy at higher leakage currents a decrease of the general chan Noise ADC o nel noise below 50 uA similar jz to the loss in calibration am A plitude shown in Fig 7 50 is C 33 visible This indicates that the E E E APV25 derivates in gain the C Ze two channels with increased jf e A a E noise are APV25 border chan 6p GSS Sa a F nels Leakage current 1A Figure 7 50 on the preceding page presents the effect of a dozen pinholes for the calibration amplitude which shows a deterioration of 30 as soon as the potential of the p implant rises to the APV25 virtual ground level compare discussion in Sec 7 7 4 on page 113 But not only the calibration amplitudes show a deterioration For the noise a similar effect is measured which is shown in Fig 7 51 Here the effect looks a little more complicated While the pinholes are noisy at higher leakage currents a deterioration of the general channel noise is visible In the low current region the pinholes have a reduced noise and in the region of the p
226. r with the shaper is about 110mV MIP with a total power consumption of 1 15mW channel The output which is sampled at the 40 08 MHz LHC bunch crossing frequency is stored into a 192 elements deep analogue pipeline memory compare schematic Fig 4 7 The pipeline enables storage of tracker data for upto 4 8 us trigger latency while decision is taken whether an interesting physics event has been taken or not APSP The pipeline is DC coupled to the analogue pulse shape processor APSP which works in either Peak Deconvolution or Multi mode The APSP is designed to discriminate between signals from consecutive bunch crossings This need arises from the fact that the amplifier and shaper stage are optimised for noise performance which calls for long shap 30 4 2 Readout and Control Chain HO MUX gain E lit 128 1 n f Differential OS inverter n MUX current a 2 pipeline gt output amp J gt Hb Li gt All SR APSP 8 Eu preamp rn shaper a Ft S H Figure 4 8 Schematic circuit diagram of the APV25 chip The schematics to the left of the 128 1 multiplexer MUX is implemented individually for each of the 128 channels Turchetta et al 2001 ing times and thus long output pulses The chosen preamplifier and shaper configuration produces pulses
227. raggling functions i pi EN Mean energy 7 in silicon for 500 MeV pions nor L i x AN loss rate malised to unity at the most prob oa i RY L Ss able value Amp x The width w ES pP E aa do a aaa aa is the full width at half maximum 100 200 300 400 500 600 Hagiwara et al 2002 A x eV um detector from the mean energy deposit as given by the Bethe Block formula at a level of about 70 Hagiwara et al 2002 Since both electronic and lattice excitations are involved the variance in the number of charge carriers N E I produced by an absorbed energy E is reduced by the Fano factor F about 0 1 in Si leading to y VFN and the energy resolution og E y FIJE However the measured signal fluctuations are usually dominated by electronic noise or energy loss fluctuations in the detector Charge collection time scales with the charge carrier velocities which can be increased by operating the detector at higher voltages This leads to an over bias i e a bias voltage exceeding the value required to fully deplete the device Nevertheless the collection time is limited by velocity saturation at high fields approaching 10 cm s at E gt 104 V cm at an average field of 10 V cm the drift velocities are about 15 ps um for electrons and 30 ps um for holes In typical fully depleted detectors 300 um thick electrons are collected within about 10 ns and holes within about 25 ns 4 3 2 Radiation Effects
228. rature of 10 C 4 The width of the distribu tion is dominated by electron ics noise and by the intrinsic o detectors resolution 6 3 20 Peak wo Inv Mode wo Cal Module 30216630300027 n 60 80 100 120 140 SNR ob N o gt a Entries 641 Mean 21 22 x Indf 59 64 72 Width 1 752 0 3362 MP 19 0 4266 Area 430 3 18 57 GSigma 5 407 0 5364 Entries 8 N o oHTTI EENS LEE TITT TITT TTTT TTT Figure 7 20 SNR in Decon 1 volution mode at a bias voltage of 300 V and a temperature of 10 C The width of the dis tribution is dominated by elec tronics noise and detectors in o trinsic resolution a a Dec wo Inv Mode wo Cal Module 30216630300027 increase of the peak position most probable value Amp compared to a pure Landau Fur thermore the fit was done automatically with fit ranges set always to 30 of the histograms mean value and to five times the mean value Finally not the energy loss itself is fitted but the cluster SNR distribution which is expected to follow the same behaviour As already described the cluster SNR is defined as SNR cluster signal Ber 7 2 lust i Cluster noise 1 NN Since the noise is very homogeneous the main difference between the energy loss and cluster SNR distribution will be a scale factor namely the noise Figures 7 19 and 7 20 show fitted SNR distributions tak
229. ren Messungen Die f r die Produktion notwendigen Testsysteme und Prozeduren konnten anhand der Prototypserien erprobt werden und erwiesen sich als au erordentlich erfolgreich 1 Introduction 1 1 Introduction To a great degree the progress of particle physics has followed from progress in accelerator science and instrumentation There is no substitute for experiment and experiment requires both inventions in hardware and software as well as continuous innovation in analysis tech niques The slogan Yesterday s sensation is today s calibration and tomorrow s background V L Telegdi embodies both the challenge and the opportunity of advances in experimental technique In the middle of the revolution we are experiencing indeed making in our conception of Nature when we deal with fundamental questions such as e What are the symmetries of Nature and how are they hidden from us e Are the quarks and leptons composite e Are there new forms of matter like the superpartners suggested by supersymmetry e Are there more fundamental forces e What makes an electron an electron a neutrino a neutrino and a top quark a top quark e What is the dimensionality of spacetime we cannot advance without new instruments that extend our senses and allow us to create and understand new experience far beyond the realm of everyday human knowledge The Large Hadron Collider LHC program addresses these questions and one of i
230. rimeter aia Han ke ln et Bee Be 3 4 1 Electromagnetic Calorimeter 0 0 0000 eee eee 3 4 2 Hadronic Calorimeter 2 2 ee 3 4 3 Hadronic Forward Calorimeter 000008 eae 3 07 Muon System pw ado Ake ee ee eae Se ee ee ee 4 Silicon Strip Tracker and its End cap Modules 4 1 Silicon Strip Tracker Modules 2 22 Coon nn Al Mech nies A a an ea i 4 1 2 Silicon Micro Strip Sensors nn 4 2 Readout and Control Chain 0 0 0 0 000002 n 4 2 1 Front End Hybrid 22 2 Corn nn dal Pi A A AAA AAA E E AZ APV25 a a DO EAE e aoe kG AP SPs go nt AA la ho SEE othe CM Suppression 2 i aeiia ne Bee eka Inverter stage so i o 002 Se RA ee A Se ee Calibration unit 2 22 a Interfaces u 28 a fideo a ada rt A213 SARPVIMUX 22 a te pastas hy bee EE Beer gh Ree AQ MA DEU as Bee ete oa he a a hs BO we Se OS 15 15 15 16 16 17 19 19 20 20 21 il CONTENTS 4 2 2 Optical Links aan aaa na pee ne eee nes 34 42 21 ALA it 31 6 gia ol a ek oh Piao hh un na en 34 4 2 2 2 Analogue Opto hybrid 2 2 2 2 on nenn 35 4 2 3 Front End Driver e 35 4 2 4 Control and Monitor Path 2 2 2 CC mn nn 35 AD ALL A AA ea gett Bete bo ne eS ee BB Sh BAA a 35 42427 HOCWUIME N a he ee yh a A a Be GE ER Ze 37 4 2 4 3 Digital Opto hybrid 222 2 2 non nn 37 A2AA MING A ar ee ann rato Beh a rd 37 4 3 Expected Module Performance 22 2 22mm nme 37 4 3 1 Signal Creation
231. rip Tracker layout 2 2 2 222m ee Average number of detectors intersected by infinite momentum tracks as func tion of pseudo rapidity for the silicon strip tracker Contributions of the CMS tracker subsystem to the interaction length A ring 5 module and its components 0 200002 eee Schematic structure of CMS silicon strip sensors 2 2 2 22m nn Close up view of a tracker end cap sensor 2 02000004 Schematic of a silicon micro strip sensor 2 2 2 Cm mn CMS Silicon Strip Tracker readout and control system Combined coding for the LHC clock and first level trigger signals Block diagram of the APV25 2 2 on E nn Schematic circuit diagram of the APV25 chip Working principle of the APV25 CM suppression 2 222 2 222m Small signal model of the APV25 CM suppression Switching between two APV25s by the APVMUX Connections of the DCU on CMS end cap modules The FEC and CCU token ring architecture En The ECU block diasram 2 3 a era ran a re Transverse momentum resolution as a function of pseudo rapidity Zimp resolution as a function of pseudo rapidity 2 2 222 Straggling functions in silicon for 500 MeV pions 2 2 2 2 2222 Schematic silicon detector front end circuit 2 22 o nn Equivalent circuit for noise analysis 2 on on nn Tracker logistics from industrial p
232. rlskog invariant J C12C23C13812523813sind An 2 11 where c and s are shorthand for the cosine and sine of the angle 6 see e g Hagiwara et al 2002 for the corresponding standard parametrisation and A A and n are the parameters of the Wolfenstein parameterisation The maximum value of J in any unitary 3 x 3 matrix is 6 3 0 1 the value in the CKM matrix is 4 x 107 which underlies the statement that CP violation in the SM is small 2 5 1 The Mass Difference Am and B _B Mixing The eigenstates of flavour B bq and B bq q d s degenerate in pure QCD mix on account of weak interactions box diagrams of Fig 2 4 The quantum mechanics of the two state system with basis 1 2 B IB is described by a complex 2 x 2 Hamiltonian matrix i M Mi i r Ta H M T 2 12 2 A de E 3 ER with Hermitian matrices M and T and for the case that CPT theorem holds The off diagonal elements in Eq 2 12 arise from AB 2 flavour changing transitions with virtual M12 or real intermediate states T12 in the latter case corresponding to decay channels common to B and B Diagonalising Eq 2 12 the physical eigenstates B heavy B less heavy and the corresponding eigenvalues Mn Tr are obtained The mass and width difference read 2Re M 1 12 Amp Mr M 2 Mi2 AT T T Ma2 2 13 The mass difference Amp is a measure of the frequency of change from B
233. rocessed twice because channels with signals are discarded from the calculation and therefore one needs at least an estimate for the pedestal and noise Thus one uses the first Thus to reduce overall noise concentrate on large noise signals 88 7 2 Noise Studies Sensor channel number Pedestal ADC 4 4 cal o dl io 1 00 Figure 7 2 Pedestal oe wu me zo dependency on inverter stage for Peak wo Cal mode data The upper plot has been taken without the inverter while the plot in the middle shows Peak wo Cal w Inv mode data The difference as shown in the lower plot reveals an obvious structure SoS Te si ar a u a which is the same for all pe channel number APV25 chips Pedestal ADC wo Cal Pedestal diff ADC Module 30200020000503 31 Peak w amp wo Inv Mode 30200020000503 100 events to calculate a pedestal and noise estimate which are then used as starting val ues for tagging signals To calculate the RMS on the fly the well known RMS relation Oy V lt x gt lt qx gt is used The RMS calculation gets a little bit more complicated if CM correction is applied Figures 7 1 on the page before shows the pedestal measured for module 30200020000503 For the modules tested so far the Peak modes have a higher pedestal than the Deconvolution modes and inverter off modes tend generally to smaller pedestals than inverter on modes Beside the slight different offsets a dee
234. roducers to tracker assembly Sensor logistics for the main production 2 2 2 En Picture of a standard test structure 2 2 2 2 Eon nn Proton irradiation beam line with thermal insulation box and plot of the ac tivity of a scanned Ni foil by autoradiography 2 2 2 222 nn Stacker for irradiation s sc cea cese za ve eR a m Bea ran Cut view of a TEC module at the FEH FHIT as used by the FEH producing companies Module logistics for tracker end caps 2 2 2 Cm m nn Automatic assembly robot as used in the CMS gantry centre in Bari Italy Karlsruhe automatic bonder from Hesse amp Knips Schematic of scratch fault in silicon strip detectors Schematic of short fault in silicon strip detectors Schematic of break fault in silicon strip detectors 2 2 22 2 2 Schematic of pinhole fault in silicon strip detectors 2222222 Pinholes and the APV25 circuit 2 2 2 CE En nn 141 142 6 1 6 2 6 3 6 4 6 5 6 6 6 7 6 8 6 9 6 10 6 11 6 12 6 13 6 14 6 15 6 16 6 17 6 18 6 19 6 20 7 1 7 2 7 3 7 4 7 8 7 6 7 7 7 8 7 9 7 10 7 11 7 12 7 13 7 14 7 15 7 16 7 17 7 18 7 19 7 20 7 21 7 22 7 23 7 24 7 25 7 26 7 27 7 28 7 29 LIST OF FIGURES Hardware layout of the Karlsruhe readout system o o 62 Karlsruhe motherboard with all cards mounted 6
235. rrection applied The shape of the CM corrected noise middle results mainly from the different strip length on the PA 30200020000638 on the DTS Sensor channel number Raw noise ADC Noise ADC A O try Noise diff ADC 400 500 DB channel number Peak wo Cal wo Inv Mode Module 30200020000503 o 3020002000503 The observed CM correction gives an explanation of the raw noise correction shape visible in Fig 7 5 Especially for the inner APV25s the CM correction shows an increase towards the edges lower plot of Fig 7 5 This results from the quadratic nature of the of RMS calculation and the fact that one uses a linear function for the correction Although the mean correction is the same for all channels of the APV25 its fluctuation increases for the mid towards the edge and this enters quadratically into the RMS calculation resulting in a larger correction towards the edges Taking the corresponding values of the CM correction and its slope the expected edge increase is of the order of 0 15 ADC This is superimposed by other edge effects 7 Test System Performance and Module Qualification Studies 91 Entries 128 Mean 2 120 RMS 0 086 Entries 128 Mean 2 047 RMS _0 063 Entries Entries 16 1 8 2 22 2 4 26 28 3 32 34 0 Te 18 2 22 24 26 28 3 32 34 Channel noise ADC Channel noise ADC APV 3 Entries 12
236. rred The other class embedded in the APV library is the ApvCluster class used to manage several APV25 through a common interface This class serves the same methods as the Apv class but with an additional argument specifying which APV25 should be written Furthermore the ApvCluster class supports the global APV25 address accessing all APV25s connected to the I C bus simultaneously PLL library contains the Pll class whose methods give access to the PLL functionalities compare Sec 4 2 1 1 on page 28 132 APPENDIX C KARLSRUHE READOUT LIBRARIES void Pll ushort amp addr 12c bus constructor call takes a reference to a variable holding the I C address of the chip and an pointer to the I2c class needed to get access to the 12C bus as arguments needed to access the PLL int setClockPhase ushort phase and int getClockPhase allows control of the PLLs clock phase adjustment ability int setL1TriggerDelay int delay and int getL1TriggerDelay give access to the delay line for the trigger signals int getStatus is implemented to check the status of the built in auto calibration circuit and of the SEU detection which is implemented as triple voting logic int restart method will restart the auto calibration circuit The return values follow the general scheme of being 0 or a positive defined result on success and being negative in case of a fault Furthermore auxiliary methods are implemented to check the internal registers of the auto
237. rs bei Pinholedefekten Hier regeniert sich das Kalibrationssignal bei kleinen Leckstr nem bevor es bei Leckstr men von mehr als 30 yA wieder degeneriert prinzipiell einen ganzen Auslesechip beeintr chtigen Durch den Ohmschen Kurzschlu der AC gekoppelten Auslesestreifen siehe Abb 8 liegt am dazugeh rigen APV25 Ein gangskanal das Potential Vimp an welches den Vorversp rker in S ttigung treibt Solange das Eingangspotential kleiner als der virtuelle Ground von 0 75V des Vorverst rkers bliebt bleibt sein Ausgang V positive und der Invertertransistor schaltet nicht durch Steigt das Eingangspotential allerdings ber die 0 75 V so schalted der Invertertransis tor durch womit dieser permanent Strom zieht Das Potential des Implantatstreifens Vimp ist abh ngig von dem Leckstrom welcher durch den Vorspannungswiderstand Rpoly und einen Me widerstand auf der R ckf hrleitung Rret auf dem Auslesehybriden fliesst Abbildung 9 zeigt das Verhalten der Kalibrationsampliude bei eines APV25 mit einem k nstlich erzeugten Pinhole Bei einem k nstlichen Leckstrom von ca 30 uA wenn das Eingangspotential Vimp mit dem virtuellen Ground des APV25 bereinstimmt wird das Pinhole im Signal der Kalibrationseinheit ununterscheidbar von normalen Kan len W hrend ein einzelnes Pinhole die anderen Kan le des APV25 nicht beeinflu t ndert sich dieses Bild drastisch sobald mehr als drei Pinholes an einem APV25 angeschlossen sind Abbildung 10 zeigt da
238. rt the deposited energy into charge the system s noise level is conveniently expressed as an equivalent noise current ENC equal to a detec tor signal yielding a signal to noise ratio of one The equivalent noise charge is commonly expressed in the corresponding number of electrons Coulombs or the equivalent deposited energy eV For a capacitive sensor the equivalent noise current ENC Qn is 2 C Qn in Ts en Foz Fup AsCio 4 11 where in is the sum of all current noise generators e the sum of all voltage noise sources Crot is the sum of all capacitances shunting the input coupling and inter strip Lutz 1999 F Fy and F r depend on the shape of the pulse determined by the shaper and T is a characteristic time for example the peaking time of a semi Gaussian pulse The form factors F Fy are easily calculated BE 5 OT awa pe zr wo dt m f D at 4 12 where for time invariant pulse shaping W t is the system s impulse response the output signal seen on an oscilloscope with the peak output signal normalised to unity The total capacitance Crot can be calculated for CMS sensors using the values measured for the coupling capacitance Ce inter strip capacitance Cint and a calculated one for the strip to backplane capacitance Cy Assuming that only inter strip capacities from first neighbouring ones contribute significantly to capacitance C strip seen from the strip into the detector we get Cstrip
239. ry eprint hep ph 0012288 2000 M Kobayashi and T Maskawa CP violation in the renormalizable theory of weak interaction Prog Theor Phys 49 652 657 1973 D Kotlinski The design of the CMS pixel detector system Nucl Instrum Meth A477 446 450 2002 R V Kowalewski B physics and CP violation eprint hep ex 0305024 2003 M Krammer 2003 private communication Z Kunszt The physics potential of the LHC eprint hep ph 9710210 1997 P Lecomte Large lead tungstate crystals for calorimetry IEEE Trans Nucl Sci 49 583 587 2002 W R Leo Techniques for Nuclear and Particle Physics Experiments Springer Verlag Berlin Heidelberg 1994 G Lutz Semiconductor Radiation Detectors Springer 1999 G Magazzu A Marchioro and P Moreira DCU2 User Guide CERN EP MIC Geneva Switzerland INFN Sezione di Pisa Pisa Italy 2001 Version 2 16 M Manneli 2003 private communication A Marchioro C Ljuslin and C Paillard CCU25 Communication and Control ASIC for embedded slow control 2002 148 REFERENCES E Migliore Measurements on PA in Torino CMS Tracker Electronics Meeting January 2002 http cmsdoc cern ch ghall TKEL_0102 Migliore_0102 pdf M Moll Radiation damage in silicon particle detectors Microscopic defects and macroscopic properties PhD thesis DESY 1999 DESY THESIS 1999 040 E Murer PMC to I2C Bus Controller User s Guide CERN EP Division MIC DG Section March 2000
240. s load the circuit with all components and connector s do the wire bonding of the APV25s power lines and finally test the FEH electronically Therefore the companies are equipped with specialised test system called FEH Industrial Tester FHIT see Fig 5 6 on the next page This test system has been developed from the CMS groups at the RWTH 2000 5 Quality Control at CMS Tracker Modules 51 ceramic et au SUITES a F spacer RACI N f y G N graphite cros 1000 SIE um LSM J Y LI Y es ER legs My hard glue araldite 0 ELL soft glue DC3140 Figure 5 7 Cut view of a TEC module at the FEH Due to the different frame design and sensor thickness of rings 1 4 and ring 5 7 two different module build ups exist Honma 2003 Aachen and the Res in Strasbourg and allows simplified functionality tests during this first step of the production Axer et al 2001 First the FEH Industrial Tester FHIT sys tem checks the connectivity to reveal shorted or open circuits This is done completely passive without powering of the DUT The second test part consists of an electrical test It includes tests of power supply currents and voltages test of the IC communication with the DCU PLL APVMUX and APV25s ASICs Both connec tivity and electrical tests reveal most of the er rors on the FEHs Finally a functionality test Figure 5 6 FHIT as used by the FEH pro is performed measuring pedestal noise and cal duci
241. s Verhalten eines APV25 mit unterschiedlich vielen Pinholes anhand der Kalibrationspulse bei zunehmendem k nstlichen Leckstrom Hierbei reduziert ein Leckstrom von mehr als 50 uA die Verst rkung des APV25 bei mehr als drei Pin holes um bis zu 30 Ursache hierf r ist dass alle Inverterstufen ber einen externen Widerstand mit Spannung versorgt werden Fliesst nun aber gr erer Strom ber diesen Widerstand so kommt es zu einem Spannungsabfall welcher auf alle Kan le des APV25 durchschl gt vgl auch Abb 8 IX Zusammenfassung foe a o N al N o o al a ei E o o Mean calibration amplitude ADC 0 2 E 4 50 100 150 200 u 10 8 2 anoles 250 300 350 400 12 ne AP Leakage Current uA Abbildung 10 Pinhole verursachte Abnahme des APV25 Verst rkungsfaktors Die Antwort auf Kalibrationspulse mit einer ein MIP quivalenten Amplitude kann als Me gr e f r den APV25 Verst rkungsfaktor verwendet werden Bei sehr niedrigen Leck str men arbeitet der APV25 noch normal welches von den geringen Potentialdifferenzen in diesem Leckstrombereich herr hrt Mit steigendem Leckstrom aber zeigt sich ein Verst rkungsverlust von bis zu 30 bei mehr als drei Pinholes Extrapoliert man die Zahl der Pinholedefekte wie sie in den Vorserien gefunden wurde auf die gesamte Produktion so erwartet man dass bei 3 5 der Module ein APV25 mit mehr als drei Pinholes verbunden ist H
242. s an extra term i e 1 p gt iTe WW zia a x Wy 2 4 This ensures gauge invariance of W ww and hence for the last term of the Lagrangian represent ing gauge kinematic energy Evidently the middle term of Eq 2 2 representing scalar mass and self interaction is invariant under gauge transformation on Finally the first term rep resenting scalar kinematic energy and gauge interaction can be shown to be invariant under the simultaneous gauge transformations Eq 2 4 However the addition of a mass term ae a M W W 2 5 would clearly break the gauge invariance of the Lagrangian In contrast the scalar mass term uo is clearly gauge invariant This phenomenon is exploited to give mass to the gauge bosons through back door without breaking the gauge invariance of the Lagrangian This is the Higgs mechanism of spontaneous symmetry breaking Roy 2003 Starting with a SU 2 doublet of complex scalar field amp with imaginary mass i e u lt 0 results in the minimum of the scalar potential 2419 A to move out from the origin to a finite value v y 2JA 2 6 i e the field develops a finite vacuum expectation value Since perturbative expansion in quantum field theory is stable only around a local minimum one has to translate the field by the constant quantity v H z 2 7 6 2 2 Standard Model Higgs g t t pH t g Figure 2 1 Higgs production in proton proton high energy collisions is do
243. s in a 9 ES slight base line shift resulting in small 2 A ES negative signals between the calibration channels 20 40 60 80 100 120 DB channel number 7 Test System Performance and Module Qualification Studies 97 i 1 129 257 385 Entries A O A O Peak wo Inv Mode wo Cal Module 30216630300027 513 Channel Figure 7 15 Hit map of cosmic ray detection shows a homogeneous distribution over the complete module 7 4 2 Cosmic Ray and Radioactive Source Signals Detection To obtain more precise informations about the detector response cosmics rays and radioactive sources are used While cosmic rays have the advantage of being perfect MIPs if detected accordingly they lack higher counting rates which makes cosmic ray runs time consuming Within both test stations the rate of cosmic rays is about one particle per minute which results in a half day run time to collect sufficient statistics gt 300 events for 5 accuracy Radioactive sources on the other side deliver sufficient rates but with lower energy resulting finally in increased energy losses compared to a real MIP 7 4 2 1 Angular Acceptance For an accurate calibration both test stations are equipped with scintillators and photomultipliers PMs for cosmic ray detection In both cases the mechanical placement of the scintillators see Fig 6 16 on page 81 of the Fast Test Station FTS and Fig 6 18 on page 83 for the Diagnostic Test Stat
244. s is stored in a dual port memory DPM FED Manual The data is fetched by an Xilinx field pro grammable gate array FPGA unit for further processing Currently only the frame search is implemented which extracts header pipeline column address error bit and channel data of an APV25 frame and passes them on to the peripheral component interconnect PCI in terface for readout Furthermore the FED has a 32 bit bunch crossing and a 16 bit trigger counter and adds their values to the APV25 data The bunch crossing number can be used as a time stamp with relative accuracy of 25ns In future versions also signal processing such as channel reordering pedestal subtraction common mode correction zero suppression and clustering algorithms will be included in the FPGA to reduce the amount of data 4 2 4 Control and Monitor Path The internal tracker control and timing distribution system consists of a set of services dedi cated to the distribution of the LHC bunch crossing clock trigger signals and the supervision of the embedded front end electronics via a dedicated network The building blocks of the CMS tracker control chain are the front end controllers FECs on the back end side and communication and control modules CCUMs on the front end side which interpret and distribute the signals received from the FECs via the digital optical links 4 2 4 1 FEC The front end controller FEC is the master of the communication network While in princ
245. s out that the BPp 10105 52 5 HV module has a power limitation of 1W which means that the current delivered by the module may be significantly larger than the specified 1mA for voltages below 1kV 66 6 2 Hardware Components arzo tU eee on ERN 32 208 1 a MIO iseg Monitor kill Kill D 0 7 BPp 10 105 5 2 5 f y sBitData Q HV return PLD lt o A u ee 62 p HV line DAC DAC cS Select 3 Remote AD 557 AD 557 lie cS Strobe ALTERA gt gt MAX7064 SLC44 6 Acknowledge Somp kil Figure 6 5 Karlsruhe HV card picture and block diagram Like all cards of the Karlsruhe system the HV card is connected to the motherboard and a PLD performs all communication tasks The iseg HV module mounted on back side delivers the voltage according to the applied remote voltage which can be set by a DAC The current drawn is measured on the HV line and the isolation amplifiers output can be read out by an external MIO card as well as the HV modules voltage monitor output via top connector Furthermore an interlock system is implemented by comparing the current measurement to a reference voltage delivered by a second DAC If the current signal exceeds the reference voltage a kill signal is delivered via the PLD to the iseg HV module Heier 2003 monitor uses the control line of the ISEG module Monitor while the current mo
246. s work on a new silicon strip detector and the confidence he had to give me the freedom in organising the works connected to the module readout I especially have to thank him for patience and the support he gave especially in periods of slow progress while waiting for early prototypes For giving me the honour of being my Co referent I have to thank Prof Dr Wim de Boer Special thanks I would have to give to Dr Siegfried Weseler f 2001 who inspired a lot of my work With his tremendous knowledge on electronics as well as experimental physics and its instrumentation he helped designing the test stations as well as the test strategies 1 am sure that he would have like to see the first modules to be qualified on a Karlsruhe test system Dr Peter Bl m gave important contributions to this work by coordinating the activities of the Karlsruhe CMS group Without his tireless work a lot of useful informations would have been missed Especially I have to thank him for reading the manuscript and for a lot of profound discussions I have to thank Dr Hans J rgen Simonis for the discussions we had not only about technical aspects Furthermore I have to thank him for the patience he had dealing with all the request and ideas I addressed to him concerning the local computer network Dr Frank Hartmann contributed to my work by giving inspiration and hints as well as by requesting and organising works to do He always answered my frequent questio
247. same as for the inter strip resistance Rint measurement expect that they are connected with poly silicon resistors to a bias ring Additionally on each side three outer strips are shorted together The inter strip capacitance Cint is measured between the central strip and its two first neighbours the six outer ones being grounded with the bias line The surface currents generated at the MIS interface in the inter strip region are charac terised with the gate controlled diode GCD This device is basically a comb shaped diode intertwined with a comb shaped MIS in such a way that the gate voltage applied through the MIS can influence the surface component of the diode leakage current The diode leak age current variation is measured versus the gate voltage for a constant bias voltage The surface current is the main parameter extracted From this IV curve the flat band voltage corresponding to the passivation insulator is also determined To check the poly silicon resistors Rpoly the p implant and the aluminium resistivity Qs a series of resistances have been gathered in a dedicated substructure Sheet There are three poly silicon resistors identical to the ones found on the sensor itself for the biasing of each strip Three p implanted strips with different widths 10 20 and 50mm are also implemented to extract the resistivity of the implant and to get a value of the global doping 5 Quality Control at CMS Tracker Modules 49 Figur
248. se for Peak and Deconvolution mode both 92 7 2 Noise Studies Sensor channel number 500 400 300 200 100 Te e EETA EE k i i z E lt 8 Peak mode 4 A det e E E 3 oa Figure 7 7 Raw noise comparison between Peak and Deconvolution z mode The upper plot amp o shows the raw noise for 2 c Peak mode and Decon z c volution mode raw noise is shown in the middle plot Both modes are without calibration and 2 a m T O without inverter circuit 2 A E A ne 3 x A 3 8 used The difference is 5 23 End a o gt 8 visible in the lower plot 3 S 3 o and shows that the Peak e N gS mode has about 25 00300 o a o 3 DB channel number E EE less raw noise than the a ss Deconvolution mode without inverter stage and without calibration circuit used The difference between these two modes shows no deeper structure and the offset between both mode is of the order of 30 40 The impression arising from Fig 7 7 that the Deconvolution mode might have a broader raw noise distribution is misleading Although the slope of the raw noise for both modes complicates the RMS calculation it turns out that for both plots the RMS is dominated by statistical fluctuations of the order of 0 08 ADC The APSP filter reduces the amplitude of the APV25s edge noise which does not surprise because the APSP algorithm adds three consecutive samples with according weigh
249. sed in the following sections 3 2 Magnet System The choice of the magnet system meant the starting point of the CMS detector design A long superconducting solenoid of 12 5 m length and 6 0 m inner diameter with a uniform magnetic field of AT has been chosen The magnetic flux is returned via a 1 8m thick saturated iron yoke which is instrumented with muon detector stations The inner coil diameter is large enough to accommodate the tracker and calorimeters Since the magnet is the main element of CMS in terms of size 21 6m length 15m di ameter weight 125004 and structural rigidity it is used as the principal structural element to support all other detector components The return yoke consists of five outer barrel rings and two end caps with three separated disks both Each barrel ring is made up of three iron layers The superconduction coil system and its sub systems cryogenics power vacuum and quenching protection are attached to the fixed central ring The other rings can slide on rails parallel to the beam direction to allow insertion and maintenance of the muon stations For details on the magnet system see the CMS Collaboration 1997c and recent publications like Herve et al 2001 16 3 3 Central Tracker SUPERCONDUCTING CALORIMETERS COIL EGAL HCAL Scintillating Plastic scintillator brass PbWO4 crystals E sandwich IRON YOKE Silicon Microstrips Pixels MUON M ON BARRE ENDCAPS Drift Tube Resistiv
250. sesignal welche im CMS Detektor in einem gemeinsamen Signal kodiert bertragen werden wobei die Ausl sesignale als fehlende Taktzyklen kodiert sind Weiterhin synchronisiert der PLL Chip die Phasenlage des Taktsignals und kann das Auslesesignal verz gern um so unterschiedliche Signal laufzeiten auszugleichen APV25 Der APV25 Chip ist das Herz der Auslesekette Er lie t parallel 128 Detek torkan le aus verst rkt deren Signale und speichert diese analog zwischen vgl Abb 3 Der Vorverst rker und der anschlie ende Pulsformer sind direkt an eine analoge Pipe line angeschlossen Auf diese folgt ein APSP Filter welcher das vom Vorverst rker und Pulsformer gefaltete Signal zu entfalten vermag Dies ist n tig da der Vorverst rker und Pulsformer das Eingangssignal auf etwa 100 ns verbreitert was zu einem Signal in mehreren aufeinanderfolgenden Takten f hrt Dies l t sich vermeiden indem man das Signal aus drei aufeinanderfolgenden Taktzyklen kombiniert und so im APSP Filter die Verbreiterung des Signals herausrechnet Falls der APSP Filter genutzt wird spricht man vom Deconvolution falls er nicht genutzt wird vom Peak Modus Weiterhin kann zwischen dem Vorverst rker und dem Pulsformer eine Inverterstufe zugeschaltet werden um so unterschiedliche Eingangspolarit ten zu verarbeiten Die Daten der 128 Eingangskan le werden ber einen 128 1 Multiplexer seriell aus gegeben wobei dessen Baumstruktur die Reihenfolge der
251. signal collection time which decreases due to larger charge carrier mobility x T On the other hand the noise decreases because the noise sources have a decreasing tem perature dependence While this is obvious for the thermal and shot noise the flicker noise decreases due to decreasing leakage current Furthermore the APV25 has been designed to be operated at 10 C which means that the amplifier and shaper time constants are best adapted for signals arising for silicon strip detector at that temperature Figures 7 21 and 7 22 on the next page show the measured SNR as function of the ambient temperature For both modes the SNR increase from room temperature 22 C down to 10 C is about 20 SNR SNR 102 7 5 Signal to Noise Ratio Studies 36 E g F 519 5 RR ae E E T 19 34 E 33 les i 325 18 ab 175 E En E Ar 28 e E 38 E i 29 28 z F zg 16 5 T E E e gt E A O E LA Eye elena eles leerer AE 10 5 0 5 10 15 20 83 10 5 0 5 10 15 20 Temperature C Temperature C Figure 7 22 SNR vs temperature in De convolution mode at a bias voltage of 300 V taken with cosmic rays on the FTS Figure 7 21 SNR vs temperature in Peak mode at a bias voltage of 300 V taken with cosmic rays on the FTS 35 x 307 ha 5 F m C C n E L he H a 30 E BT A La C s A a a 20 25 as B lo E E To 15
252. sistors is used to bias the implant strips these resistors are connected to one end of the implant strips Their mean value is around 1 8 MQ which is a compromise between two different needs a low noise contribution which implies a high value and low voltage drop across the resistor when the detector is irradiated and leakage currents are increased On the backside a uniform metallised n layer is present providing a good ohmic contact In addition on the junction side an n implant is required over the entire cutting line as well as metallised pt guard and bias rings surrounding the active area of the sensor To protect the sensors during the assembly phase the front side of the sensor except some regions required for contacts and bonding to the metal layer will be passivated The alignment tolerances with respect to any mask is required to be 2 um maximum Sensors should be stable in time without breakdown below 500 V and excesses of noisy or faulty strips before and after irradiation The goal is to receive sensors that once assembled in a module will result in a fraction of bad channels below 2 The leakage current for a 6 sensor 80cm measured at room temperature and 450 V reverse bias should not exceed 10 pA The total effective capacitance of each strip C strip measured as a sum of coupling capac itance to the two neighbouring strips Cint on each side and the capacitance to the backplane Cy is expected not to exceed 1 3 p
253. strip Migliore 2002 Raymond 2001b For a typical ring 6 module the measured values for the interstrip capacity of an individual sensor results to 3 6 pF calculated from the sensors total capacity the coupling capacitance to 498 pF and polysilicon resistor values to 1 85 MQ While the coupling capacity matches the specifications perfectly is the interstrip capacitance better and the specification which results in a total capacity of 15 6 pF which is significantly lower than the 22 7 pF obtained from the sensor specification The other values of interest for noise calculation are the metal strip resistance Rs of 200 Q and the 50 ns shaping time of the APV25 For this values Tab 4 5 lists the contributions of the different noise sources Noise source Peak mode Deconvolution mode Reverse bias current 8 1068 3 480 Bias resistors Rpoly 165 74 Metal strip resistors Rs 425 627 Front end electronic 808 1316 Total 928 1415 1460 1537 scaled from 0 1 A for non irradiated module to 1mA after 10 years of LHC Table 4 5 Noise contributions for Ring 6 Modules in equivalent electron charge Calculated for T 50 ns inp 0 05 u A 1mA homogeneous distributed over all 512 strips and both sensors Ryoly 1 85 MQ Rs 220 2602 Crop 15 6 pF This total noise values together with the expected MIP signal calculated of an effective thickness of 470 um predicts a SNR value for non irradiated modules of 37 2 in Peak mode and 23 6 in Deconvoluti
254. sults in a gain of fault detection capability As discussed in Sec 5 5 3 on page 56 about error detection most of the strip errors have a characteristic coupling to the noise Finally the noise measured by the readout system is the fluctuation in the digital data captured by the FED Hereby the digitalisation itself adds a quantisation noise of the order of Q1sB v12 where Qrspg is defined as the charge when the LSB changes Due to the fact that the FED uses a 10 bit ADC and removes the LSB the integral non linearity INL error deviation from a linear characteristic as well as the differential non linearity DNL error variation of the 1 LSB change corresponding analogue step size can be neglected Friedl 2001 Nevertheless the quantisation noise has to be taken into account for FEH testing In Peak mode the APV25 has a ENC of 246 e 396 e for Deconvolution compare Tab 4 4 on page 44 and Q sp is of the order of 500 e see Tab 7 2 on page 101 7 2 1 Pedestal and Raw Noise Pedestal and raw noise are measured by calculating the mean and RMS of a raw data sample of sufficient size Typical pedestal and noise runs within this work consist of 1000 events if not stated otherwise Thus the resulting accuracy is of the order of 3 for the noise values The analysis software calculates the pedestal and noise on the fly which means that the data set is not reprocessed after the mean calculation Nevertheless the first 100 events are p
255. symmetries at higher mass scales such as Supersymmetry which would manifest itself in a new spectrum of high mass particles or left right symmetry which would entail the existence of new vector bosons More hypothetical but spectacular is the prospect of finding indications of extra dimensions in space Last but not least high precision and high rate beauty physics are to be explored at the LHC To pursue these physics topics high resolution track and vertex reconstruction are vital Search for the Higgs boson is one of the prime goals of the LHC It turns out that for light Higgs bosons decaying mainly to a b quark pair excellent b tagging performance and good mass resolution as well as jet separation are essential for the discovery potential of a given detector which calls for a high resolution tracking detector The CMS silicon tracker consists of 15 232 detector modules as the smallest indepen dent units Production and assembly of these modules will span two and a half years period during which it is essential to guarantee a continuous production quality using defined control procedures and acceptance criteria to avoid expensive and non replaceable production failures Part of this quality control chain has to ensure functionality and reliability of the final silicon modules produced The CMS group in Karlsruhe is involved in the construction of the silicon trackers end caps and will produce and qualify the 1600 modules of ring 5 Therefore
256. system based on peltier elements and is designed to keep time needed for a complete test cycle short enough to stay compatible with the production scheme of producing four modules per day including a complete qualification test The other setup has to be as flexible in usage as possible This system is designed to be used for debugging purposes and is called Diagnostic Test Station DTS It will make use of additional test possibilities like laser micro manipulation probes sources and cosmic rays Both systems are based on the same hardware devices developed from the Karlsruhe CMS group 6 1 Karlsruhe Readout System Figure 6 1 on the next page shows the basic hardware setup Based on a standard personal computer PC equipped with the FED as ADC an I C card additional readout compo nents are developed as replacement for the much more complex and expensive trigger timing and control system used in the final experiment Furthermore embedded slow control and additional diagnostic tools for quality control are developed 62 6 2 Hardware Components Test Statio TI mene MT O A Figure 6 1 Hardware layout of the Karlsruhe readout system Based on an external mother board several dedicated functionalities are realised as small devices controlled via the moth erboards bus system which is connected to the PC either through a DIO an PC link or via parallel port For data digitisation a FED is used and for the PC link
257. t and this will lead to a damping of non correlated CM noise The inverter stage s influence on the raw noise behaviour is shown for Peak mode in Fig 7 8 on the facing page The raw noise with the inverter stage on is reduced to the same level as the CM corrected Peak mode without inverter compare Fig 7 5 on page 90 Obviously the inverter stage reduces the CM sensibility which can be explained by the fact that the inverter stage is driven through a 100 Q resistor connected to VDD 2 50 V while for inverter off configuration is driven from GND 1 25V Also the APV25 edge noise is reduced significantly in inverter modes Civinini 2003 The calibration circuit also influences the raw noise behaviour which is shown in Fig 7 9 on the facing page Two effects are obvious just by comparison of the upper and the middle plot of 7 9 First one sees that the calibration increases the raw noise asymmetrically and the second is that the edge noise is increased The difference plot on bottom of Fig 7 9 on the next page shows a systematic in the noise increase This effect of the calibration circuit is only visible in Peak mode with inverter on For the other Peak mode nearly no influence is visible at all and for Deconvolution mode only a slight edge effect is detectable Nevertheless the effect is small and for the detector qualification 7 Test System Performance and Module Qualification Studies Raw noise ADC Raw noise ADC Raw noise d
258. t forward bias alone gives not a sufficient tag for a clear and unambiguous pinhole identification 7 7 5 Bad Poly Resistors Although bad poly resistors are detected once in a while from the QTC no module with an bad poly resistor has been measured so far Furthermore from the readout side a bad poly resistor fault is hard to disentangle from other fault types because there is no clear signature available In principle a strongly reduced poly resistor should be sensitive to CM noise from the bias line and show an increased noise similar to the sensor edge channel noise discussed next compare also Sec 5 5 3 6 on page 59 7 Test System Performance and Module Qualification Studies 115 Sensor channel number 500 400 300 200 100 Figure 7 46 Forward bi ased 9V module noise re ME A l veals among other also the pin hole at channel 420 On this module channel 162 has a pin hole too and in the inverter on mode both can be tagged clearly Without inverter the short of the channels 61 amp 62 is also visible Furthermore a distortion of unknown ori gin is visible in the last APV25 Noise ADC o Noise ADC o z PAE E eet nn Without Inverter fe 3 i i i 100 200 300 400 500 Peak wo Inv Mode wo Cal Module 30200020000638 DB channel number channels 385 512 Sensor channel number Sensor channel number 500 400 300 200 100 Noise ADC gt a MPAA AT ATT
259. t lines from the corresponding FEH which are connected via a short lt 30cm 100Q transmission line About 12900 of these opto hybrids will be used for the CMS tracker readout links The total power dissipation of individual chips must remain constant regardless of the modulation signal to minimise cross talk and noise injection in the common power supplies Troska et al 2002 4 2 3 Front End Driver On the back end side of the readout chain the front end driver FED receives and digitises the analogue data streams from the APV25s While the current prototype based on a PCI Mezzanine card PMC format has only an electrical receiver implemented the final FED will be a 9U Versa Module Eurocard VME board including analogue optical receivers for 96 channels Furthermore the final one will have a TTCrx receiver for clock and trigger input and a high speed interface DAQ link to the subsequent event builder The overall input data rate of the final FED will be 3 1 GB s condensed to about 100 MB s at the DAQ link output Both prototype and final FED digitise the analogue inputs by 10 bit ADCs of which 9 bits are read out at 40 MHz Similar to the PLL ASIC the final FED can adjust the sampling time within a clock period for optimum sampling performance for every channel while the prototypes supports this feature only globally which makes sampling performance optimisation difficult After receiving a trigger a programmable number of sample
260. term test while during the mass production this reliability test will be done on a full scale for equipped petals 5 4 3 Petal Integration Centre The five BC distribute the modules to the six Petal Integration Centres PICs located in Aachen Brussels Karlsruhe Louvain Lyon and Strasbourg Each PIC takes the responsibil ity for one of the 6 groups of petals These groups consist of front or backside petals of the discs 1 3 46 and 7 9 Independent of this all petals have the same structure Based on a honeycomb structure with embedded cooling pipes the interconnect boards ICBs are mounted These ICBs pro vide all needed services to the modules and establish the connection between the FEH and the Opto hybrids as well as their connection to the CCUM and power supplies During the assembly to the petals functionality tests have to be performed after mounting of new components because of the high density of fragile components First the CCUMs and Opto hybrids will be mounted and tested After that the non overlapping modules of a ring are installed and their functionality is tested by the means of an 12C scan and pedestal noise analysis before mounting the second layer This procedure is repeated till a power group is completely installed and tested Afterwards the integration of the next power groups starts following the same procedure As soon as a complete petal is integrated it is placed in a dry storage and the long term test is start
261. the low resistance of the bulk material during irradiation by putting both alternating current AC pads and bias ring on ground and 1V on the back plane In fact for practical reasons a voltage of 10 15 V is chosen After irradiation all the structures are stored at 18 C except for the time of controlled annealing 80 minutes at 60 C to reach the minimum of full depletion voltage The structures are then qualified at 10 C Typical changes due to irradiation are the leakage current Ileak increases significantly as expected and influences the measurement of the bias resistance the coupling capacitance C e decreases slightly by about 1 pF and no increase in the dielectric current Igie is observed the inter strip resistance Rint is initially out of the measurement range and after irradiation in the order of few hundred MQ up to GQ The important parameter with respect to noise is the inter strip capacitance Cint which increases only slightly 10fF Dierlamm 2003 5 3 Quality Test Strategy for the Front End Electronics For the FEH the CMS group at the University of Strasbourg acts as control and distribution centre They get the ASICs from the groups in the United Kingdom and distribute them to the industrial producers Cicorel Boudry Swiss and Hybrid SA Chez Le Bart Swiss to build the FEH 5 3 1 Front End Hybrid Industrial Test The FEHs are assembled outside of the CMS collaboration The companies produce the flex PCB
262. the supersymmetric structure of the theory requires at least two Higgs doublets to generate mass for both up type and down type quarks and charged leptons which leads to at least five supersymmetric Higgs bosons one lightest CP even h one heavier CP even H one CP odd A and two charged HF as it is described in Kazakov 2000 If supersymmetry is realised by nature and connected to the SU 2 symmetry breaking of the electroweak interaction the LHC will be able to find traces of it For the MSSM Higgs sector the LHC covers nearly all the theoretical allowed parameter space although ensuring a 50 discovery over the entire plane requires more luminosity Womersley 1997 Furthermore if supersymmetry is relevant to the electroweak symmetry breaking problem then most of the supersymmetric particles will be in a mass range that is observable at LHC Both ATLAS and CMS have enormous potential to discover supersymmetry if it exists at mass scale less than about 2 TeV Asai 2002 2 High Energy Physics at the Large Hadron Collider 9 2 4 New Physics Beside the search for Higgs bosons also other theoretical models will be probed at the LHC This includes e New gauge bosons as they are a generic prediction of superstring theories which have additional U 1 gauge groups motivating the search for additional W and Z bosons e Technicolor as a model of strong electroweak symmetry breaking predicts resonances which deca
263. tool rack as they are used for the TEC community in the module assembly centres MACs in Brussels and Lyon This so called gantry system is equipped with high speed X Y linear motor stages and Z and rotary drives Furthermore the Z drive is equipped with a charge coupled device CCD camera to spot particular fiducial marks on the individual detector components An additional camera is installed on the mechanical support structure of the Y linear motor to view the dispensing of glue With this dedicated setup up to four detector modules can be assembled at the same time All components are first placed on the supply platform and the frames on the assembly plate Glue is then dispensed on the carbon fiber frames using specially developed glue dispensing tools The components are then placed by the robot arm in their final location on the carbon fiber frame The proper handling of the very delicate sensors during the pick up and operations is assured by means of flat Teflon coated vacuum pick up tools with built in pressure control sensors Once the components are positioned vacuum valves are enabled to 5 Quality Control at CMS Tracker Modules 53 Figure 5 9 Automatic assembly robot as used in the CMS gantry centre in Bari Italy In front the assembly plate for four TID modules is visible Behind it the supply plat form and the tooling rack can be seen Sur row 2001 secure the detectors and the frames into place on the vacuum chuc
264. trans mitted to the FEH and the corresponding trigger signal which is send to the FED as well setLedLatency is quite similar to the setLatency except that it manipulates the dis tance between the trigger signal send to the LED system and the trigger for the FEH and the FED sendReset sends the 101 reset pattern to the FED enableExtTrig and enableExtLedTrig switch on the external input To prevent pile ups the method set WaitCycles allows to set an inhibit for up to 999 clock cycles The external trigger input stays active until the next software generated signal is send to the SEQ HV library contains the implementation of the HighVoltage class which controls the access to the HV card The constructor takes a pointer to the Config class as argument and gets from the Config class the slot number of the HV card and the slope and offset values needed to calculate the DAC values to be applied when the set Voltage float method is called Furthermore it serves methods to open the interlock relay hvSwitch and to set the interlock reference voltage setInterlock Voltage Led Array library contains the LedArray class Beside the default constructor taking like most of motherboard hosted cards a pointer to the Motherboard class and the slot number it contain the methods powerSwitch turns on or off the LED array itself as described in Sec 6 2 5 on page 67 set Voltage which set the DAC output voltage and control the amount of light emitted
265. ts de tectors under construction is the Compact Muon Solenoid CMS detector with its Central Tracker Hereby the Silicon Strip Tracker SST alone will instrument an area of 210m with 15 232 individual detector modules of in total 15 different geometries A detector of this size can only be built in a collaboration of many institutes all partic ipating in an industrial like mass production Over 20 institutes and companies worldwide are involved in the Silicon Strip Tracker SST production each of them specialised on one or several steps of the production Hereby the Institut fiir Experimentelle Kernphysik of the Universitat Karlsruhe TH participates in sensor testing wire bonding of modules their testing and their integration into larger substructures of the tracker end caps In Karlsruhe 1600 modules will be bonded and tested While the bonding is done on an industrial automatic bonding machine the test systems of the final modules have to be developed by the collaboration This contains not only the technical part but also the test strategies and the final automatisation suitable for an usage of the test systems by technicians This thesis introduces all these aspects as well as the impact of the author s work on the testing procedures and the final design of the Silicon Strip Tracker SST module In Sec 2 a short overview of the Large Hadron Collider LHC physics program is given followed by a brief introduction to the Com
266. ts of the bond foot is still visible on the first AC pad of the unconnected channel 323 bonding process and will report failed bonds immediately Usually the bonding machines will bond the sensor to sensor or the sensor to PA connections without interruption and a bonding speed of the order of one bond per second producing a very uniform bond quality in terms of shape and strength Nevertheless bad bond connections are a common problem although they are not frequent Two classes of bad bond connections have to be distinguished A single bad bond in the form of a lost connection due to a surface irregularity or even more frequent simply due to a dust particle And as a second class a bad bond may be an indicator for a general bonding problem resulting e g in a series of bonds with reduced pull strength While the occurrence of the first type can be minimised by careful cleaning and working in a proper environment reflects the second class a more general problem The detection of a single missing bond has to be covered by a fast functionality test performed right after bonding which allows an immediate repair Figure 7 39 on the next page shows the signature of a missing bond in the calibration amplitude and the noise for the Peak without inverter mode The calibration amplitude is increased due to the reduced capacitive load For the noise a reduced capacitive load would result in a decreased noise but only if no CM noise is in the system Du
267. ual to its board ID which can be selected by a hex switch on the motherboard If the PLD find its own ID it will select or deselected the slot coded in the three LSB 2 0 Thus before communication with a card plugged onto the motherboard the corresponding slot has to be selected All cards on the motherboard have to implement the handshaking via Logic true is represented by voltage level 0 64 6 2 Hardware Components Figure 6 3 Karlsruhe motherboard schematic the DS and ACK lines them self Therefore they are equipped with a PLD for performing communication As soon as the communication is finished the slot has to be released again to unblock the communication to the other slots For details see Heier 2001 6 2 2 PLD Sequencer The sequencer SEQ is the most important card of the Karlsruhe readout system because it provides all clock and trigger signals and thus replaces the complete fast control path of the final readout system compare Fig 4 5 on page 27 The layout of the sequencer SEQ can be seen in Fig 6 4 on the next page The idea behind the SEQ is to use a random access memory RAM and feedback part of its data lines to the address lines The final design uses 10 out of 18 data lines for the feedback address which results in sequences of a maximal length of 21 1024 The needed logic of the SEQ is implemented in two PLDs One performs all communication tasks to the PC via th
268. ure additional cooling done by ventilation 6 Karlsruhe Test Stations 73 6 2 9 I2C The I C card has been designed from the EP group at CERN implementing a I C bus on a PMC format compare App E on page 137 for details on the I C bus specifications Due to its PCI compatible interface and plain design it can be easily incorporated as well in standard PCs as in VME based systems The building block of the 12C card is a PCF8584 from Philips Philips 1997 providing one I C bus with a maximum speed of 90kHz The PCF8584 acts always as the master of the connected 12C bus which means that only this card generates a clock On the other side the PCF8584 is connected to a local 8 bit parallel bus which is interfaced by a PCI9050 from PLX PLX 1999 bridge to the PCI bus The bridge configuration is stored in a EEPROM which makes this card fully PCI 2 0 compatible see Fig 6 11 on the preceding page The output lines of the I C bus are galvanically decoupled by using opto couplers driven from the slave side which keeps the GND of the PC and connected slave side decoupled Furthermore one has to ensure that the slot cover does not get into ohmic contact with the PC cover because it may also get in contact to the slaves sides GND by the connectors plug 6 2 10 DIO A digital input output DIO card is used to address the motherboard in some setups and therefore the DIO card PCI 6503 from National Instruments is used National Instruments 199
269. urns to the sender This node verifies the FEC acknowledgement and a checksum If the transmission was successful an empty token is inserted in the ring instead of the data frame otherwise the data packet is resent The token ring architecture minimises the connections needed between the stations but relies on the functionality of all links If a single connection is broken the whole ring will get of control To overcome this risk the final configuration of the CMS tracker control token ring will be redundant in a way that there is a second staggered ring Each station will have two interfaces and automatically detects which one is in the currently active ring 4 2 4 3 Digital Opto hybrid The digital optical link uses the same components as the analogue link for the transmitter section located in the radiative environment of the ex periment Since two way communication is required on the control part pin diodes and a radiation tolerant receiver ASIC will be included on the digital opto hybrids as well The best digital signal transmission on fiber optics is obtained with DC balanced data containing the same amount of low and high level bits on average as it is implicitly the case with the clock For the data line this is achieved with a special four to five bit encoding and the non return to zero with invert 1 on change NRZI scheme CMS Collaboration 1998 Basically this scheme uses a line transition to represent a 1 and no transition to r
270. us Decon Modus minimale Korrektur mittlere Korrektur SNRmp 34 53 0 37 19 00 0 43 19 00 0 43 Korrekturfaktor 1 015 1 075 1 180 SNR Korr 35 0 0 4 20 440 5 224 0 5 Rauschen Mess 1 97 2 72 2 72 ENC Mess 986 12 1691 45 1540 45 ENCT heo 928 1460 1460 ADC 501 33 621 39 566 35 Tabelle 2 SNR Messung f r Peak und Deconvolution Modi Ein MIP in 500 um starken Sensoren 470 um effektive St rke erzeugt 34500 Elektron Lochpaare Die Messung der wahrscheinlichsten Energiedeposition SNRyp und die akzeptanzkorrigierte Deposition SNReorr verglichen mit dem berechneten Wert zeigt eine gute bereinstimmung welche f r eine absolute Kalibration der ADC Werte genutzt werden kann Abbildung 7 zeigt das Signal zu Rauschverh ltnis SNR im Deconvolution Modus Das SNR wird in diesem Fall durch Winkelakzeptanzen und die Ausl seimpulsakzeptanz verbreitert wobei insbesondere letztere gro en Einflu haben kann Zeitgleiche Signale von den Photomultiplieren werden hierbei direkt und den Sequenzer weitergeleitet welcher dann entsprechend um Laufzeiten korrigiert ein Auslesesignal f r den Detektor und die Auslese generiert Hierbei kommt es zu zeitlichen Schwankungen welche selbst bei op timaler Konfiguration zu Signalverlusten f hren W hrend der Peak Modus relativ un empfindlich auf diese Schwankungen reagiert 1 5 ist der Deconvolution Modus aufgrund seiner deutlich k rzeren Signalform sehr empfindlich auf diese Sc
271. ven address useful for debugging purposes void FED_status to check the configuration and to print it to the console I2C library The I2c library contains the I C class which encapsulates the access to the I C card The I C class opens and closes the device in its constructor and destructor methods Reading and writing to the 12C bus is performed through read and write methods taking both three arguments The first argument specifies the slave address which will be attached the second argument is a pointer to a buffer either containing the data to be send or being the destination for received data The last argument gives the number of bytes to be transmitted The return value of the read write methods indicates failures during the transmission like got no slave acknowledge or bus busy APV library as shown in Fig 6 12 on page 75 contains two classes The first one is the Apv class containing the implementation of methods needed to control a single APV25 via its 12C interface This class contains two overloaded methods setReg allows setting of the APV25s internal registers either by their address or by their name both referring to the APV25 manual Jones 2001 or by a construct holding a complete set of all APV25 registers getReg reads back the internal registers and is also overloaded like the setReg method getError evaluates the error register of the APV25 used to identify whether a FIFO overflow or an pipeline error occu
272. visual inspections or to control the placement of micromanipulation probes as they are used for IV measurements before bonding Furthermore a laser fiber or a source can be attached to it which allows strip scans to detect localised defects even in the sub pitch regime Finally radioactive sources can be used inside the station Therefore two fiber scintillator arrays are placed below the module which are connected to PMs on the backside one PM is partially visible on the left 6 5 1 Linear Gate System The linear gate system embedded in the DTS is built of components from isel which include two belt drives a linear motion stage moving through a spindle and two stepper motors The linear gate system is assembled on an aluminium plate of 10 mm thickness and spans 28 cm in the x direction parallel to the front of the DTS and 45 cm in the y direction The gates accuracy is 6 ym in the x direction and 150 um in y the latter one is sufficient if the modules readout strips are in parallel to the y axis of the gate system The system is driven by two stepper motors which are controlled by a commercial stepper motor control unit The controller is connected to the readout PC via the serial port During the initialisation of the gate the controller drives the head to the park position in the right backside of the station Here are two switches used as reference to determine the heads position Afterwards the systems keeps track of the heads positio
273. w oriented way of programming like it is used by LabView supports the programming of flow oriented control tasks The user can control the behaviour of the readout system by changing the corresponding parts in the control files These files are stored in the config folder and contain settingsconfig contains the definition of the keys that can be used by the other control files a description used by the GUI limits and the corresponding socket command name as well as the clients addressed by this command The settingsconfig file is parsed by the GUI and it keys are accessible through the settings menu compare Fig 6 14 on the next page settings file hold the default values for the keys defined in the settingsconfig file These 6 Karlsruhe Test Stations 79 OHistogram View Settings y Write to Database v Change Device Y Load Settings v YI Curve O Fasttest o gdi data run gdi pinhole run Pedestal Run O LED Run o Particle Run Calibration Run Calib Pulse Shape Laser Run Pipelinescan Pedestal Pipelinescan Calibration Single Event Run Single Command o Testrun o Testentry Stop Temperature Control Humidity Control Motor Control 03 04 13_15 40 13 RUN_PED finished 1062 events done 03 04 13_15 40 14 RUN_PED started going to take 1000 events 03 04 13_15 40 21 RUN_PED finished 1062 events done 03 04 13 15 40 21 RUN_CAL started going to take 3072 events 03 04 13_ 15 40 34
274. which causes a current of 260 yA that is limited by the poly resistors and the 22kQ resistor in the HV return line 7 Test System Performance and Module Qualification Studies 109 Sensor channel number 450 445 440 T TER E T Figure 7 37 Shorted strips signature in noise for Peak modes solid lines and Decon volution modes dashed lines The upper plot shows the ef fect of shorts for modes with out the inverter and the lower one for modes with inverter used For Peak w Inv Mode lower plot solid curve a de fect is visible in the noise but the structure does not reflect the topology of a short Noise ADC Noise ADC Peak w amp wo Inv Mode wo Cal Module 30200020000638 70 75 DB channel number Bias Peak Mode Deconvolution mode voltage wo Inv w Inv wo Inv w Inv Noise Cal amp Noise Cal amp Noise Cal amp Noise Cal amp 400V tag tag visible tag visible tag visible tag 20V tag tag visible tag visible tag visible tag OV invisible tag tag tag invisible tag tag tag 9V invisible tag tag tag invisible visible tag tag Table 7 3 Signature of a short as measured for channels 61 amp 62 on Module 30200020000638 for different bias voltages 9 V means that a forward bias of 9 V has been applied to the module A tag requires a clear structure which can be identified easily and is ranked fro
275. with a wavelength of 1050 nm As discussed in Sec 4 3 3 one expects an increasing noise in the order of one ADC from the bias resistor Rpoly compare Tab 4 5 and Tab 7 2 on page 101 The mean noise in Fig 7 29 on the facing page increases from 2 23 ADC at a leakage current of 132 uA to 2 91 ADC at a leakage current of 898 yA This corresponds to an increase of 30 while the calculation from Sec 4 3 1 on page 38 predicts an increase of 36 The difference probably results from an inhomogeneous distribution of the leakage current over the sensors strips compare Fig 7 30 on the facing page 7 6 2 Infrared LED signals Both types of infrared LEDs used SHF 4301 from Infinion and LED1050 03 from Roithner Lasertechnik have rise and fall times of 10ns which is comparable to the signal collection 7 Test System Performance and Module Qualification Studies 105 Sensor channel number 500 400 300 10 Leakage current SA r r A 132 pA 9 f i ee E E ENS ae 169 HA lt 206 uA p 27 ee o TOMA Figure 7 29 Module noise on 4 x A 2 e m as function of artifical leakage a ee z Ho current in Peak mode The ES H i j 5 Hd nda alt dd 792 uA illumination is performed on 859 pA EENET TAE O the DTS with 1050 nm wave length LED homogeneously distributed over the full silicon As one might expect the mod ule noise increases with the E leakage current generated due ooo a gt DRAE TE 500
276. wo Cal w Inv Mode Dec wo Cal wo Inv Mode Dec wo Cal w Inv Mode oi Ao o DR Peak amp Dec w amp wo Inv Mode wo Cal Module 30200020000503 10 200 300 400 500 creases the noise measured by DB channel number 20 30 Mean 0 09105 Mean 0 6924 a E RMS 0 7268 2 E RMS 4 311 o p E Er Integral 1052 40 Integral 1052 250 5 uE aof 30 u E ar E 38 L E 28 E 23 10 F ss F E 0 0 x 31 3 2 0 2 3 15 10 5 0 5 10 15 eS Common mode correction ADC Common mode slope x 10 ADC channel Figure 7 4 Common mode noise and slope distribution 7 2 2 Common Mode Noise Common mode CM correction is done on an APV25 basis by a linear fit to the pedestal reduced raw data Hereby channels are excluded which show a signal of gt 20 In Fig 7 4 the constant of the CM correction defined for the middle of the APV25 and its slope are presented which both show a Gaussian distribution Comparing the slope and the constant shows that the impact of the slope to the CM correction is small but significant This implies that a linear fit is a reasonable approximation for CM As the CM correction is done on an APV25 basis its dependence on the position of the APV25 on the FEH is investigated Table 7 1 on the following page shows the measured CM noise and its slope for the four APV25s of a ring 6 module Independent of the readout mode the CM correction is typically increased by 10 1
277. y can be used for constant illumination of sensors inducing an artifical leakage current This led to the discovery of gain losses of complete readout chips induced by shorted AC coupling capacitances of several readout channels which are called pinholes Therefore pinholes must be unbonded from the front end preamplifier which requires faultless identification techniques for pinholes CONTENTS Contents German Abstract Zusammenfassung 1 Introduction 2 High Energy Physics at the Large Hadron Collider 2 1 Large Hadron Collider 2 2 2 2 EC non 2 2 Standard Model Higgs coner nn nommen 2 2 1 Low mass Higgsi cac aa nura uti an am han ana een 2 2 2 Intermediate and high mass Higgs 22 nn e 2 2 3 lt High massHiges ir y A a a ES eo EE SS 2 3 Supersymmetry au anne en a a a ab dk ed aed oe eG 2 4 New Physics S14 bit SL OE BE eR ee Ae 229 2B Physics 2 a Sti A ges Ge eek ioe ee eis Theoretical background 2 2 2 22m rn nennen 2 5 1 The Mass Difference Amp and B B Mixing 2 5 2 CP Violation in the B System 2 022004 2 6 Heavy Ton Collisions s sms scr moso hs pee Para Re a ee eG 3 Compact Muon Solenoid 3 1 Detector Design Goals 2 2 LE ee 3 2 Magnet System e sop code a ba as bbe anne ns 3 3 SGEntr l Tracker 4 een Jf os ee Ba Ken de Hess Mie 3 3 1 Pixel Vertex Detector 2 22 Co oo nn 3 3 2 Silicon Strip Tracker e 3 4 Calo
278. y into vector bosons These signals are very striking since they are produced with large cross sections and may be observed in the leptonic decay modes of the W and Z where the backgrounds are very small e Strongly interacting W s will reveal new dynamics at a scale of s 1TeV if no Higgs like particles exists e Compositeness of the quarks and leptons has no a priori reason but could explain the observed mass spectrum If they have a substructure it will be revealed in the deviations of the jet cross section from that predicted by QCD 2 5 B Physics The exploration of physics with b flavoured hadrons offers a very fertile testing ground for the SM description of electroweak interactions One of the key problems to be studied is the phenomenon of CP violation which although already discovered in 1964 by Christenson et al in the neutral Kaon system Christenson et al 1964 is still one of the experimentally least constrained phenomena Another main topic is the study of rare b decays induced by flavour changing neutral currents FCNCs transitions b s d which are loop suppressed in the SM and thus very sensitive to new physics effects During the last few years B physics has received a lot of attention both from theorists and experimentalists and we are presently at the beginning of the B factory era in particle physics The BaBar SLAC BELLE KEK and HERA B DESY detectors as well as CLEOII Cornell CDF II and DO II FermiL
279. ysics Nevertheless we know that the Standard Model SM is incomplete and several problems arising from the theory wait for experimental clues Hierarchy problem what is the origin of the large difference between the electroweak symmetry breaking scale and the Planck scale Fine tuning problem arises from quadratic divergences of the radiative corrections caused by the hierarchy problem Why are there three generations of quarks and leptons Neutrinos are described as massless particles in the SM although neutrino oscillations show that they are massive particles Still the SM has in total 19 parameters the three coupling constants of the gauge theory SU 3 c85U 2 8 U 1 y three lepton and six quark masses the mass of the Z boson which sets the scale of weak interactions and the four Cabbibo Kobayashi Maskawa CKM quark mixing parameters One of the remaining two parameters is a CP violating parameter associated with the strong interaction The other parameter is associated to the mechanism responsible for the breakdown of SU 2 U 1 y to U 1 em This can be taken as the mass of the Higgs boson The couplings of the Higgs boson are determined once its mass is given Unfortunately within the SM we have no guidance on the expected mass of the Higgs boson The current summer 2003 experimental lower bound is 114 5 GeV c With larger Higgs boson mass self coupling and coupling to the W and Z boson grow This feature has a very
280. ystem the SEQ is connected to the motherboard and a PLD performs all communication related tasks ALTERA chip on the left The SEQ provides five output lines consisting of up to three different signals each from which four can be configured with small adapter PCBs to the preferred electrical format two of them mounted on the picture All output signals are accessible through the SCSI connecter on the lefthand side Heier 2001 6 2 3 Aachen readout and control system ARCS Repeater The APV25 is not able to drive longer lines While later in the detector the analogue opto hybrid is located close to the FEH a repeater card has to be used to drive the lines from the FEH to the FED The Karlsruhe readout system uses the ARCS repeater card Bei el 2001 developed for the ARCS system at the RWTH Aachen This repeater card also regulates the low voltages needed for the FEH 2 5 V and 1 25 V and adapts the I C level for the FEH 6 2 4 HV Card The high voltage HV card utilises a ISEG BPp1010552 5 HV module ise 2002 which supports voltages up to 1000 V with a current of 1mA at maximum The noise ripples on the HV line are well below 50mV The HV card allows monitoring of the voltage of the HV line and of the current drawn on it by giving corresponding output signals which can be read out by a multiple input output MIO card see Fig 6 5 on the following page The voltage Although these are the values specified by the producer it turn

Entwicklung und Umsetzung von Strategien zur Qualitätssicherung

Contents

Download Pdf Manuals

Related Search

Related Contents

Entwicklung und Umsetzung von Strategien zur Qualit&auml;tssicherung

Contents

Download Pdf Manuals

Related Search

Related Contents

Entwicklung und Umsetzung von Strategien zur Qualitätssicherung