High rate tests of pixel readout chips for the upgraded CMS pixel
Contents
1. SYNC OUTPUT TRIGGER INPUT Figure 11 The front panel of the data generator DG2020A 4 Pattern Data Output Connector EEE STD 488 Connector Power Connector WARNING A ATTENTION A 1 0 0 wu wen omms ger RS 232 Connector CLOCK IN Connector PRINCIPAL POWER SWITCH Figure 12 The rear panel of the data generator DG2020A 4 14 2 A rooe Repeat upoaTe AUTO Pi ON Poo 1997 86 19 14 21 13 Hardcopy Block Name D lof 1 000 Resolution file C ol 0 000 A 0 009 ns ce an 10000 000 ns Mark 0 000 ns Load p 10 9 40 70 80 a o oo Save o Data s al Ben MR OO import 2 O UV UL UU UV o gi Data YO INNE DATAG3 o Export DATAOZ e Data Tie 869 DATAGO NER Fill count up pattern step 1 Move cursor to prevnext Execute Y Enhanced Make sarge mc Y See Smee baka 1 si Figure 13 The display of the data generator DG2020A showing the typical content of the EDIT menu and 2 are controlled via the bottom and side bezel buttons respec tively 4 File tk 15 Connecting the DG2020A To use the data generator only the power connector needs to be connected while the principal power switch mus2. o Sf a 50 Entries 100 E Mean 107 2 F RMS 2 255 40 30 20 10 Er JE qr du Ep qp lh ES pe 4 Ss foz 106 108 110 112 114 116 tet hit distribution pulselength 10 g E a m 50 Entries 100 L Mean 1077 H RMS 3 189 40 30 20 10 O AA AA hai 1 Lilia Poo 102 104 106 nla 108 110 112 hit distribution pulselength 12 La 1 114 116 118 tet W B toL a ki 50 Entries 100 L Mean 108 4 E RMS 4 053 40 ao 20 10 C la a tira tr br lb br dos 102 104 106 108 110 112 114 116 118 tet Figure 38 The amount of hits as a function of tct The unit length of a pulse is 25 ns 50 hit distribution pulselength 15 hit distribution pulselength 20 hits F a BUE a 50 Entries 100 L Entries 100 L Mean 110 8 60 Mean 115 4 t RMS 5 793 E RMS 8 266 E C 40r 50 30 40 6 307 20 E E 20 10 E L 10 ee ULA Loans a aandaa gobo haina Haaa 02 104 106 108 110 112 114 116 118 120 122 100 105 110 115 120 125 190 135 tot tot hit distribution pulselength 25 hit distribution pulselength 31 ggF 8 E 80 a E80 a its Entries 100 w E Entri
3. A typical distribution of the pulse height per hit rate vs Vthr a Entries 145 Mean 101 4 RMS 28 6 m N N e FTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT rate MHz cm 2 ete Pa Sac eese nce pap nee teo teen ee 20 40 60 80 100 120 140 V thrComp ce Figure 22 The rate is depicted as a function of the comparator threshold 26 4 4 Pulse Length While the early measurements where done with a haphazardly chosen pulse length the necessity of a proper investigation of its influence became pressing subsequently It was impossible though to analyse particular clock cycles with a random trigger Therefore the trigger and the pulse had to be synchronised The synchronisation was achieved by adding a synchronising signal to the se quence as depicted schematically in figure 26 One of the programmable outputs of the testboard was programmed to put out the synchronisation signal as it was done with the clock cf section 4 2 This particular sig nal output was then connected to the trigger input of the data generator which was programmed as to send the sequence only once after receiving the synchronisation signal from the testboard cf the paragraph about the run mode in section 3 3 Since the synchronisation signal was synchronous w r t to the testboard trigger and w r t the pulse of the data generator the trigger was syn chronous w r t the pulse As illustrated in the
4. BC Enes 0 B E es Meanx 0 E Meanx 20 nE Meany 0 TW Meany 10 t IT E os E ma so E 07 50 cd oe a 40 10 5 XE pa oa E F 10 3 je 20k ar E o2 10E o E lo p i l l T 10 20 36 ao 5 b 10 20 30 40 50 en column Figure 29 Two desynchronisation measurements plotted as a function of position The amount of hits is Left The Voca expected in an even trigger vanished The precedent and following trigger contain no Vo as expected Right The Vca expected in an even trigger was shifted into the following odd trigger 35 Desynchronisation 40 30 20 10 lumi La T aca aa bdo 1215 3 1215 4 1215 5 1215 6 1215 7 1215 8 event Figure 30 A desynchronisation measurement The amount of hits is plotted as a function of the event Desynchronisation 25 20 15 10 0 i Un un mn gt 1 OS 234 3 234 35 234 4 234 45 234 5 234 55 event Figure 31 A desynchronisation measurement The amount of hits is plotted as a function of the event 36 4 7 Chip Efficiency After the hit rate was well understood and a procedure was found to produce a desired rate the next step was to use the acquired knowledge to perform a more advanced test The more advanced test was in fact an early chip efficiency test for which the setup was built The efficiency of a pixel was tested by switching through a Vea an
5. 33 One can clearly see a peak at 100 96 In addition on the left a separate region can be distinguished The pixels in this region belong to 37 the particularly affected double column in figure 32 The mean efficiency of such a test was then extracted and plotted against the rate as shown in figure 34 The error in the rate was taken from the rate distribution in figure 27 The error in the efficiency was taken from the distribution in figure 35 where the distribution of 30 efficiency test is shown The efficiency tests for this distribution were however done only over the rows 40 60 to decrease mea surement duration In figure 34 the efficiency does not change within the error as the rate increases from 300 to 350 MES This is attributed to the method used to increase the rate to 350 ML In section 4 5 an upper limit to the measurable rate is mentioned This limit was reached soon after 300 MES To increase the rate anyway two of the five cables were chosen which had a negligible overlap in terms of affected double columns This means that the clusters of those two cables did lie in different double columns such as not EN to increase the data in the double column data channel 3 buffers To increase the rate an additional pulse was sent to one of those cables at the crane same time when a pulse was sent through the other cable This situation is schematically shown in the schematic on the right This method all
6. D1 D2 had to be programmed accordingly before each test Examples of how to program this are contained in the appendix The figure 16 shows two hits per non zero event distributions for the asynchronous and the synchronous case Aside from the synchronisation of 21 the clock these two distributions were obtained using the same test param eters In the asynchronous case there are LHE on the left of the clear peak Also the total number of entries was significantly larger However the rate stayed within the measurement error the same Asynchronous Synchronous event distr for 100000 triggers and rate 61 003750 event distr for 100000 triggers and rate 62 573750 S Entries 1666 O Entries 1480 500 Mean 58 59 gr Mean 67 65 3 Soo o o BR a o o o o e D o o E o a na nz pi ain Lus is isla mh lg la pay pi r dm bog ly 20 40 60 80 100 120 0 20 40 60 80 100 120 hits event hits event o T Figure 16 A hits per non zero event distribution for each the asynchronised clock and the synchronised clock case A closer look at LHE revealed the data shown in the figures 17 and 18 where the former displays a ROC map of only one non zero event and the latter shows the result of the whole rate measurement containing all hits As can be seen in the two figures the LHE are in the same clusters as usual non zero events In figure 18 the diff
7. Especially since the the local rate did not decrease compared to the mea surement done at a total rate of 353 MES Efficiency map o Entries 4160 3 80 Meanx 25 38 00 Meany 39 42 70 90 60 80 50 40 30 20 0 10 20 30 40 50 column Figure 32 An efficiency map at 300 MES The amount of readout Vos is plotted as a function of position 39 Figure 33 The analysis of the pixel efficiency map at 300 pixel efficiency of all pixels o E o EL x 5 Entries 4160 510 E Mean 95 34 E RMS 1412 10 10 tt Bi i L I L L I L 1 j i L j 0 20 40 60 80 of pixels as a function of their efficiency Mean Efficiency 26 Pixel Efficiency vs Rate 90 50 100 150 200 250 300 100 Efficiency MA The number 350 Rate MHz Figure 34 The mean efficiency as a function of the rate 40 Efficiency distribution 8 gL entries 30 5 E mean 94 776802 3 sigma 0 131535 o L O25 2 1 5 1 0 5 olo li hog 94 4 94 6 94 8 95 95 2 95 4 Efficiency Figure 35 A distribution of 30 efficiency measurements Each measurement was done for the rows 40 60 41 Efficiency and Local Rate per Double Column Efficiency versus local rate per double column F 5 100 x KI E m x x 9 E Pe E IG E H x
8. MHz cm 2 Local rate MHz cm 2 Efficiency and Local Rate per Double Column 100 MHz cm 2 T z 100f s SE x 120 S JE x x x 3 ees x x 4 Ea x x 4 a E x 53100 sok X xx E 80 98 5 x al E x og a zl El x 9 J sE o o 160 E x o o Ki 975 E a 40 E a o zs 9 o o J tE o Io Ex x o 4 ssp X i 9 1 L 1 LT 10 15 20 25 0 Efficiency and Local Rate per Double Column i 195 MI 2 EZ E 3 z 7 BessE x a S x al 3 E o x x x x J 5 se di x 200 985E x Kg x o zl E xc 1150 sE a pi o y J E o o 4 975 x o o J aE 0 A o o 9 7100 E 9 x oj es O x kaaa o 5 s 750 E a zl i i Q fi fi L Lal 10 15 20 25 0 Double Column Efficiency and Local Rate per Double Column 300 MHz cm 2 E 100 E 2 xxx TX amp x E xx x A X X x zn 5 o 3 oof 300 E j E o 3 sof 7250 E o J 7oE B oo A200 E e o o 2 Y mE 59 o 150 E q SE o s o 100 E o 3 WES 50 o E ob 9 1 tb E 1 ny pg t 10 15 20 25 0 Double Column Figure 43 The efficiency x and the local rate 0 cf equation 3 as a function of the double column The total rate is shown in the top right corner of each plot 55 Local rate MHz cm 2 Local rate MHz cm 2 Local rate MHz cm 2 Efficiency Efficiency versus local rate per double column 50 MHz cm 2 Efficiency versus local rate per double column 100 MHz cm 2 2 El Bun 99 5 99 98 9
9. event as discussed in the previous subsection was significantly influenced by LHEs and therefore impaired a comparison with estimates Since such comparisons were exceedingly important at this stage of the experiment the clocks were synchronised for all following test unless mentioned otherwise Map o 1 row 0 9 e FTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTT 0 8 6 o a 0 7 5 o 0 6 4 0 5 3 0 4 o 0 3 2 o 0 2 E o 0 1 L L L L L L L L L L L L L L L L L 0 10 20 30 40 50 column cr Figure 17 A ROC map displaying a single LHE The amount of hits is plotted as a function of position 23 z 80 45 Nd z A 70 40 60 35 r BF L 30 50 E 25 407 a LE 20 E 15 20 E m 10 E E Jas 5 0 E 1 L L L L L L i i i 1 L L L i 1 L i L 0 0 10 20 30 40 50 column Figure 18 A ROC map where the amount of hits is plotted as a function of position The red pixels indicate LHE while the green pixels indicate an expected hit corresponding to the peak in figure 16 event distr for 100000 triggers and rate 123 330000 Entries 3936 Mean 50 13 D Q o occurrence a o o 400 300 200 100 o TEME acil eae E aarda 0 30 40 50 60 70 80 hits event o o 2 Figure 19 A hits per non zero event distribution The difference to the data shown in figure 16 is the amount of pulses
10. is then entered in units corresponding to the pattern unit length which is 25 ns if the internal clock has been set to match the clock of the testboard To edit a pattern select EXECUTE ACTION SET DATA TO HIGH or SET DATA TO LOW Select the channel using the arrow buttons Then use the general purpose knob or the arrow buttons to select the bit and press the EXECUTE button to edit the bit To edit multiple bits at once in the same channel press the CURSOR button change the cursor size with the general purpose knob and press the CURSOR button again Editing the run mode To choose the run mode select the SETUP menu and press RUN MODE Now a variety of run modes can be selected where the most important ones 16 are the REPEAT and the SINGLE mode To send the patterns repeatedly select REPEAT If the trigger which the testboard sends to the ROC shall be synchronised with the pattern select SINGLE The SINGLE mode will send the pattern once each time the data generator receives a trigger corresponding to the synchronisation signal output by the testboard in the trigger input cf figure 11 Assigning channels to the pod Press the SETUP button and select the POD ASSIGN menu This opens a new window in which the general purpose knob can be used to select the internal channels e g DATA01 The arrow buttons are used to select the output channel of the pod e g A 01 Press the START STOP button to start or stop the outpu
11. schematic the WBC value determines which clock cycle is read out while the tct value determines the distance between the calibrate signal and the trigger signal In the following measurements the tct value instead of the WBC value was varied to scan through the region of the pulse The usage of the tct value or the WBC value is equivalent and the unit of both values is given in clock cycles The figures 23 24 and 25 show the amount of hits plotted as a function of the tct value for different pulse lengths As mentioned before cf section 3 the unit length of the pulse length is given in clock cycles as well In figure 23 it is clear to see that there are only hits in one clock cycle when using the smallest possible pulse length On the other hand figure 25 indicates clearly that the rising edge as well as the falling edge cause hits Keeping this in mind the contribution of the rising edge as well as the falling edge can also be identified in figure 24 However there are also hits in between the two edges The contribution of hits in between the edges was found to be largest at pulse lengths of about 5 7 clock cycles At a pulse length of about 25 clock cycles this contribution is very small yet further increasing the pulse length does not necessarily extinguish this contribu tion It can be avoided though by increasing the comparator threshold In the same way the whole contribution from the rising edge or the total contribution fr
12. shown in figure 29 The function adc2 defines the sequence and is not given by default It can though easily be rewritten using the adc function given in the file containing the testboard functions DTB py doubleC py performs the desynchronisation test as shown in e g figure 30 This test also makes use of the adc2 function eff cxx takes an efficiency map as input provided by either pixelalive py or pa py In the former case it produces three pixel efficiency plots One for all pixels a second one for all affected double columns and a third for all clusters In case of a map provided by the pa py program it produces the efficiency for all pixels only as for instance in figure 33 efferror C produces the plot shown in figure 35 The data is hardcoded effhistol C produces the plot shown in figure 34 The data is hardcoded functions py provides a couple of functions to use the pulse height cali bration data lhe py produces 5 ROC maps containing single LHEs as e g in figure 17 map py measures the rate and produces a ROC map a hits per non zero event distribution a non zero event per loop distribution and a pulse height distribution AT pa2 py performs an efficiency test and produces an efficiency map It arms each pixel separately instead of a whole row and was used for debug ging pa error py performs numerous efficiency tests It provided the data for the efferror C makro In the beginning a rate measu
13. when choosing cables for a rate measurement 31 Occurrence N gt o co 2 D 3 S El S e TTTTTTTTTTTTTTTTTTTTTTT 0 Rate Distribution ils Entries Mean RMS 600 120 6 2 065 S dl 114 116 ia ER BER ER 118 120 loy 122 sl s 124 CA 126 Lo 1 1 128 130 Rate MHz Figure 27 A distribution of 600 rate measurements all done with 100 loops and 1000 triggers each Occurrence 10 N B a 2 5 3 S S TTT TTTTTTTTTTTTTTTT o Rate Distribution Entries Mean RMS 500 177 7 7 315 L wol yg 200 210 Rate MHz Figure 28 A distribution of 500 rate measurements all done with 1 loop and 100 000 triggers each Table 1 The mean and RMS of rate distributions for different settings Measurements loops triggers per loop mean rate 4 te RMS e te 100 10 10 124 8 83 56 100 100 10 141 48 33 13 100 10 100 128 91 45 83 100 100 100 115 74 9 14 100 100 1000 136 47 4 07 32 4 6 Desynchronisation There were reports of an issue when reading out data Specifically it seemed that data belonging to one trigger were readout with the next or the previous one This shift which is called desynchronisation could also be investigated with the current setup To do so a new seq
14. 42 62 120 60 F 100 50 F 800 E BH 30 600 20 er 10 200 geese ls lass lass P 0 10 20 30 40 50 column Figure 14 The typical outcome of a rate measurement Only one cable is connected The amount of hits is plotted as a function of position 20 events per loop distribution o T TT I LT T T T TI T c Entries 100 Mean 17 24 occurrence oH para ar e t er da ix Tog ac elg o 0 5 10 15 20 25 30 35 40 events per loop Figure 15 A distribution of non zero events per loop 1000 triggers were sent in 100 loop 4 2 Clock Synchronisation At this stage every pulse of the length of one clock cycle was assumed to induce hits in only one clock cycle Calculations as in the example of the previous section are based on this assumption When the hits per event distributions were plotted two clear peaks were expected One at zero the other at about 80 hits since 8 clusters 10 hits per cluster 80 hits However measurements done with the clocks from the testboard and the data generator running asynchronously led to the detection of so called low hit events LHE Thereafter the clocks were always synchronised by putting the clock of the testboard into the data generator cf sention 3 3 To put the clock out of the testboard one of the programmable outputs labelled
15. 7 5 97 96 5 mani TTTTTTFETTTTTTTTTT Efficiency o 50 60 Local rate MHz cm 2 Efficiency versus local rate per double column 148 MHz cm 2 Efficiency 24 100 99 98 97 96 95 TIT TTTTTTTTTTTTTTLTTTTTTTTT Xx Efficiency 100 120 Local rate MHz cm 2 Efficiency versus local rate per double column 195 MHz cm 2 o ry 98 97 96 95 Nx 140 160 180 Local rate MHz cm 2 100 120 Efficiency versus local rate per double column 253 MHz cm 2 e US A a Te 40 60 80 100 120 140 160 180 200 220 240 Local rate MHz cm 2 Efficiency versus local rate per double column 300 MHz cm 2 Efficiency 26 100 99 987 97 96 95 FTT TTTT TTTT TTTTTTTTTTTTTT Efficiency j 3 60 40 20 To CEST US ETT x OK PO xxx x x 350 1 1 L 1 50 100 150 200 250 300 Local rate MHz cm 2 Efficiency versus local rate per double column 353 MHz cm 2 Efficiency 100 90 80 70 60 FFTT TTTTTTTTT TTTTTTTT Y 50 xx X x x Xx x x x x x L L kag 200 250 sa loy 1 350 L 1 1 1 50 Too 150 300 lisas boo 0 3 yy litis o IT 50 100 150 200 250 300 350 Local rate MHz cm 2 e 1 1 200 250 300 350 Local rate MHz cm 2 Figure 44 The efficiency as a function of the lo
16. Electrical High Rate Test Stephan Wiederkehr Department of Physics ETH Z rich July 29 2014 Contents 1 Introduction 1 1 1 2 1 3 1 4 LHC add CMOS an eee eae asian Pixel Detector x zoo ek RG eRe ee ae in Luminosity 4 2 scales Sh dom OUR ee BANA Electrical High Rate Test nn Testing Procedure Experimental Setup 3 1 3 2 3 3 3 4 Toa DOIE e 2 6 ra AT caw Mena aden E aded AI d e Go ne Da end Excitation Harness le DG2020A Data Generator 2 nn nn Rate Estimate and Measurement Results 4 1 4 2 4 3 4 4 4 5 4 6 4 7 Early Results Proof of Concept 1 222 nn Ene Clock Synchronisation lll Pu lse Height cis mE o al ed a ee S Pulse Length ice ra yo eec pmo ees pom Gru te pU Rate Measurements eA Desynchronisation ee eee Chip Efficiency a Conclusion Possible Further Studies Acknowledgments Appendix 43 44 45 47 1 Introduction 1 1 LHC and CMS The Large Hadron Collider LHC is the latest and most powerful accel erator added to the European Organization for Nuclear Research CERN Conseil Europen pour la Recherche Nuclaire It consists of a 27 km long ring in which 2 beam pipes carry the particles The particles are forced into a circular path by a magnetic field provided by superconducting electromag nets There are several accelerating structures along the ring to boost the partic
17. Entries 4160 xt o Meanx 25 5 El x Meany 39 37 2 Entries 4160 510 Mean 98 95 3 E RMS 2 106 10 F L f 10 iku kagat stat sibat AAA 20 30 0 20 40 60 80 100 column Efficiency Pixel Efficiency of all pixels 100 MHz Pixel Efficiency of all pixels 148 MHz a E a E l x E us m Entries 4160 a Entries 4160 5 ok Mean 98 72 5 Oe Mean 98 05 zc RMS 2 907 Z E RMS 4 422 10 il 102 ib F E lu n ih tok l ok d tt tr AEN LI SENE nt 0 20 40 60 80 100 0 20 40 60 80 100 Efficiency Efficiency Yo Pixel Efficiency of all pixels 195 MHz Pixel Efficiency of all pixels 253 MHz 2 pF 2 F o H o F X x x a p Entries 4160 2 y Entries 4160 Sol Mean 98 13 Sip Mean 97 65 z E RMS 4745 2 E RMS 6 11 105 E 10 i d F Al d 10 n 10 E ij E Eua a d oa rcg go lll Maga rox d os Es aia ll ea lg 0 20 40 60 80 100 0 20 40 60 100 Efficiency Efficiency Figure 41 Efficiency map and plots where the pixel efficiency is shown as a func tion of position and the amount of pixels is shown as a function of their efficiency respectively The rate is shown in the top right or top left corner of each plot or map respectively 53 Pixel Efficiency of all pixels 300 MHz Pixel Efficiency of all pixels 353 MHz a p a F o o H M x L x 5 Entries 4160 5 Entries 4160 103 Mean 95 34 210 Mean 95 41 ae RMS 1412 E RMS 11 95 107 10 10 tok tt 1
18. H E 1 L il I i L L Ii L 1 L 1 E 1 i L 1 1 0 20 40 60 80 100 0 80 100 Efficiency 90 Efficiency 90 Figure 42 Pixel efficiency plots where the amount of pixels is shown as a function of their efficiency The rate is shown in the top right corner of each plot 54 Efficiency Efficiency 96 Efficiency Efficiency 96 Efficiency and Local Rate per Double Column 50 MHz cm 2 100 E x x o x T X x ssh x x x o E M Ea E x j 99 x X x 3 F e o I 98 5 y o J F x o o Y E x oo E x 5 o o Ho o o o zl 9s o E 9 d 97 i 2 1 l 1 ee 10 15 20 25 Efficiency and Local Rate per Double Column 148 MHz cm 2 E a xS 99 ie J x x 9 x p 98 x o x x x o o 9 o pa x x o a pa ox e x o o x o o o o t In L 1 L 10 15 20 Double Column Efficiency and Local Rate per Double Column m 253 MHz cm 2 m x es o x x s x x E a x o F xy 8 x x x E x o ii o o o 9 2 o x YA r o o o o 96 d x o o po x 3 pa 95 pr x o E go 1 pA CNET pi 10 15 20 25 Double Column Efficiency and Local Rate per Double Column 353 MHz cm 2 100 Xd EEES TTE x x x sok pa o L o L o so o E o C o o 7o o o po o o r o 60 o L 9 E 4 9 1 1 1 13 5 10 15 20 25 Double Column 180 Local rate MHz cm 2 Local rate MHz cm 2 Local rate
19. L T 350 C 9 N 1 L L 1 L 1 x SI o so 100 150 200 250 300 350 x so lt Ap Dos ls d acacaca a a a APP NA 10 15 20 25 0 50 100 150 200 250 300 350 Double Column Local rate MHz cm 2 Figure 36 Left The efficiency x and the local rate 0 cf eq 3 asa function of the double column Right The efficiency as a function of the local rate The plots contain the data starting on top from the efficiency maps for 253 MHz 300 MHz and 353 VHZ cm cm cm 42 5 Conclusion The electrical high rate test produces indeed clusters of pixel hits at a tunable rate Moreover the rate agrees well with estimates which allow to predict the measured rate and program the data generator accordingly However high rates can only be measured with a specific choice of cables The accuracy of rate measurements depends on the amount of statistics as expected Yet the required amount of statistics depends on the random ness of the readout Using a total of 100 000 triggers yielded an error of about 2 MES The total amount of triggers was split using a for loop In future experiments a more sophisticated method to realise a random readout might help to decrease the necessary amount of triggers Synchronising the internal clocks of the data generator and the testboard proved to be important to control the amount of non zero events and the amount of hits within each With the clocks being synchronised the pulse length does
20. Mean 106 6 H RMS 1 631 40 30k 20 10 E a A A A 1 104 106 108 110 11 4 Figure 24 The amount of hits as a function of tct The unit length of a pulse is 25 ns hit distribution pulselength 25 a Entries 100 Mean 118 2 RMS 10 83 60 50 40 30 20 Figure 25 The amount of hits as a function of tct The unit length of a pulse is 25 ns 29 Figure 26 A schematic of the signal synchronisation Additionally the post flank oscillations are drawn 30 4 5 Rate Measurements As the hit rate is generated by the pulse pattern which was sent repeat edly during rate measurements the exact rate is given by reading out all clock cycles of the pulse pattern once and averaging the hits contained in them For a random readout this means that every clock cycle of the pulse pattern should be read out equally often Hence the amount of statistics determines the precision of the rate measurement It was found that at least 100 000 triggers have to be sent to achieve a reasonably small error Moreover the influence of the loops was found to be significant Figure 27 shows the distribution of 600 rate measurements all done with 100 loops and 1000 triggers each Figure 28 shows a simi lar distribution but the measurements were done with only one loop and 100 000 triggers in it The total amount of triggers in the measurements of both distributions is therefore the same H
21. S gh o x x x x T S eb x x xx 5 E x i 23 5 E xX a C i x o 4 FE x 5 go Xx 98H x x x Too E x E x o x H x C o 9 o n o bi x ka 9 x ETA x 415 E E o a C o 7 96 x s x T kai E x a 95 x x o o x o o o x L 50 E Me Eg j ga x x 7 a AAA AAA AAA PP PP AAA 10 15 20 25 0 50 100 150 200 250 300 Double Column Local rate MHz cm 2 Efficiency and Local Rate per Double Column Efficiency versus local rate per double column 10 xod dk EEE xx xxx 15 T100 x X X xx xx BOB x x g Wu eS KA x x gt E mO Xx RX xy x x E E 5 Asg 5 100 S w 8 E S a xx E o J x so 25 L x x j n x E o 88 x x 7o ka dod 20 soj E o o o o C E E 97 x so o Y H x E o 315 a dj e 96 x SG o o 9 10 u o 4 95 4o a 20 50 M gt 7 m 1 1 E o x I Ha Ng 30 700 186 200 256 300 350 ee d 3178 oso so Dis A E EE E S E O E 10 15 20 25 0 50 100 150 200 250 300 350 Double Column Local rate MHz cm 2 Efficiency and Local Rate per Double Column Efficiency versus local rate per double column E 100 7 3 kg 0 x X X X x3 X x 100 pag o u og ee o xd ue x X x g5 x x Xx x 2 E x x E 2 E 3e x x IK x x x x xj 3 Eo x 3 x E S EE gt Y E 90 E o o 480 u F x ox L o E E z x E 25 E os x x L p o pi so x 80 3 L os E 720 L sw po i id z gt m 70f s E o o 718 x 70 o E 94 i E o i Lo o o 10 o x o y E 60 o L s
22. an keep its level of performance the luminosity will be further increased along with the centre of mass energy which is planned to reach 14 TeV As a next step the spacing between bunch crossings will be reduced to 25 ns after the first long shutdown LS1 The development of the luminosity in the past and the predictions for the future can be seen in figure 2 With an increasing luminosity even going beyond the initial design luminosity also the pixel detector is confronted with an increasing rate of particle collisions To cope with the increased requirements a new chip was designed and is presently under development It will be installed in the layers 2 3 and 4 Therefore in addition to the desire to reproduce hits in a common labo ratory similar to those produced in the LHC also the rate of those hits experiences an increasing importance Already existing methods to gener ate high hit rates are the beam test and the X ray setup The former can emulate the situation encountered in the LHC very well but it is a large installation itself where the latter can be used in a common laboratory yet it does only induce hits in single pixels instead of inducing hits in clusters as it happens in the LHC The electrical high rate test was developed as a small setup suitable for laboratories which can induce clusters of pixel hits It is the latest method to induce pixel hits at a high rate for testing purposes 1 Peak luminosity Integ
23. ate 17 non zero events The delay consisted of a fixed part and a random contribution The fix delay was set to 10 us to provide enough time for the chip to read the data out The random part varied between 0 and 50 us The sequence can be sent in two ways either via the Pg_Single command or via the Pg_Loop command The Pg_Loop command sends a defined sequence and a defined delay repeat edly The user can define the start the time during the command shall run and the stop The amount of sent triggers can be calculated using the time of one loop sequence delay and the duration of the command Since the calculation may differ from the actual amount of triggers sent the Pg_Single command was used The Pg Single command sends the afore defined sequence one Hence to send e g 1000 triggers the Pg_Single command must be executed 1000 times It was found that sending those single shots led to an additional not constant delay Therefore the total delay amounted to about 275 us on average corre sponding to a trigger rate of roughly 3 6 kHz The triggers were sent to the testboard using a single command However in the test program the total amount of triggers was split up using a for loop as shown in the pseudo code below In the following the amount of loops inside such a for loop will be referred to just as loops for loops send x sequences each sent as Pg_Single The random delay and the loops were applied to random
24. c drawing of the basic principle Isolated wires pointing on Kapton foil lying on the ROC The Kapton foil acts as a dielectric of a capaci tor ROC inr Channels Cables Wires zb qp 3 E Voltage gas Figure 4 An schematic illustration of the setup featuring the data genera tor the cables the wires and the ROC A more detailed description of the setup can be found in section 3 2 Testing Procedure The first prototype ready for enhanced measurements of the excitation harness cf section 3 which is the pivotal component of the eletrical high rate test was built at the Paul Scherrer Institute PSI before it was brought to the ETH to test it The goal of this work was to obtain a testing routine to do experiments which require a certain hit rate Therefore the first step was to reproduce the basic results of the setup first and foremost the ability to induce hits in a ROC As a second step mechanisms which influence the rate were identified and investigated Also the precise procedure and setup for a rate measurement was a priori not clear and was developed Rate estimates and the deviation from the measurements were analysed as it is important to provide a desired rate for more advanced measurements Finally the chip efficiency was being tested The efficiency test analyses the output of the ROC with respect to the input which is given by the provided rate and determines the efficiency of the read
25. cal rate cf equation 3 for each double column The total rate is shown in the top right corner of each plot 56
26. d reading it out again while the chip had to deal with the hits induced by the data generator Therefore the efficiency of a pixel corresponds to the readout Vos divided by the total amount of Veus sent The latter is equivalent to the total number of triggers sent since the sequence sent to the chip was the same as for a rate measurement cf 3 4 Therefore the efficiency corresponds to the following formula where Noa is the number of readout Vos and Nrg is the number of triggers NG al pixel ef ficiency Ning All pixels of the ROC could not be armed at once due to memory limitations but it was armed row by row The full procedure of a test was the following e A single reset signal was sent e A whole row of pixels was armed and another single reset signal was sent e The sequences were sent to the ROC A total of 100 000 triggers were sent in 100 loops and 1000 triggers each e The readout Vo s were plotted in a ROC map as a function of position e The whole ROC was unarmed and the procedure started anew After the last row the test ended According to the procedure the outcome of such a test was a ROC map where the pixel efficiency was plotted as a function of position Figure 32 shows a so called efficiency map at a hit rate of 300 MES As one can see by the clusters of lesser efficiency these tests were performed using five ca bles to distribute the clusters over the whole ROC The analysis is shown in figure
27. erence between a LHE and an expected non zero event can clearly be distinguished by the amount of hits and there fore the colour in which they are plotted The green pixels correspond to an expected hit belonging to the peak in the distribution on the left of figure 16 The red pixels saw additional hits coming from LHE of the mentioned distribution It is unclear why the LHE manifest themselves only in some pixels instead of being equally distributed over all clusters The contribution of LHE to the rate was roughly up to 40 Yo where the percentage did depend on the cable To emphasise this figure 19 shows a measurement for another cable The pulse height of LHE did appear to be randomly scattered within the expected pulse height distribution as a change in the comparator threshold 22 did not extinguish the LHE Neither did the relative contribution to the rate change significantly It was not investigated whether the pulse height of a particular pixel changes when the pixel is part of an LHE w r t when it is part of an usual hit It is being assumed that running the clock asynchronously splits some pulses into two clock cycles where each contains a fraction of the total amount of hits This would explain the increased number of non zero events as well as the occasionally reduced amount of hits within such an event The LHE were not investigated more intensely and remain not fully understood though The probability of a non zero
28. es 100 E Mean 118 2 EE Mean 124 4 WF RMS 10 83 To RMS 14 26 60f 60 50 50 40 40 30 30 20 20 105 10 aaa Da apat ang Lpa a ig ha ar gabi rna tia EL fra boo 105 110 115 120 125 130 135 140 105 110 115 120 125 130 135 140 145 150 tot tot Figure 39 The amount of hits as a function of tct The unit length of a pulse is 25 ns 51 50 MHz Efficiency map o z 80 Entries 4160 tf 2 Meanx 25 61 o Meany 39 45 100 MHz Efficiency map Entries 4160 510 Meanx 25 62 Meany 39 44 10 20 30 10 20 30 148 MHz Efficiency map o 195 MHz Efficiency map Entries 4160 Y Entries 4160 N Meanx 25 65 jt Bo Meanx 25 63 dl Meany 39 41 Mean y 39 4 80 row 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 o 0 30 40 50 0 10 20 30 40 50 column column 253 MHz Efficiency map o 300 MHz Efficiency map o 80 Entries 4160 Y 80 Entries 4160 N Meanx 256010 E Meanx 2538 1100 Meany 39 41 Meany 3942 10 20 row 70 70 60 60 50 50 40 40 30 30 20 20 Figure 40 Efficiency maps where the pixel efficiency is shown as a function of position The rate is shown in the top left corner of each map 52 Efficiency map o Pixel Efficiency of all pixels 50 MHz x PO baa MHz
29. estboard 2 The single ROC used in this project was wire bonded to a so called chip board Furthermore the chip board was connected to the testboard with an adapter Figure 9 shows the described setup fully connected On the adapter several labelled dot contacts can be identified With an oscillo scope they can be used to measure the signals corresponding to the labels 11 Figure 9 A picture of a single ROC on a chip board An adapter is con necting the chip board to the testboard The small installation on the right hand side of the chip board is a basic version of the excitation harness and was not installed for the project presented here 3 2 Excitation Harness The excitation harness is the central part of the setup guiding the elec tric pulses to the ROC and inducing signals in it Following the previous description it consists of wires grouped in cables which are gathered in a block of plastic to focus the wires on the ROC The Kapton foil is fixed to the bottom of the block Figure 10 shows a picture of the excitation harness already attached to a ROC When attaching the excitation harness to a ROC the block is screwed on the chip board The chip board is a small printed circuit board PCB to which the ROC is wire bonded The two drilled holes in the chip board define the position of the block with respect to the ROC A few splitters of the ROC waver were placed at the sides of the ROC before screwing the block on top
30. gt a RE Entries 100 E Entries 100 80 Mean 105 5 60 Mean 105 8 E RMS 0 1758 F RMS 0 4539 70 50 60 F E 40 50 F 40 8o 30g 20 20 E E 10k 10E B A ANA ls 1 DT Lara Lira lin oC ara La ca Lada La lu ls bara la Va 100 102 104 106 108 110 112 114 100 102 104 106 108 110 112 hit distribution pulselength 5 tot hit distribution pulselength 6 tet 2 F 2 F zc a 2c a y Entries 100 50 Entries 100 50 Mean 1062 E Mean 106 5 F RMS 0 9227 L RMS 1 307 E 40 40 E C 30 30 20 20 10 10 el ala Ea La anri ai ad el lalala alas lin 100 102 104 106 108 110 112 102 104 106 108 110 112 114 tet tet Figure 37 The amount of hits as a function of tct The unit length of a pulse is 25 ns 49 hit distribution pulselength 7 eo oe E L a e 5O Entries 100 L Mean 106 6 E RMS 1 631 40 30 20 10 E uu ene paca Pado gag lolas leah 102 104 106 108 110 112 114 tet hit distribution pulselength 9 2 F EF qa 50 Entries 100 E Mean 107 5 L RMS 2 876 40 30 20 10 pe Lala aaa o ioo o rr bah lpg 104 106 108 110 112 114 116 tet hit distribution pulselength 11 2 Teal a ka sor Entries 100 E Mean 108 3 E RMS 374 40 30 20k 10 a a tide tid 102 104 106 108 110 112 114 116 118 tct hit distribution pulselength 8
31. he information used in the introduction was partially taken from the CMS TECHNICAL DESIGN REPORT FOR THE PIXEL DETEC TOR UPGRADE ISBN 978 92 9083 380 2 D Kotlinski R Baur K Gabathuler R Horisberger R Schnyder and W Erdmann Readout of the CMS Pixel Detector http ph collectif lecc workshops web cern ch ph collectif lecc workshops LEB00 Book tracker kotlinski pdf online 20 08 2012 The original picture was taken from a presentation of Jorgen Beck Hansen Physics at run 2 The original pictures are taken from User Manual DG2020A P3410 amp P3420 Data Generator and Pods Sony Tektronix Corporation P O Box 5209 Tokyo Int l Tokyo 100 31 Japan Tektronix Inc P O Box 1000 Wilsonville OR 97070 1000 46 8 Appendix Figure 37 38 and 39 show all pulse length scans cf section 4 4 Since the data generator did run with the clock of the testboard the unit length of the pulse is given by 25 ns All efficiency maps and their analyses are summarised in figure 40 41 and 42 The error in the rate is 2 MER according to figure 27 All important programs and root makros which where used can be found on the provided CD ROM A brief description is given in the following dceffl C Takes an efficiency map pa py as input and outputs the effi ciency and local rate per double column as well as the efficiency as a function of the rate for each double column double py performs the desynchronisation test as
32. igure 1 During the winter break in 2016 2017 the pixel detector will be replaced with an 4 layer barrel 3 disk endcap system The fourth layer added at a radius of 16 cm will increase the tracking precision and also serve as backup in case the first layer of the strip detector degrades faster than expected The layers are positioned at the radii 30 mm 68 mm 109 mm and 160 mm while the length of the barrel extends to 548 8 mm The support tube carrying the pixel modules will also be redesigned to avoid a large increase in material due to the added layer and endcap discs The detector is built in a modular concept A module consists normally of 16 readout chips ROCs A single ROC has 80 x 52 pixel channels each of the dimension of 100 um x 150 um yielding a total of 78 8 million channels for the whole barrel part of the detector To ease the installation of the current detector around the beam pipe it is built in two halves motivating the construction of half modules Figure 1 Layout of the pixel detector before the upgrade The three barrel layers drawn in black and the four forward discs in purple 2 1 3 Luminosity The LHC was designed to operate at a luminosity of 1 10 ri while the particle bunches where set to collide every 25 ns However when it started in 2009 the spacing between bunch crossings was 50 ns long Increas ing the centre of mass energy led to a peak luminosity of about 7 1033 E in 2012 If the LHC c
33. ise the read out Ideally the hits would occur randomly However since they have to be programmed and therefore are not random the readout was randomised instead The program measuring the rate cf the appendix essentially executed the for loop shown above collected the data and calculated the rate The data contained the amount of events corresponding to the number of triggers sent Each event comprised the amount of hits including the information about 18 the their pulse height not calibrated and their pixel address The rate was calculated by dividing the number of hits and by the total number of triggers loops triggers per loop according to equation 2 where Nnits is the amount of hits Nirg is the amount of triggers 7 25 ns is the clock cycle length Aroc is the area of the ROC and C 40 converts the rate from MEZ into cm cm MHz Nnits 1 C rate 2 m Nego T Aroc 2 4 Results 4 1 Early Results Proof of Concept The result of an early rate measurement is shown in figure 14 The figure shows a map of the ROC where the amount of hits is plotted as a function of position Indeed each wire induces a cluster of roughly 10 pixel hits However only eight clusters can be seen even though ten clusters were expected after connecting one cable Unfortunately the observation of less than ten wires was the generic case The reasons were found to be one of the following e The wires did not poi
34. le energy The beams travel opposite to each other and cross in spe cific areas where detectors are stationed to observe the collision of particles One of the four experiments located at the LHC the compact muon solenoid CMS has been designed to search for the Higgs boson and for signs of Supersymmetry alongside the ATLAS experiment To meet all the requirements of such a task the CMS detector consists of the following elements listing from centre e a tracking system including pixel detector silicon strip detector e calorimeters measuring the particles energy electromagnetic calorimeter ECAL hadronic calorimeter HCAL e solenoid e muon chamber detectors The solenoid which is 7 meters in diameter and 13 meters long provides the experiment with a magnetic field of 3 8 Tesla Since the silicon strip detector cannot cope with the design luminosity close to the beam the innermost detector is made of silicon pixels which can provide three dimensional information of each charged particle 1 2 Pixel Detector The CMS pixel detector is designed to measure precise hit information withstand radiation damage for multiple years of operation and using as little material as possible to minimise multiple scattering of the particles The barrel part consisting of three layers and two forward discs are arranged such as to enclose the bunch crossings as good as possible The described layout of the detector is shown in f
35. not matter when it comes to rate calculation since the contri butions of the edges and between are very reproducable and can be included in the estimate However for most measurements the pulse length of one clock cycle is the easiest to handle as there are only hits in one clock cycle per pulse Furthermore the setup was successfully used to test the chip efficiency The mean efficiency per pixel was found to be over 98 96 for rates below 150 M Br Comparing the result shown in figure 34 with data obtained by X ray experiments the efficiency is already relatively low even below 150 MF The reason for this difference is attributed to the different amount of hits per non zero event While in this experiment there are about 80 hits in a non zero event there are only about 3 hits per non zero event using X rays at the same rate Similar is the situation for the affected double columns In this experiment an affected double column contains usually about 5 hits while there is rarely more than one hit per double column using an X ray setup Therefore in an affected double column the rate can easily reach 100 MHz cm or more To increase the comparability the amount of wires per cable could be re duced to decrease the amount of hits caused by one pulse 43 Looking at the rates in the double columns separately the expected decrease of the efficiency with increasing rate is merely hinted It is unclear why there is a fluctuati
36. nt on the ROC e The wires were not close enough to the Kapton foil let alone touching 1t This is discussed further in the section 6 e The wires were broken or badly soldered Figure 15 shows a distribution of non zero events per loop where an event describes a readout clock cycle As the readout process is assumed to be random a Poisson distribution was expected Furthermore since the depicted mean corresponds well to the maximum of the distribution the mean is equivalent to the probability to encounter a non zero event in a loop The probability can be read off dividing the mean by the number of triggers per loop in this case 1000 This probability can be compared to the probability given by the data generator The latter is simply the amount of pulses divided by the pattern length assuming a pulse induces hits in only one clock cycle As an example The measurement shown in figure 15 was done with 1000 triggers per loop 19 which corresponds to a probability of 1 724 The measurement was done with only one cable The pulse pattern was 58 clock cycles long and contained one pulse of one clock cycle Therefore the expected probability is equal to x ez 1 724 Generally a good agreement between those probabilities was found Hence the clock cycles were roughly readout equally often confirming the assumption of a random readout Map 80r 2 a 140 E a Entries 97606 707 L Meanx 20 32 E Meany
37. od in which it was found in the odd events It is not clear whether the Vc shifted twice in the same direction or if it shifted back and therefore restored the original synchronisation In summary the desynchronisation was not found to depend on the hit rate Only the blocksize and buffsize values were found to influence the ap pearance of the effect The desynchronisation was not further investigated because it did not appear with the blocksize and buffsize values mentioned above 34 good good p z 80 TA 3 80 t SIE m 7 leon E 500 F 1800 sol u a Ps se a E E 700 H a 400 H a o m En BE SE nes soma eno E Meanx 2319 E Meanx 23 19 0 Meany saes go 40 E Meany 347 i500 E ma lao0 200 E 300 20 mpa 20 200 E 8 100 E E to 1o Fat o 1 1 1 lo oF i 1 1 lo b 10 20 30 40 50 y 10 20 30 40 50 column column previous previous 80 5 80 b BUE Enes 0 B L Entries 3 Meanx 0 E Mean x m Meany 0 70 Mean y amp 0 so sE 50 40 40 a 30 30 20 20 105 107 oF 1 1 1 1 ot 1 f 1 b 10 20 30 40 50 b 10 20 30 a0 column column error error c c 80 80 zF Es o EE Ens 0 E Meanx 0 E Meanx 0 m Meany 0 7 Meany 0 60 605 50 50 0 P 30 30 20 20 107 107 oF 1 NG 1 1 b 10 20 30 40 50 y 10 20 30 40 50 column column after first error after first error 80 80 d Y Envios 1
38. of it to protect the ROC from pressure The actual effect of those splitters was not investigated and is not known As one can see in figure 10 the cables are all connected to each other This is to connect all of the cables to ground which is done by the black cable shown in the figure 12 z u a Figure 10 A picture of the excitation harness attached to a ROC The black cable is the connection to ground 33 DG2020A Data Generator In this section a more detailed description the the data generator will be given along with a few examples of how to use it The goal is to help using the presented setup not to substitute the manual It is highly recommended to read the latter The data generator in question was the model DG2020A from Tektronix Sony where the pod model was P3410 The figures 11 and 12 show schematics of the front and rear panel of the data generator respectively The figure 13 shows a picture of its display while in the EDIT menu 13 GENERAL PURPOSE KNOB BE DG2020A DATA GENERATOR Bottom and side bezel buttons The main use of the bottom buttons is to call up sub menus and the side buttons are used to execute more detailed operations within the sub menus ON STBY Button CLEAR MENU Button EVENT OUTPUT O 3 ro mO
39. om both edges can be erased All performed pulse length 27 measurements are shown in the appendix In the schematic in figure 26 also the pulse as it is assumed to behave inside the ROC is drawn along with the comparator threshold voltage The chip is designed such that it reacts only on falling edges Therefore whenever a falling edge crosses the level of VinrComp a hit is induced in a pixel Thus in theory only the falling edge of the pulse was expected to induce hits The contribution of the rising edge as well as the contributions in between the edges are attributed to the post flank oscillations It is however not understood how these oscillations cause what is depicted in figure 24 and 25 For instance in figure 25 and considering the schematic it is unclear how in the contribution of the rising edge pulse even the second falling edge post flank osc seems to induce hits while in the contribution of the falling edge pulse only the first falling edge post flank osc induces hits hit distribution pulselength 1 eo 70 a 55 Entries 100 Mean 104 5 B RMS 0 5 e 4i e 3 e 2 e E o cs al hara Dope Er nn 06 98 100 102 104 106 108 110 112 ict Figure 23 The amount of hits as a function of tct The unit length of a pulse is 25 ns 28 hit distribution pulselength 7 a E 2 a 50 Entries 100 E
40. on of the efficiency even at relatively low rates of up to roughly 4 6 Possible Further Studies An issue encountered with the actual setup was the mechanical sensi tivity of the excitation harness Connecting and disconnecting cables can require a strength which is enough to pull a wire from the ROC As all the wires are tightly compressed when focusing them on the ROC the wire will not return once it has been pulled away This could be avoided by using e g a PCB as interconnection between the excitation harness and the data generator The PCB could be part of a fixed structure such that the user does hardly touch the excitation harness once the ROC has been attached Generally the control over the cluster position should be increased At the same time one could decrease the amount of wires per cable Ideally the wires should be connected separately such that the distinction of cables and wires becomes redundant As a suggestion one could place the wires on top of the Kapton foil in a well defined pattern Since the patter is known this would enable the determination of the cluster position 44 7 Acknowledgments I would like to thank Prof Dr Rainer Wallny for making this work possible and letting me be part of the very enjoyable and helpful atmo sphere Furthermore I would like to thank Dr Andrey Starodumov and Dr Wolfram Erdmann for their assistance and support during the project 45 References 1 T
41. out process Such efficiency tests were done before with other methods e g with X rays Therefore the efficiency test was used as a first example making use of a previously defined rate and it was a good opportunity to compare the results with already established data 3 Experimental Setup A picture of the whole setup is shown in figure 5 Also a schematic is shown in figure 6 where the connections to the particular power supplies are not drawn In the real setup a so called pinboard appears Its task is to connect cables coming from the pod to the cables of the excitation harness Hence the pinboard defines which channel being output by the pod is connected to which cable of the excitation harness A picture taken of the pinboard is shown in figure 7 Another difference between the picture of the real setup and the schematic is the connection between the testboard and the clock input of the data generator The connection was added after taking the picture and enables the synchronisation of the internal clocks However the connection alone does not suffice to synchronise the internal clocks A description of how the synchronisation can be achieved is given in the section 3 3 Independently of the synchronisation unless mentioned otherwise the clock of the data generator was running with 40 MHz as was the clock of the testboard Therefore the data generator could be programmed in units of 25 ns Unless mentioned otherwise the minimal
42. owed the measurement of a higher rate On the other hand a further drop in the efficiency of the strongly affected double column shown in figure 32 was clearly avoided as the additional hits were all in other dou ble columns Voltage Time Figure 36 shows the efficiency and the local rate as a function of the double column in the left column and the efficiency as a function of the local rate in the right column The plots contain the data for the rates 253 IL 300 i and 353 i as these were the least expected results The remaining data can be found in the appendix The local rate was calculated as in equation 3 were Np4 5 is the number of hits N ry is the number of triggers 7 is the clock cycle length Apc is the area of the double column and C converts the rate to MHz MH Nhi C local rate cm 3 Ning T Apo 38 The efficiency was expected to decline steadily as a function of the lo cal rate Yet the difference in efficiency between the double columns at a relatively low local rate is too large to see any tendency The only clearly visible decline was observed if the local rate exceeds 16 ME per double cm column at a total rate of 300 MES There the efficiency dropped below 40 96 Including the data for 353 AIO the reason for this jump becomes unclear Contrariwise it is not understood why at 353 AES there are two double columns which have an efficiency over 90 96 at a local rate of over 16 MES
43. owever the root mean square RMS shown in the inset box in each distribution indicates that the loops do contribute significantly to the randomness of the readout For compari son table 1 shows the mean and the RMS for different settings Using measurements such as the hits per non zero event distribution to get more precise information than merely vague estimates the formula in equation 1 yielded values very close to the measured rate The formula proved to be the most accurate way to program the data generator to pro vide a certain hit rate Due to the finite width of the mentioned hits per non zero event distribution and the error in the measured rate a fine tuning of the data generator was necessary in most cases It should be noted that the rate can only be measured correctly with a pixel efficiency close to 100 Increasing the rate will lead to a strong decrease in the pixel efficiency which will finally reach 0 such that the measurable rate has an upper limit This limit could be overcome by induc ing hits in double columns which were previously not hit by any clusters This means that the upper limit depends on the choice of cables Hence this limitation is attributed to an overflow in the double column data buffer Since each cable induces up to ten clusters which might all lie in the same double column this limit can occur at relatively low rates e g at about 100 u E Therefore this limitation should always be considered
44. pectively Figure 29 shows the outcome of two measurements In the measurement on the left the Vc simply vanished while in the measurement on the right the Vocal was shifted to the next odd trigger The error was typically found after 300 000 50 000 triggers and did not depend on the rate as it also occurred with the data generator turned off As a first step the data flow on the testboard was changed This was done by increasing the parameter buffsize from its default value 2097152 to 5 000 000 and additionally increasing the parameter blocksize successively starting at its default value 8192 As a result increasing those parameters did also increase the trigger num 33 ber at which the first error occurred At the value blocksize 1 000 000 no error was found during ten measurements each sending 2 000 000 triggers The figures 30 and 31 show essentially the same test though it was mod ified not to stop at the first error but to plot all the events in a certain range around it The hits caused by the electric pulses are clearly visible as those events contain more than the one hit due to the Vc The entries plotted in green correspond to events which passed the error test while the entries plotted in read produced errors according to the definition above To see errors at all the blocksize was set to its default value again Figure 31 is particularly interesting because the VCa returned to the even events after a peri
45. pulse length of one clock cycle 25 ns was used to induce hits The testboard was in the end running with the firmware 2 18 Primarily USB problems which seemed to be linked to the operating system Ubuntu led to installing this version of the firmware Figure 5 A picture of the whole setup taken before the first measurements The connection for the synchronisation of the clock cf figure 6 was added later clock input Data generator O Figure 7 A picture of the pinboard which connects the pod output cables to the cables of the excitation harness The blue jumpers show possible con nections where the black jumpers suppress a crosstalk between the cables 10 3 1 Testboard The purpose of the testboard is to emulate all commands which can be addressed to a single ROC or even a whole module The acquired data can be transferred to a computer via a USB connection A picture of the testboard taken of its front is shown in figure 8 There are four LEMO connections on the front of the testboard where two are fixed connections to put in a clock or a trigger and thus are labelled CLK and TRG The other two connections labelled D1 and D2 can be programmed Both connections were used in the presented setup One to put out the clock the other to put out the synchronisation signal gt gt 1 cK re 2 Ww HU et o D d oc a Po pier 9 Figure 8 A picture of the front of the t
46. rated luminosity 1000 0 300 fb 100 0 2 50E 34 2 00E 34 1 50E 34 LS1 LS2 LS3 1 00E 34 Luminosity cm s Integrated luminosity fb 5 00E 33 0 00E 00 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 Year ending Figure 2 The evolution of the luminosity including the data until and the prediction for the time after LS1 3 1 4 Electrical High Rate Test In the electrical high rate test hits are induced using electric pulses These pulses are guided via isolated wires to a Kapton foil which lies on top of the ROC The Kapton foil acts as a dielectric of a capacitor such that charge induced on top of it is also induced in the pixel unit cell PUC within the ROC A simple illustration is shown in figure 3 The pulses are generated by a data generator which allows to define pulse patterns in a tunable pattern length The used data generator was able to program different patterns into a number of channels where the pattern length was the same for all channels Using a pod each channel could be connected to a bunch of wires which henceforth will be denoted as cable Each cable consists of ten wires which were pointing on the previously mentioned Kapton foil Figure 4 illustrates the explanation above All the cables including the structure to focus the wire on the ROC are denoted as the excitation harness Wires Kapton foil Figure 3 A schemati
47. rement is included to avoid a reinitialisation of the ROC pa py performs an efficiency test and produces an efficiency map It per forms a rate measurement in the beginning to avoid a reinitialisation of the ROC param py contains the testing parameters used in map py pixelalive py pa py pa2 py lhe py and rate vs thr py pixelalive py performs an efficiency test and produces an efficiency map rate to histo py takes a text file containing the output of map py and produces a rate distribution rate vs thr py performs rate measurements while sweeping through an in terval of Vinrcomp It produces a plot as show in figure 22 scan tct py performs a pulse length scan as shown in figure 23 The syn chronising signal must be added to the sequence 48 hit distribution pulselength 1 hit distribution pulselength 2 ar a EY a E50 a BRE Entries 100 PESCE Entries 100 m Mean 1045 E Mean 104 9 60 RMS 0 E RMS 0 4933 40 50 L Ls 40 80 30 L E 20 20 L F 10 10 Exc pea Pana Ne Tues ep tp pcs T3 glee sc Dona f f EAS O O et Qc 9g 100 102 104 106 108 110 112 100 102 104 106 108 110 RC NER tet TM na tet hit distribution pulselength 3 hit distribution pulselength 4 B E o E 90 a
48. t be switched on cf figure 12 To output channels a pod needs to be connected to one of the three pattern data output connectors The clock synchronisation can be achieved by con necting the testboard to the clock in connection Note that the testboard supports LEMO connections while the clock in supports SMB connections Once a pod is connected the channels of the data generator must be as signed to the pod outputs The assignment is explained below To navigate through the menus described in the following use the bottom and side bezel buttons cf figure 11 Setting the internal clock Enter the SETUP menu and open the OSCILLATOR menu First the source can be chosen To synchronise the clock with the testboard clock choose EXT To enable the generation of an internal clock select INT If an inter nal clock is used the PLL should be turned ON If an external clock is used the PLL must be turned OFF Using an internal clock its frequency can be chosen pressing INT FREQUENCY If an external clock is used the EXT FREQUENCY influences the displayed frequency e g in the EDIT menu and should therefore be chosen as close as possibly to the actual external frequency Editing the channels To edit the channels enter the EDIT menu Subsequent a display will ap pear similar to figure 13 The leftmost entry in each row displays the name of the channel The pattern length can be edited pressing SETTINGS SET MEMORY SIZE The memory size
49. t of the pulse patterns 3 4 Rate Estimate and Measurement It is very important for the understanding but also to program the data generator accordingly to be able to estimate the hit rate In principle the rate corresponds to Since for all rate measurements the data generator sent the programmed pulse pattern repeatedly and without a delay between two patterns the relevant time is the pattern length Within the pattern length the amount of hits is given by the amount of hits generated by each pulse multiplied by the amount of pulses This is summarised in equation 1 where N is the amount of pulses Np is the number of hits per pulse M is the pattern length given in us and Aroc is the area of the ROC With a desired rate given the formula can be applied to estimate the pattern length and the amount of pulses to place in it setia BATAAN sal 1 m2 M A ROC To measure the rate a sequence consisting of a CAL calibrate a TRG trigger a TOK token and a delay was sent to the ROC The CAL signal injects charge in a pixel to emulate a hit Unless the pixel is armed though the charge is not being switched through such that it does not influence the pixel The TRG signal indicates the data to be readout before the TOK signal starts the readout process The data readout with one trigger is sometimes also called an event In general an event may or may not contain hits In the following events containing hits are labelled r
50. uence cf section 3 4 was defined consisting of the following signals CAL TRG TOK delay TRG TOK delay The delay was 510 clock cycles long the time between CAL and TRG tct was 106 clock cycles long and the time between TRG and TOK ttk was 32 clock cycles long A particular pixel was armed such that the voltage defined by Voc was switched through and read out by the first trigger only At the same time the data generator was providing pulses at an afore determined rate The address of the armed pixel was chosen not to lie within a cluster such that the test could search for a particular pixel address to identify the Vo As the readout was random both triggers may or may not contain hits due to the electric pulses The CAL signal was however a fix part of the se quence such that the Vc was expected in every even trigger as the counting started at zero The first test did break when encountering the first error and produced four ROC maps in total each plotting the amount of hits as a function of po sition An error was produced when either the Vo was found in an odd trigger or no Vo was found in an even trigger The first map contains all hits gathered during the test The map labelled error contains the data belonging to the trigger which produced the first error while the maps labelled previous and after first error contain the data belonging to the trigger preceding or following the error res
51. used and the connected cable 24 4 3 Pulse Height As whole clusters of hits are induced it was a priori not clear how the pulse height would be distributed Beforehand a pulse height calibration was done to assign an amount of charge to the digital value delivered by the pixel Figure 20 shows a ROC map where the pulse height was plotted as a func tion of position From the figure it becomes clear that the pulse height is not uniform over a cluster which leads to the broad pulse height distribution shown in figure 21 As a consequence the rate can be tuned gradually by varying the comparator threshold Figure 22 shows a plot where the rate is plotted as a function of the comparator threshold Indeed the increase in the rate is gradual until the threshold is on the level of the noise which occurs at VinrComp Value of about 135 VihrComp is a digital to analogue converter DAC which determines the height of the comparator threshold Map 801 z x103 o B E Entries 202376 300 705 5 Meanx 20 97 E Meany 43 99 700 NE er E y a 600 cr E g 50 E 500 M 400 E a us 300 Eu E EN 200 E E eos 100 0 E i i i 1 L 1 1 i L L 1 Ii Ii i i i i i i L 0 0 10 20 30 40 50 column Figure 20 AROC map The pulse height is plotted as a function of position 25 ph per hit occurence 2000 1500 1000 500 e Lila 0 50 100 150 200 250 300 350 400 450 500 pulse height Vcal Figure 21
Download Pdf Manuals
Related Search