Home

Testing and Troubleshooting Digital RF Communications

1. Center 850 MHz Span 3 MHz Figure 63 In channel interfering tone not visible in a the Frequency domain but detectable in b the constellation and c error vector spectrum Figure 64 Power amplifier TRACE dB 95 52 In channel spurious cause interference in the modulation A single spur combines with the modulated signal and the result depends on their phase relationship The spur is rarely high enough to be detected in the frequency domain but it may be identified in the constellation because it forms circles around the reference points The radius of the circle corresponds to the magnitude relationship between the interfering tone and the desired I Q signal There may be some randomness caused by noise and if the spur is very small the circle might not be clear even when zooming onto a single constellation point as in Figure 63b The best way to determine if an in channel spur is present is by looking at the error vector spectrum The magnitude and frequency offset of the spur from the unmodulated carrier can be measured from this display For instance the error vector spectrum in Figure 63c shows a spur at 850 053710 MHz 53 710 kHz away from the unmodulated carrier frequency R sr n TRACE A Ch QPSK Meas Tine Chi Spectrum 8 731 66 Const b 5 n div 681 66 ml 0 78621107165 TRACE B Ch QPSK Err Y B Marker
2. Time Ofs 2 4 us Pilot Freq Error 0 0 Hz Paging 0 dB Max IT Carrier FT 0 0 dB Sync 0 dB Avg IT Code domain power is affected by any I Q impairment Basically any impairment that degrades EVM will cause an increase in the code domain power noise floor that is an increase in the level of non active channels Figure 37 shows an increase in the code domain power noise floor for a cdmaOne system with an I Q gain imbalance of 3 dB Base Ch Freq 1 80000 GHz PN Ofs x 64 chips Code Domain IS 95A Averages 10 Ref 0 00 dB 63 Act Set Th 20 00 dB Time Ofs 12896 7 us Pilot Freq Error 0 5 Hz Paging Carrier FT 28 8 dB Sync 4 2 dB Avg RT 8 7 dB 4 5 dB Max IT 48 4 dB 10 5 dB Avg IT 50 2 dB I Q impairments are not linear errors and therefore cannot be removed by applying equalization Base Ch Freq 1 80000 GHz PN Ofs x 64Ichips fverases 10 7 Code Domain IS 95R Act Set Th 20 00 dB Time Ofs 12896 8 us Pilot Freq Error 3 2 Hz Paging Carrier FT 29 8 dB Syne 4 4 dB Avg RT 8 8 dB 4 6 dB Max IT 29 5 dB 10 6 dB Avg IT 33 5 dB Figure 38 Symbol encoder 37 3 2 3 Incorrect symbol rate 1001 Baseband ya E Filters Modulator IF Filter Amplifier Power Control IF LO RF LO The symbol clock of a digital receiver system dictates the sampling rate of the baseband I and Q wavef
3. 0 63806292835 Spec 850 053 710 937 5 Hz 1 861 dB EQ I Wm Center 1 8 GHz Span 3 MHz Center 850 MHz Span 3 90625 MHz An interfering tone within the frequency bandwidth of the modulated signal also results in uncorrelated energy that raises the code domain power noise floor Interfering tones are not linear impairments and cannot be removed by applying equalization although the signal quality may improve somewhat since the equalizer helps minimize EVM 3 2 9 AM PM conversion Other Channel 7771 Other Channel 1 0 Modulator Baseband Filters IF Filter IF LO RF LO Other Channel Other Channel 2 4 Figure 65 AM PM conversion Figure 66 Polar trajectory of a signal with AM PM conversion relative to ideal trajectory 53 As mentioned earlier the frequency and amplitude responses are important characteristics of power amplifiers Apart from compression AM AM conversion power amplifiers may cause phase distortion for high levels of signal amplitude This effect is known as AM PM conversion illustrated in Figure 65 AM PM conversion typically occurs at the linear range of the amplifier that is for amplitude levels below compression It is particularly relevant for signals with high peak to average power ratios where different amplitude levels suffer different phase shifts How can you verify AM PM conversion The best wa
4. dB MaxP 12 3 ExtAt 0 0 Trig Free 1 930 GHz 1 945 GHz Res BH 30 00 kHz Segment Center Channel Power 15 26 dBm Offset Freq a from Limit a from Carrier Horst Spur 9 113 MHz 4 48 dB 49 48 dBc Marker Out of band measurements are those outside the system frequency band 2 4 1 Spurious and harmonics While spurious are caused by different combinations of signals in the transmitter harmonics are distortion products caused by nonlinear behavior in the transmitter They are integer multiples of the transmitted signal s carrier frequency Out of band spurious and harmonics are measured to ensure minimum interference with other communications systems Figure 24 2 3 Figure 24 Out of band spurious and harmonics measurement 2 5 Best practices in conducting transmitter performance tests 26 TRACE B Chl Spectrum A Offset 850 000 000 Hz B Offset 350 000 000 Hz aa ogo tt LL w LL IN Lui LogMag 85 dBri Center 1 5 GHz Span 2 GHz By following certain guidelines in conducting design verification tests you can greatly increase the probability that the transmitter will operate properly in the real world environment The test equipment should be carefully chosen to reduce measurement uncertainties and increase confidence in correct transmitter operation When performing absolute pow
5. 2 3 1 6 Timing measurements Timing measurements are common on TDMA systems where the signal is bursted The measurements assess the envelope of the carrier in the time domain against prescribed limits Measurements include burst width rise time fall time on time off time peak power on power off power and duty cycle 17 Figure 13 Timing measurements 90 Amplitude Points 50 Amplitude Points 10 Amplitude Points gt n Peak Power x m Burst Top Amplitude Overshoot Burst Base Amplitude Burst Width 9 T 4 Off time gt Rise Time Fall Time A Burst Interval or Period Timing measurements are mainly important to ensure minimum interference with adjacent frequency channels or timeslots during signal turn on and turn off For instance if the transmitter turns off too slowly the user of the next timeslot in the TDMA frame experiences interference If it turns off too quickly the power spread into adjacent frequency channels increases 3 4 2 3 1 7 Modulation quality measurements There are a number of different ways to measure the quality of a digitally modulated signal They usually involve precision demodulation of the transmitted signal and comparison of this transmitted signal with a mathematically generated ideal or reference signal as we saw
6. mismatch between the RF amplifier and the antenna or between any components in the IF or RF section of the transmitter may cause reflections that result in distortion of the overall transmitter s frequency response The distortion is either tilt or ripple depending on the distance between the mismatched components and the bandwidth of the signal As you can see in Figure 54 for narrow bandwidths relative to the inverse of the distance between the components the frequency shape appears tilted For wide bandwidths it appears rippled The effect of mismatch between components is usually negligible for the bandwidths used in wireless RF applications BW lt lt 1 d while it becomes more important in applications using wider bandwidths such as LMDS Local Multipoint Distribution System or satellite applications IF Filter Upconverter Amplifier l l l i je ifBW lt 1 d lt if BW gt gt 1 0 gt How can you verify filter tilt or ripple Filter tilt or ripple causes distortion of the demodulated baseband signal Therefore the constellation and EVM are degraded The best way to verify IF filter errors or equivalent effects is by applying equalization to the signal and checking the bits to RF frequency response Equalization removes linear distortion Therefore both the constellation and magnitude of the error vector versus time should improve noticeably Assuming that there ar
7. Span 4 00 MHz Total Pur Ref 11 38 dBm 1 00 MHz ACPR Lower c Upper Offset Freq Integ Bu dB dBm dBc dBm 1 50 MHz 1 00 MHz 61 87 73 25 61 98 73 37 When making ACPR measurements it is important to take into account the statistics of the signal transmitted CCDF curves can be used for this purpose as we saw earlier Different peak to average ratio values have a different impact on the non linear components of the transmitter such as the RF amplifier and therefore on the ACPR as well Higher peak to average ratios in the transmitted signal can cause more interference in the adjacent channel ACPR measurements on the same transmitter can provide different results depending on the statistics of the transmitted signal When measuring ACPR in CDMA base stations for example it is important to consider the channel configuration used Different standards have different names and definitions for the ACP measurement For example for TDMA systems such as GSM there are two main contributors to the ACP the burst on and off transitions and the modulation itself The GSM standards name the ACP measurement Output RF Spectrum ORFS and specify two different measurements ORFS due to modulation and ORFS due to switching 3 In the case of NADC TDMA the ACP due to the transients and the modula tion itself are also measured separately for mobile stations Additionally a weighting function that corresponds to the receiver baseband filter r
8. fverages 10 Ref 0 00 dB Power Mixing Products Ref 0 00 4B 5 00 dB a Figure 29 Sync Code domain power for a non compressed signal versus b compressed signal Time Ofs Freq Error Carrier FT 1 See Glossary for definition 12896 7 us Pilot 28 8 dB Sync pU fict Set Th 20 00 dB 63 Walsh Channel Act Set Th 20 00 dB 12896 8 us Pilot 0 3 Hz Paging 28 5 dB Sync 8 9 dB 33 8 dB 44 2 dB Time Ofs Freq Error Carrier FT 8 7 dB 48 4 dB 50 2 dB 4 2 dB Avg AT 4 4 dB Max IT 10 7 dB Avg IT 4 2 dB Avg RT 4 5 dB Max IT 10 5 dB Avg IT 0 5 Hz Paging Compression is not a linear error and cannot be removed by equalization Figure 30 1 Q modulator Figure 31 1 Q gain imbalance 32 1 Q Impairments Amplifier I I i Power Control RF LO I I I I Baseband Filter 2 AI I Q impairments can be caused by matching problems due to component differences between the I side and Q side of a network The most common I Q impairments are listed below 1 VQ gain imbalance Since I and Q are two separate signals each one is created and amplified independently Inequality of this gain between the I and Q paths results in incorrect positioning of each symbol in the constellation causing errors in recovering the data see Figure 31 This problem is rare in systems where
9. l Actual l l i i Symbol Period l I I A 1 I I I I I I l 1 I Sample Period 2 1 Graph of Error at Each Sample 0 2 1 0 1 2 The smaller the symbol rate error the more symbols are required to detect the error that is to form the V shape For instance in Figure 39b for a QPSK system with a symbol rate specified at 1MHz 100 symbols are measured to form a V shape in the magnitude of the error vector versus time display for an actual symbol rate of 1 0025 MHz In the same case about 500 symbols are required to form a similar V shape for an actual symbol rate of 1 00025 MHz Figure 41 Baseband filters Figure 42 a Time and b Frequency response of raised cosine filters with different alphas 39 The actual transmitted symbol rate can be found by adjusting the symbol rate in the measuring instrument by trial and error until magnitude of the error vector versus time looks flat Small symbol errors also affect the code domain power measurement The code domain power noise floor increases proportionally to the magnitude of the error Large errors The best way to verify large symbol rate errors that produce unlock conditions in the measurements is by measuring the signal s channel bandwidth to roughly approximate the symbol rate as explained in section 2 3 1 1 Errors in the symbol rate are not linear and cannot be minimized by applying equalization 3 2 4 Wrong filter co
10. 0 e Channel Power O e e e e Occupied Bandwidth e L e e e Peak to Average e e e CCDF e e ACPR or Equivalent eo O e eo O In Band Spurious e e e e e Stor Hermans 2 68 GHz a GHz e O e Timing e e e e e EVM or Phase Error 96 rms e O e orbhase Errar korsua inek 0 e Error Vector Spectrum e Rho e e Code Domain Power e USUS EEUU Oa e o Polar and Constellation e e e Eye or Trellis Diagram e O Symbol Table o o e Sync Word as Trigger eo e e varado e Baseband and Q Inputs e Mathematical Functions e Time Capture e Spectrogram e SEE PT o o o Notes 1 Measurement personalities and or options might be required to perform some of the measurements 2 Measurements pre configured for cdmaOne GSM NADC PDC W CDMA and cdma2000 3 Measurements pre configured for specific wireless systems depending on the measurement personality used GSM cdmaOne NADC PHS PDC or DECT 4 Only available for cdmaOne 5 Only available for PDC 6 Search must be manual or programmed by user 7 Only available for W CDMA 8 Only available for GSM 9 Only available for DVB C Digital Video Broadcast Cable with the appropriate measurement personality 63 5 Glossary SG at O E d ces e SRM NS Third Generation ACP ch ome es vu eT ee SOR trs Adjacent Channel Power ACPR iuh o Se ERU Adjacent Channel Power Ratio bits to RF frequency response Frequency response from baseband to RF using bits as the sti
11. Modulator Different implementations and systems have different design problems and often require specific measurements For instance in TDMA technologies burst parameters must be measured to ensure that interference with adjacent frequency channels and adjacent timeslots are within acceptable limits The most common measurements and problems associated with particular technologies will be discussed in the following chapters Amplifier Power Control The typical receiver Figure 4 is essentially an inverse implementation of the transmitter Although an I Q demodulator is often used there are other digital communications receiver designs Low Noise Amplifier with Automatic Baseband Gain Control IF Filter Filters CHAOS Output Data or Voice The receiver configuration also depends on the specific system and performance required For instance in high performance cellular receivers equalization is commonly used to combat ISI caused by impairments in the transmitter the air interface or the early stages of the receiver itself A more detailed description of digital receivers can be found in the companion Hewlett Packard application note Testing and Troubleshooting Digital RF Communications Receiver Designs 5l 2 Testing Transmitter Designs 2 1 Measurement model Figure 5 Measurement model There are several testing stages during the design of a digital communications transmitter The different
12. al I I Implemented digitally 1 I Power Control RF LO Baseband 1 0 Filters Modulator IF Filter Upconverter Amplifier b Power Control Implemented digitally IF LO RF LO Digital implementations of the baseband and IF sections of the transmitter avoid some of the typical impairments caused by analog hardware Drift errors due to component aging are also eliminated However digital hardware can also cause impairments as we will see in the last chapter of this application note Figure 3 Block diagram of a GMSK transmitter using a Frequency modulator 1 2 Digital communications receiver Figure 4 Block diagram Preselecting Filter of a digital communications receiver 1 See Glossary for the meanings of these acronyms 1 1 2 Other implementations In practice there are many variations of the general block diagrams discussed above These variations depend mainly on the characteristics of the technology used for example the type of multiplexing TDMA or CDMA and the specific modulation scheme such as OQPSK or GMSK 1 For instance GSM transmitters can be easily implemented using analog frequency modulators Figure 3 Since intersymbol interference is not as critical in GSM systems Gaussian baseband filtering is used instead of Nyquist filters 1 I Q modulators are only used in high performance GSM transmitters 1 0 1 1 Gausian Baseband Filter Frequency i
13. analyzers and vector signal analyzers can perform frequency domain measurements The main difference between them is that traditional spectrum analyzers are swept tuned receivers while vector signal analyzers capture time data and perform Fast Fourier Transforms FFTs to obtain the frequency spectrum In addition the VSAs measure both the magnitude and phase of a signal Amplitude SS ES Frequency Figure 7 Demodulating the signal and calculating a reference signal for modulation quality measurements 1 Measurements in the frequency domain are especially important to ensure that the signal meets the spectral occupancy adjacent channel and spurious interference requirements of the system 2 2 3 Modulation domain If the RF signal is demodulated the quality of the baseband signal can be analyzed by comparing it to an ideal reference This reference is usually mathematically derived by the instrument provided that the original data sequence can be recovered Demodulation involves applying the appropriate filtering before recovering the baseband I and Q signals and sampling these signals at the symbol rate to recover the actual symbols Figure 7 Tx Under Test Providing Tx Filtering Demodulated Data From Incoming Signal Measuring Instrument Measuring Instrument Ref Tx Ref Rx gt Reference Signal Providing Tx Filtering Providing Rx Filtering Measuring Instrument Providing Rx Filtering Measured Signal V
14. earlier The definition of the actual measurement depends mainly on the modulation scheme and standard followed NADC and PDC for example use Error Vector Magnitude EVM while GSM uses phase and frequency error cdmaOne uses rho and code domain power These and other modulation quality measurements are described in the following sections 2 3 1 7 1 Error Vector Magnitude EVM The most widely used modulation quality metric in digital communications systems is Error Vector Magnitude When performing EVM measurements the analyzer samples the transmitter output to capture the actual signal trajectory The signal is usually demodulated and a reference signal is mathematically derived The error vector is the vector difference at a given time between the ideal reference signal and the measured signal The error vector is a complex quantity that contains a magnitude and a phase component It is important not to confuse the magnitude of the error vector with the magnitude error or the phase of the error vector with the phase error A graphical depiction of these differences can be seen in Figure 14 Figure 14 Error vector and related parameters Figure 15 a Polar diagram b magnitude of the error vector versus time c error vector spectrum d summary table and symbol table 18 Magnitude of Error Vector Error Vector Phase of Error Vector Measured Signal Phase Error Ideal Signal Reference
15. periodically based on the data measured and arbitrarily assigns the phases to I or Q Using an appropriate sync word as a trigger reference makes the constellation stable on the screen permitting the correct orientation of the symbol states to be determined Therefore the relative gains of I and Q can be found for gain imbalance impairments and the phase shift sign between I and Q can be determined for quadrature errors I Q offset errors may be compensated by the measuring instrument when calculating the reference In this case they appear as an I Q offset metric Otherwise I Q offset errors result in a constellation whose center is offset from the reference center as seen in Figure 33 The constellation may tumble randomly on the screen unless a sync word is used as a trigger for the same reason indicated above Delays in the I or Q paths also distort the measured constellation However if the delay is an integer number of samples the final encoded symbols transmitted appear positioned correctly but are incorrect The error cannot be detected unless a known sequence is measured Mathematical functions in the measuring instrument can help compensate for delays between I and Q by allowing you to introduce delays in the I or Q paths In this way you can confirm and measure the delay For any of these errors magnifying the scale of the constellation can help detect subtle imbalances visually Since the constellation is affected these
16. small errors in the symbol rate is by looking at the magnitude of the error vector versus time display If the symbol rate is slightly off this display shows a characteristic V shape as in Figure 39b 38 TRACE A Chi QPSK Meas Tine 1 5 Const 7 Figure 39 300 i a Constellation and a yaiv b magnitude of the E error vector versus 1 5 time with V shape 4 4444442987 4 444944929874 TRACE B Chi QPSK Err V Tine caused by incorrect symbol rate xL T y Start O syn Stop 99 syn This effect can be understood by studying Figure 40 For simplicity a sinewave is used instead of a digitally modulated signal and its frequency symbol rate is slightly higher than the specified sample frequency symbol rate chosen in the measuring instrument At one arbitrary reference sample called 0 the signal will be sampled correctly Since the symbol rate is slightly off any other sample in the positive or negative direction will be slightly off in time Therefore the signal will deviate by some amount from the perfect reference signal This deviation or error vector grows linearly on average in both the positive and negative directions Therefore the magnitude of the error vector versus time shows a characteristic V shape 2 1 0 1 2 com or 1 l I Figure 40 Error 2 or 2 Symbol rate slightly higher than specified i i i i i i i
17. 3 1 6 Timing measurements 200 16 2 3 1 7 Modulation quality measurements 17 2 3 1 7 1 Error Vector Magnitude EVM 17 2 3 1 7 2 VQ offset 0 ee eee 20 2 3 1 7 3 Phase and frequency errors 20 2 3 1 7 4 Frequency response and group delay 21 A SA ME Lm 22 2 3 1 7 6 Code domain power 005 22 2 3 2 Out of channel measurements usus 23 2 3 2 1 Adjacent Channel Power Ratio ACPR 23 2 5229 PUTOS 2 3 aon toe HEC dete eek YA Ad Pedes 25 2 4 Out of band measurements 00 0 0000 cee eee 25 2 4 1 Spurious and harmonics 000000 25 2 5 Best practices in conducting transmitter performance tests 26 3 Troubleshooting Transmitter Designs 3 1 Troubleshooting procedure 00 002 ee eee ee 9 2 mpairments or eo ined Ae agg E Ste ERU RACE eg Appendix A Appendix B 5 Glossary 3 2 1 COMPression 0 0 ce eee ee 3 2 2 IQ impairments 0 0 0000 cee 3 2 3 Incorrect syMbolTate o oo oooooooooooooo 3 2 4 Wrong filter coefficients and incorrect windowing 3 2 5 Incorrect interpolation 0 0 0 00 e eee 3 2 6 Filter tilt or Tipple o o oooooooooooo oo SOMOS ADA ti aaa 3 2 8 Interfering tone 0 00 cee eee 3 2 9 AM PM conversion 0 000 000 cee eee 3 2 10 DSP and DAC impairments 0 3 2 11 Burst shaping impairm
18. 3 Halsh Channel 63 Act Set Th 20 00 dB Act Set Th 20 00 dB b with phase noise P Time Ofs 12896 7 us Pilot 4 2 dB Avg AT 8 7 dB Time Ofs 6179 1 us Pilot 4 2 dB Avg AT 8 8 dB Freq Error 0 5 Hz Paging 4 5 dB Max IT 48 4 dB Freq Error 0 2 Hz Paging 4 5 dB Max IT 34 3 dB Carrier FT 28 8 dB Sync 10 5 dB Avg IT 50 2 dB Carrier FT 45 2 dB Sync 10 5 dB Avg IT 41 7 dB LO instability is not a linear error and cannot be removed by equalization Figure 61 Mixers Figure 62 Out of channel spur spectrum view 51 3 2 8 Interfering tone 0 1 Baseband AS Filters IF Filter A tone or spur generated anywhere in the transmitter can interfere with the transmitted signal if it falls in the signal s bandwidth In chamnel interfering tones are usually masked by the signal in the frequency domain If the interfering tone is outside of the signal s bandwidth it can cause interference with other channels or systems Interfering tones are typically caused by interactions of internal signals in active devices such as mixers and amplifiers How can you verify the presence of interfering tones Out of channel or out of band spurious are easily detected by spurious harmonics measurements if the test instrument has sufficient dynamic range see Figure 62 TRACE A Chi Spectrun _ B Parker 850 997 500 Hz 64 423 dBn dBin LogMag 105 dBn
19. 55 4026 1 800 452 4844 Canada Hewlett Packard Canada Ltd 5150 Spectrum Way Mississauga Ontario LAW 5G1 905 206 4725 Europe Hewlett Packard European Marketing Centre P O Box 999 1180 AZ Amstelveen The Netherlands 31 20 547 9900 Japan Hewlett Packard Japan Ltd Measurement Assistance Center 9 1 Takakura Cho Hachioji Shi Tokyo 192 Japan Tel 81 426 56 7832 Fax 81 426 56 7840 Latin America Hewlett Packard Latin American Region Headquarters 5200 Blue Lagoon Drive 9th Floor Miami Florida 33126 U S A 305 267 4245 4220 Australia New Zealand Hewlett Packard Australia Ltd 31 41 Joseph Street Blackburn Victoria 3130 Australia 1 800 629 485 Asia Pacific Hewlett Packard Asia Pacific Ltd 17 21 F Shell Tower Times Square 1 Matheson Street Causeway Bay Hong Kong Tel 852 2599 7777 Fax 852 2506 9285 Data Subject to Change Copyright 1999 Hewlett Packard Company Printed in U S A 7 99 5968 3578E
20. Error Vector Magnitude is the root mean square rms value of the error vector over time at the instants of the symbol clock transitions By convention EVM is usually normalized to either the amplitude of the outermost symbol or the square root of the average symbol power 4 Apart from the constellation and polar diagrams other important displays associated with EVM that are mentioned in this application note are magnitude of the error vector versus time the spectrum of the error vector error vector spectrum phase error versus time and magnitude error versus time Figure 15 shows some of these displays TRACE B Chi PI 4 Err V Tine B Marker 39 200000 syn 98 492 wx 1 0 300 a n c fdiv 1 5 800 EVM 288 63 maras A MA pk at syn e Err gp tag M pk at syn Linffag 124 15 80 h fh PE COT oT b mo 10 Offset d fdiv Amp Droop 0 x Left 29 601 syn Right 39 201 sym Figure 16 Constellation zoomed and magnitude of the error vector versus time a without equalization and b with equalization a 19 EVM and the various related displays are sensitive to any signal flaw that affects the magnitude and phase trajectory of a signal for any digital modulation format Large error vectors both at the symbol points and at the transitions between symbols can be caused by problems a
21. Figure 10 Use the frequency error metric given in the summary table when performing modulation quality measurements as shown in Figure 15 2 3 1 3 Channel power Channel power is the average power in the frequency bandwidth of the signal of interest The measurement is generally defined as power integrated over the frequency band of interest but the actual measurement method depends on the standard followed 21 3 4 NAAA A eee Base Ch Freq 1 93125 GHz Channel Power Averages 20 _ Ref Lvl 11 49 dBm 10 00 dB Figure 9 MaxP Channel power 11 3 measurement ExtAt 0 0 Trig Free Center 1 93125 GHz Span 2 00000 MHz Res BH 27 857 kHz Channel Power Power Spectral Density 11 62 dBm 1 23000MHz 72 52 dBm Hz Figure 10 Occupied bandwidth measurement 14 Power is the fundamental parameter of any communication system The goal in wireless systems is to maintain each link sufficiently with minimum power This gives two benefits overall system interference is kept to a minimum and in the case of mobile stations battery life is maximized Output power therefore is controlled within tight limits If a transmitter produces too little power link performance is compromised too much and interference with other transmitters may be too high and battery life too short In the case of CDMA systems where total interference is a limiting factor for capacity controlling the power of each mobile is also essential to a
22. Filter liue ELE mt The compensation can be implemented digitally somewhere before the DAC or in combination with the analog reconstruction filter see Figure 69 or the IF filter Figure 70 It is a common error not to compensate for this sin x x shape Figure 69 illustrates the importance of the compensation for a transmitter with analog I Q modulation Non compensation causes distortion in the spectrum of the transmitted signal If the symbol period T is comparable to the pulse width x not compensating for the sin x x function can cause an effect as dramatic as incorrect interpolation 56 Baseband Reconstruction Filter Filter Compensation Figure 69 H F Compensating for H f a H f sin x x in a transmitter with an analog I Q modulator fsample 2f sample fsample 2f sample S Non Compensated Correct Frequency Response Frequency Response Reconstruction Filter Removes Higher Frequency Components and Compensates for sin x x Figure 70 Compensating for Figure 70 illustrates the effect of compensation for a transmitter with a digital sin x x in a transmitter IF If there were no compensation the spectrum of the transmitted signal with a digital IF would be tilted Baseband 1 0 IF Filter Filters Modulator Compensation Implemented Digitally 1860 Hat H A IF 2 sample IF 1F 2 iw IF 7 IF Non Compensated Correct Frequency Response Frequency Response Reconstruction Fi
23. LA Sickan He Wireless Test Solutions Testing and Troubleshooting Digital RF Communications Transmitter Designs Application Note 1313 A a Introduction The demand for ubiquitous wireless communications is challenging the physical constraints placed upon current wireless communications systems In addition wireless customers expect wireline quality from their service providers Service providers have invested a lot in a very limited slice of the radio spectrum Consequently network equipment manufacturers must produce wireless systems that can be quickly deployed and provide bandwidth efficient communications At early stages of equipment design rigorous testing is performed to ensure system interoperability The increasingly complex nature of digital modulation is placing additional pressure on design teams already faced with tight project deadlines Not only must the designer test to conformance he must also quickly infer possible problem causes from measurement results The objective of this application note is to help you understand why the different transmitter tests are important and how they identify the most common impairments in transmitter designs The application note focuses on cellular communications transmitters although some of the measurements and problems described may also apply to other digital communications systems This application note covers the following topics 1 How digital communica
24. MHz Span 94 921875 kHz TRACE B Chl PI 4 Err V Tine error vector versus e Harker 0 0000 syn 465 92 wx time improves with A EQ equalization ibi Linlag 800 mA div 1 Left O syn Right 20 syn 3 2 5 Incorrect interpolation Figure 46 Baseband filters 1 0 0 1 Baseband IF Filter Amplifier Baseband Filter Power Control RF LO As we saw earlier baseband filters in the transmitters are usually implemented as digital filters Digital techniques ensure that we can reproduce filters as many times as desired with exactly the same characteristics Since the RF signal is analog the digital signal has to be converted to analog at some point in the transmitter After the DAC used for that purpose an analog reconstruction filter smooths the recovered analog signal by filtering out the unwanted frequency components As per the Nyquist criteria the sampling frequency should be at least twice the highest frequency component of the signal being sampled In the case of a digitally modulated signal sampling at the symbol rate does not meet the Nyquist criterion When recovering the analog signal filtering could be an issue and significant aliasing can occur as shown in Figure 47 Figure 47 Sample rate symbol rate Effect in a the time domain and b Frequency domain Figure 48 Interpolation sample rate 4 x symbol rate Effect in the
25. a time domain and b frequency domain 43 Reconstruction ET Baseband ET Incorrect Amplitude Values at Symbol Clock Transitions f 2f f sample sample sample Aliasing P d sample 2fsample Therefore it is recommended to use a sample rate at least twice as high as the symbol rate This is accomplished by interpolating The technique generally used consists of adding zeros between the samples before applying baseband filtering as shown in Figure 48 Since the samples added are zeros the time response of the filter is not affected The only effect is to provide extra data points between the symbols This better describes the transitions between waveforms In the frequency domain the images are now shifted much higher in frequency thereby reducing the requirements of the reconstruction filter to a reasonable level Reconstruction Filter Baseband Filter Interpolated Correct Amplitude Samples Values at Symbol I I Clock Transitions f sample sample 44 interpolation versus It is a common mistake to interpolate the I and Q signals as if they were pulses instead of impulses In other words instead of adding zeros it is a common mistake to add samples that have the same value as the original samples The response of the filter is then affected as shown in Figure 49 The frequency response of an impulse is a constant amplitude When multiplied by the frequency response of the baseband filter the
26. and small at the symbol points Al How can you verify errors in the alpha coefficient and the windowing The main indicator of a wrong alpha coefficient and incorrect windowing is the display of magnitude of the error vector versus time In systems that use Nyquist filtering an incorrect alpha or a mismatch in alpha between transmitter and receiver causes incorrect transitions while the symbol points themselves remain mostly at the correct locations Therefore the EVM is large between the symbols while it remains small at the symbol points Figure 44 shows that effect for a mismatch between the transmitter filter alpha 0 25 and the receiver filter in the measuring instrument alpha 0 35 as specified for a particular system Incorrect windowing has the same effect since the actual baseband response of the transmitter and the baseband response applied in the measuring instrument no longer match The amplitude overshoot of the baseband signal depending on the alpha can be observed in the polar diagram TRACE C D3 PI 4 Meas Tine 1 5 4 4444442987 4 444449429874 TRACE D D4 PI 4 Err Y Tine 8 T x LinMag b goo ir fdiv 0 y A i N Left O syn Right 20 sym Small Error Vector at Symbol Points Baseband filtering errors may also appear in the following measurements Channel bandwidth The 3dB bandwidth of a root raised cosine
27. cause different delays in the I and Q paths Different electrical lengths in the I and Q paths may also cause significant delay differences between the two paths especially for signals with wide bandwidths high symbol rates Refer to Figure 34 Figure 34 Q delayed relative to Figure 35 1 Q swapped 34 1 Symbol Rate 1 Symbol Rate _ gt _ gt ke 1 bR p T bR 9 5 VQ swapped Swapping the I and Q signals reverses the phase trajectory and inverts the spectrum around the carrier Therefore swapping the I and Q or changing the sign of the shift 90 or 90 degrees makes a difference in the signal transmitted cos torot jsin rot versus cos wyol jsin wyot The mapping of the IF I and Q signals is reversed which causes symbol errors as seen in Figure 35 10 4 e 00 land O Swapped l I l t l l 1 6 01 35 How can you verify the different I Q impairments The best way to verify most I Q impairments is to look at the constellation and EVM metrics I Q gain imbalance results in an asymmetric constellation as seen in Figure 31 Quadrature errors result in a tipped or skewed constellation as seen in Figure 32 For both errors the constellation may tumble randomly on the screen This effect is caused by the fact that the measuring instrument decides the phases for I and Q
28. chieve maximum capacity Therefore accurate control of the transmitted power is critical in defining a system s capacity coverage and signal quality 2 3 1 4 Occupied bandwidth Occupied bandwidth is closely related to channel power It indicates how much frequency spectrum is covered by some given percentage often 99 of the total power of the modulated signal For instance in Figure 10 the bandwidth that includes 99 of the power is 1 260 MHz Any distortion harmonic or intermodulation produces power outside the specified bandwidth a REF 3 6 dBm AT 18 dB SHFL SCF BS CH 758 CENTER 892 740 MHz SPAN 3 009 MHz RES Bl 38 kHz HUBN 388 kHz SUP 28 8 msec OCCUPIED EH 99 66 PASS Carrier Frequency Error Delta Frequescy 1 266 HHZ r7 5 kHz 2 3 1 5 Peak to average power ratio and CCDF curves Peak to average power ratio and CCDF defined below are statistical measurements on the time domain waveform Peak to average power ratio is the ratio of the peak envelope power to the average envelope power of a signal during a given period of time Some instruments may provide peak to average power statistics that is the peak envelope power is given not as an absolute peak but rather as the power level associated with a certain probability For example in Figure 11 the measurement shows that the power is below a level that is 9 455 dB above the average 99 99 of the time that is there is a probability of 01 that the pow
29. components and subsections are initially tested individually When appropriate the transmitter is fully assembled and system tests are performed During the design stage of product development verifi cation tests are rigorous to make sure that the design is robust These rigorous conformance tests verify that the design meets system requirements thereby ensuring interoperability with equipment from different manufacturers This chapter covers testing of the highlighted part of the transmitter in Figure 1 It describes the conformance tests and other common measurements made at the antenna port Most of the transmitter measurements are common to all digital communications technologies although there are some variations in the way the measurements are performed Particular technologies such as CDMA or TDMA require specific tests that are also described Transmitter measurements are typically made at the antenna port where the final signal is emitted In this case the measurement equipment is used as an ideal receiver It may also be necessary to examine the transmitter at various test points as the different sections are designed see Figure 5 In this case a stimulus signal might be required to emulate those sections that are not yet available The equipment for doing this acts as an ideal substitute for the missing circuit or sections Unmodulated carrier signals have traditionally been used as the stimuli for some component and subsyste
30. correct spectrum is obtained The frequency response of a pulse in the time domain is a sin x x function where x aft and v is the width of the pulse When the pulse signal is filtered its frequency response gets multiplied with the filter s response The result for a transmitter with analog I Q modulation is an incorrect spectrum that has a rounded top instead of a flat one In the case of a transmitter with digital IF this error could cause different effects Baseband a Filter i N Himpulself Hilf Baseband Filter K Houself sin x x Hilf D A If different sections of the transmitter are designed with different sampling rates an interpolation problem may also occur For instance if the FIR filter or the digital IF section and the DAC work at different rates some kind of interpolation between these sections is required In this case either adding zeros or samples equal to the original ones to the signal between the two sections may distort the signal How can you verify interpolation errors Since incorrect interpolation modifies the overall baseband filter frequency response intersymbol interference occurs Therefore the polar or constellation diagram and EVM are degraded The best way to verify interpolation errors is by applying equalization to remove the effects of those errors in the polar or constellation diagram and EVM see Figure 50 and checking the bits to RF frequency response Assuming there are
31. cy Channel coding and interleaving are common techniques that provide protection from errors The data is also processed and organized into frames The frame structure depends on the specific system or standard followed wee ee ee A A A A A A A A A A A A A A A A A A A a a a ee ee Baseband 1 Q Filters j Modulator IF Filter Upconverter Amplifier Power Control I I I I I I Processing I I I I i RF LO I I The symbol encoder translates the serial bit stream into the appropriate I and Q baseband signals corresponding to the symbol mapping on the I Q plane for the specific system An important part of the encoder is the symbol clock which defines the frequency and exact timing of the transmission of the individual symbols Once the I and Q baseband signals have been generated they are filtered Filtering slows the fast transitions between states thereby limiting the frequency spectrum The correct filter must be used to minimize intersymbol interference ISI Nyquist filters are a special class of filters that minimize ISI while limiting the spectrum In order to improve the overall performance of the system filtering is often shared between the transmitter and the receiver In that case the filters must be compatible and correctly implemented in each to minimize ISI Figure 2 Usual implementations of a transmitter a analog 1 Q modulator b digital IF The filtered I and Q baseband si
32. e 1 will help you identify the most common impairments that might be affecting a specific measurement when you are testing your design 3 2 Impairments Impairments M e 5 8 S 9 9 E 3 23 82 vj J ca 9 pa s 5 8 S E 2 83 Ss 5 5 3 3 a sj E Ss 2e S 2 L3 3 Bof amp Jit S o s a lt kJ e 3 ae L Pi gt c c5 x S o an o ce j amp 2 8 ss s amp 2 5 s S 2 S SA SSS S EJ JZ a Table 1 Channel BW e o0 06 e Impairments versus measurements affected Channel Power OIOOIO ojo CCDF eo O O 3 ACP e O e g Spurious e 2 5 Timing O E 5 EVM o ojo o o ojo ojo ojo 2 or Phase Error Code Domain Power or rho 0 0 000 dl dll Bits to RF Frequency Response evo e amp Group Delay High probability that impairment affects the measurement Q Severe cases of impairment may affect the measurement For instance a high level of ACP is probably caused by one of the following impairments Compression at the amplifier Wrong filter coefficient or incorrect windowing at the baseband filter ncorrect interpolation LO instability Burst shaping error DAC DSP error A severe case of symbol rate error You can further analyze and verify if any of these impairments is present by following the directions given in the following sections Figure 25 Power amplifie
33. e Error Rho Code Domain Power Phase Error lt lt Mag Error Magnitude Error Versus Time Phase Error gt gt Mag Error Phase Error Mag Error Constellation Asymmetric Phase Error Versus Time 1 0 Imbalance Amplitude Droop Tilted Quadrature Offset 1 0 Offset Higher Phase Error at Larger Signal Amplitudes Phase Noise Waveshapes Residual PM Symmetrical AM PM Conversion Mag of Error Vector Versus Time Uniform Noise Error Vector Spectrum Flat Noise Frequency Response Error Vector Small at Symbol Times and Large in Between Incorrect Alpha or Windowing V Shape Symbol Rate Error DSP Overflow Discrete Tones In Channel Interfering Tones Tilt or Ripple IF Filter Problems or Non Compensated Sin x x Sin x x Wrong Interpolation or Non Compensated Sin x x 62 Appendix B Instrument capabilities This table shows the current capabilities of HP vector signal analyzers and spectrum analyzers for testing and troubleshooting digital communications transmitter designs Vector Signal Analyzers Spectrum Analyzers ep Hpo Panamitan e wh HP asen HP ESAE Capabilities Series Tester Measurement E Series Series Personality Carrier Frequency e e e e e Channel Bandwidth e O O
34. e no other linear errors in the transmitter the bits to RF frequency response shows the ripple or the tilt as seen in Figure 55 48 TRACE A Chi QPSK EQ Chan cab EQ Figure 55 Log ag Bits to RF frequency response of a transmitter with IF filter tilt T T 2 5 1 1 J Center 1 8 GHz Span 3 90625 MHz In the case of SAW filters the time delays in the filter impulse response can be further analyzed by increasing the equalizer filter length and observing the equalizer impulse response Filter tilt or ripple may also appear in the following measurements e Channel bandwidth Tilt and ripple may show subtly in the frequency domain if averaging is applied CCDF curves Tilt or ripple affect the statistics of the signal and therefore the CCDF curves Code domain power Filter tilt or ripple can also raise the code domain power noise floor 3 2 7 LO instability 100 1 Baseband 1 0 Filters Modulator IF Filter Upconverter Amplifier Figure 56 IF and RF LOs Power Control The RF LO characteristics are passed on to the final RF signal If the frequency is unstable the channel occupation restrictions may be violated and interference with adjacent channels may occur LO instability may be caused by hum or other signals that run close to the LO For instance data lines typically CMOS carry signals with steep edges that can couple RF ener
35. ector signal analyzers can demodulate the signal and make modulation quality measurements A swept tuned spectrum analyzer with additional hardware and software can also demodulate and analyze modulation quality Various display formats and capabilities can be used to view the baseband signal characteristics and analyze modulation quality e Q polar vector and constellation diagrams e Summary table with I Q quality metrics such as Error Vector Magnitude EVM magnitude error phase error frequency error rho and I Q offset e Magnitude of the error vector versus time and error vector versus frequency error vector spectrum e Magnitude error and phase error versus time or frequency e Eye and Trellis diagrams e Symbol table e Equalization which allows frequency response and group delay measurements e Code domain analysis Some of these display formats and capabilities are briefly described in the modulation quality measurement sections in this chapter For more detailed descriptions of the common modulation quality displays please see M Combinations of the display formats and capabilities above can be used to troubleshoot potential problems in the design as we will see in the next chapter Additionally analog demodulation tools such as phase demodulation or frequency demodulation can be used to troubleshoot particular problems in a digital communications transmitter For example phase demodulation is often used to tr
36. ed only the measurement technique TRACE C io MN Meas Tine t 0 0000 syn 1 0008 45 055 deg 754 93 EQ M Const b i Wi x LN o LEM 0 772179353984 Figure 17 a Ideal constellation versus b offset constellation Figure 18 Phase and frequency error measurement 20 2 3 1 7 2 VQ offset DC offsets at the I or the Q signals cause I Q or origin offsets as shown in Figure 17 1 Q offsets result in carrier feedthrough Some instruments compensate for this error before displaying the constellation or polar diagram and measuring EVM In that case 1 Q offset is given as a separate error metric Offset Constellation Ideal Constellation a b 2 3 1 7 3 Phase and frequency errors For constant amplitude modulation formats such as GMSK used in GSM systems the I Q phase and frequency errors are more representative measures of the quality of the signal than EVM As with EVM the analyzer samples the transmitter output in order to capture the actual phase trajectory This is then demodulated and the ideal or reference phase trajectory is mathematically derived The phase error is determined by comparing the actual and reference signals The mean gradient of the phase error signal is the frequency error The short term variation of this signal is defined as phase error and is expressed in terms of rms and peak see Fig
37. efficients and incorrect windowing Baseband A IF Filter Amplifier Baseband Filter Power Control RF LO Baseband filtering must be correctly implemented to provide the right baseband frequency response and avoid intersymbol interference and overshoot of the baseband signal In systems using Nyquist filters the roll off parameter alpha defines the sharpness of the filter in the frequency domain As shown in Figure 42 the lower the alpha the sharper the filter in the frequency domain and the higher the overshoot in the time domain It is important to verify that the transmitter has the appropriate baseband frequency response for the alpha specified 1 0 0 2 0 4 0 6 fi F symbol 40 Incorrect windowing distorts the frequency In many communications systems using Nyquist baseband filtering the filter response is shared between the transmitter and the receiver The filters must be compatible and correctly implemented in each The type of filter and the roll off factor alpha are the key parameters that must be considered The main causes of error in baseband filtering are the following 1 Wrong filter coefficients For Nyquist filters an error in the implemen tation of alpha may result in undesirable amplitude overshoot in the signal or interference in the adjacent frequency channel It may also degrade intersymbol interference IST caused by fading 2 Incorrect windowing of the transmi
38. enna port However substandard performance may be caused by various parts of the system so troubleshooting is usually done at several points in the transmitter The source of impairments can be difficult to determine This difficulty is magnified by these practicalities e Part of the transmitter is generally implemented digitally e Some parts of the transmitter may not be accessible e It may be unclear whether a problem is rooted in the analog or digital section of the system The ability to look at the signal and deduce the source of a problem is very important to successful design The ideal troubleshooting instrument has the flexibility and measurement capabilities to help you infer problem causes from measurements at the RF IF and baseband sections of the transmitter The measurements described in this chapter are performed at the antenna port assuming that other parts of the transmitter are not easily accessed The objective is to help you recognize and troubleshoot the most common impairments from measurements performed at the antenna port To assist you in this task the following information is included in this chapter e A general troubleshooting procedure see Appendix A for a more detailed procedure e A table that links measurement problems to their possible causes in the different sections of the transmitter e A description of the most common impairments and an explanation of how to verify each one of them The foll
39. ents 05 Detailed Troubleshooting Procedure Instrument Capabilities 6 References o eae lat iateless 7 Related Literature ccc ccc ce eee cee oro rro rooo 1 Wireless Digital Communications Sytems 1 1 Digital communications transmitter Figure 1 Block diagram of a digital communications transmitter Digital Data Channel Coding Interleaving The performance of a wireless communications system depends on the transmitter the receiver and the air interface over which the communications take place This chapter shows how a digital communications transmitter works and discusses the most common variations of transmitters Finally it briefly describes how the complementary digital communications receiver works Figure 1 shows a simplified block diagram of a digital communications transmitter that uses I Q modulation I Q modulators are commonly used in high performance transmitters This application note focuses on the highlighted section of Figure 1 Measurements and common impairments for this section of the transmitter will be described in the following chapters Previous stages in the transmitter include speech coding assuming voice transmission channel coding and interleaving Speech coding quantizes the analog signal and converts it into digital data It also applies compression to minimize the data rate and increase spectral efficien
40. er measurements such as channel power the accuracy of the measurement is limited by the absolute amplitude accuracy of the instrument In the case of relative power measurements such as ACPR the accuracy is limited by the relative amplitude accuracy and dynamic range of the instrument As a rule of thumb the noise floor or distortion of the instrument should be at least 10 dB below the distortion of the signal being measured Since the signal is noise like averaging the power over several measurements is extremely important for more repeatable power measurements 8 In the case of timing measurements the accuracy of the measurement is mainly limited by the time accuracy time resolution and amplitude linearity of the instrument Since there are a number of parameters to measure the use of masks and pass fail messages makes it simpler to ensure that all the timing parameters meet their specifications The accuracy of modulation quality measurements is mainly limited by the accuracy of the test instrument which is usually given as a percentage Typically the test equipment should be ten times more accurate than the specified limit so measurement results can be attributed to the unit under test UUT and not to the measuring instrument 27 3 Troubleshooting Transmitter Designs 3 1 Troubleshooting procedure Transmitter designs are tested to ensure conformance with a particular standard and are typically performed at the ant
41. er of the signal is higher than a level 9 455dB above the average 15 TRACE A Chi Main Tine 5 dBVpk N n n LogMag Figure 11 i HT i i Peak to average i A MI hn a power ratio statistics F h 1 Jl TA LL BE EUNI ll aeva Peak Hvg 9 455 dB Start Os Stop 38 1696428571 us The power statistics of the signal can be completely characterized by performing several of these measurements and displaying the results in a graph known as the Complementary Cumulative Distribution Function CCDF The CCDF curve shows the probability that the power is equal to or above a certain peak to average ratio for different probabilities and peak to average ratios The higher the peak to average power ratio the lower the probability of reaching it TRACE B Chi CCDF 100 5 P p 3 3 Cnt 736k Avg 11 088 dBm 3 Figure 12 E 3 Code Channel Signal CCDF curves Probability m 0 001 Start 0 dB Stop 20 dB dB Above Average gt 16 The statistics of the signal determine the headroom required in amplifiers and other components Signals with different peak to average statistics can stress the components in a transmitter in different ways causing different levels of distortion CCDF measurements can be performed at different points in the
42. errors deteriorate EVM I Q swapped results in an inverted spectrum However because of the noise like shape of digitally modulated signals the inversion is usually undetectable in the frequency domain In the modulation domain the data mapping is inverted as seen in Figure 35 but the error cannot be detected unless a known sequence is measured In CDMA signals I Q swapping errors can be detected by looking at the code domain power display Since these errors result in an incorrect transmitted symbol sequence the measuring instrument can no longer find correlation to the codes This causes an unlock condition in which the correlated power is randomly distributed among all code channels as shown in Figure 36 Some vector signal analyzers have an inverted frequency mode that accounts for this error and allows you verify it Figure 36 a Normal versus b unlock condition for code domain power power randomly distributed among all code channels Figure 37 Increase in code domain power noise Floor right versus Left 36 Base Ch Freq 1 80000 GHz PN Ofs x 64 chipsl Code Domain IS 95A Averages 10 Ref 0 00 dB 5 00 dB Sync 63 Act Set Th 20 00 dB Time Ofs 12896 7 us Pilot Freq Error 0 5 Hz Paging Carrier FT 28 8 dB Sync 4 2 dB Avg RT 8 7 dB 4 5 dB Max IT 48 4 dB 10 5 dB Avg IT 50 2 dB Base Ch Freq 1 80000 GHz NEC Code Domain IS 95A Averages 1
43. esponse is applied to the measurement for both base and mobile stations 4 Spectral splatter is a term often associated with the ACP due to transients Spectral splatter can be caused by fast burst turn on and turn off clipping saturation and Digital Signal Processor DSP glitches or other errors due to scaling High spectral splatter may occasionally be caused by phase transients Since transients are very short events time capture can be useful to locate and analyze them Spectral splatter can also be analyzed using a spectrogram which displays spectrum versus time as shown in Figure 74 section 3 2 11 Figure 23 In band spurious measurement 2 4 Out of band measurements 25 For cdmaOne systems the ACPR is not defined in the standard but it is often used in practice to test the specified in band spurious emissions 2 Spectral regrowth is a measure of how much the power in the adjacent channel grows how much worse it gets for a specific increment of the transmitted channel power 2 3 2 2 Spurious Spurious signals can be caused by different combinations of signals in the transmitter The spurious emissions from the transmitter that fall within the system s band should be below the level specified by the standard to guarantee minimum interference with other frequency channels in the system see Figure 23 21 B Base Ch Freq 1 93125 GHz Spur Close fverages 5 pass Ref 15 26 dBm Spectrum 10 001
44. f signal a with incorrect dB fdiv interpolation T T 110 dB L 1 Center 850 MHz Span 2 MHz CCDF curves Incorrect interpolation may distort the signal and therefore its statistics may change ACP Incorrect interpolation may cause an increase in the ACP Code domain power Since EVM is degraded the code domain power noise floor may increase 3 2 6 IF filter tilt or ripple 1 0 0 1 Baseband Up ur j Filters Amplifier Figure 53 IF filter gt Power Control IF LO RF LO The IF filter eliminates out of channel interference created during the I Q modulation Error characteristics in the design of this filter can affect the resulting signal Ideally the filter should be flat across the frequency band of interest and its group delay should be constant across the same frequency band Common IF filter impairments include filter tilt or ripple in the frequency response and variations of group delay For instance SAW Surface Acoustic Wave filters may have internal reflections that cause long delays which produce fine grain ripple in the frequency response Filter tilt or ripple in the frequency response causes linear distortion in the signal Figure 54 Mismatch between IF or RF components may cause tilt or ripple in the frequency response 47 An effect equivalent to filter tilt or ripple is often caused by improper matching of any component from the IF filter to the antenna For example
45. filter is independent of the filter coefficient alpha However incorrect windowing may cause a dramatic change of the spectrum that may affect the 3dB bandwidth e CCDF curves The roll off factor alpha affects the amount of overshoot Therefore a filter with the wrong parameter may affect the statistics of the signal e ACP Incorrect filtering may affect the degree of interference in the adjacent channel The windowing applied in time also affects the ACP The shorter the time window for a particular alpha the worse the ACPR More abrupt time windows also cause a larger ACPR e Code domain power Distortion caused by mismatched filters can result in an increase in the code domain noise floor Frequency response The frequency response of the baseband filter affects the total frequency response of the transmitter Therefore the effect of wrong filter coefficients or incorrect windowing can be analyzed by applying equalization and examining the transmitter s bits to RF frequency response as seen in Figure 45a 42 Equalization minimizes the errors caused by baseband filtering impairments Figure 45b shows how equalization improves magnitude of the error vector versus time compare to Figure 44b TRACE A Chi PI 4 EQ Chan A Marker 849 952 539 062 5 Hz 0 003 dB 1 05 dB T EQ Figure 45 a Frequency response for incorrect alpha b Magnitude of the Center 850
46. gnals are fed into the I Q modulator The local oscillator LO in the modulator can operate at an intermediate frequency IF or directly at the final radio frequency RF The output of the modulator is the combination of the two orthogonal I and Q signals at IF or RF After modulation the signal is upconverted to RF if needed The RF signal is often combined with other signals other channels before being applied to the output amplifier The amplifier must be appropriate for the signal type 1 1 1 Analog I Q modulator versus digital IF Although digital communications transmitters can be designed using analog hardware there is a clear trend toward implementing part of the system digitally The section of the system designed digitally can vary from radio to radio Therefore the location of the Digital to Analog Converter DAC varies For instance while baseband filters are usually implemented digitally as FIR finite impulse response filters the I Q modulator has traditionally been designed using analog hardware In this case two DACs one for each path are applied prior to the I Q modulator However there is an increasing tendency to implement the I Q modulator digitally digital IF which leads to more stable results In this case a DAC is located at the IF In any case the signal is analog at RF 1 Baseband 1 0 Filters Modulator IF Filter Upconverter Amplifier I I I I I l Symbol i tu
47. gy to an LO Phase noise in any LO in the transmitter can also cause noise in the phase of the recovered I Q signal Phase noise at far offsets in the RF LO may be fed through and cause energy to be spilled into the adjacent channel 1 See Glossary Figure 57 Phase error versus time for a signal with hum Figure 58 Constellation display degraded by phase noise 49 How can you verify LO instability A large phase error relative to the magnitude error gives the best indication of instability in an LO The error can be further analyzed by examining the phase error versus time display This display shows the modulating waveform of any residual or interfering PM signal Random phase errors indicate phase noise while sinewave shapes or periodic waveforms indicate interfering PM tones For example Figure 57 shows the phase error versus time for a signal with a symbol rate of 50 kHz Since the phase error versus time shows two cycles over 2000 symbols the frequency of the interfering PM signal is 2x 50x103 2000 50 Hz TRACE A Chi QPSK Phs Error 25 deg Wr Phs deg div 25 Start O syn Stop 1 999 ksyn If the LOs in the transmitter are accessible the problem can be tracked down to a specific LO by making analog PM measurements on the different LOs of the transmitter Bad cases of LO interference cause a constellation display like that sh
48. he results If measurements in front of the amplifier are not possible you can lower the amplitude of the transmitted signal and compare measurement results In the case of ACPR if the peak amplitude of the transmitted signal drives the amplifier into compression distortion occurs and the distortion in adjacent frequency channels is larger than expected Therefore the measured ACPR is smaller 30 With Compression Without Compression TRACE B Chi Spectrun 20 dBn Figure 26 LogMag ACP increases when compression occurs 10 dB div 120 dBn Center 850 MHz Span 5 MHz In the case of peak to average ratio and CCDF statistics if compression occurs the peak levels of the transmitted signal are clipped Clipping causes a peak to average ratio reduction Therefore the CCDF curve shows lower probabilities of reaching large peak to average ratios that is the peak to average ratio is smaller for a certain probability as shown in Figure 27 TRACE C D3 CCDF 100 F F Cnt 130k Avg 21 062 dBm Cnt 206k Avg 4 45 dBm Figure 27 CCDF curves for signal with and without compression Probability Compressed M 0 001 Start 0 dB L J Sto
49. ina 4 bit processor the MSB Most Significant Bit would be lost and the result would be 0010 To avoid this error DSPs can operate in a special mode called saturation In that case the DSP saturates to all ones if the result is greater than the largest possible value and to all zeros if the result is less than the smallest acceptable value How can you verify errors caused by internal overflow If the DSP s algorithms yield values greater than its design allows artifacts can occur in the modulation For example as we saw above the DSP can roll over signal values that are higher than its design allows to voltage levels that are very low This error is usually confined to a single output which causes spikes in the magnitude of the error vector versus time display as shown in Figure 71b IKHLE H Chl 164AM Meas Tine 1 5 EQ Const o 300 mre a M alt p fdiv ey e 9 o 1 5 4 4444442987 4 44444429874 TRACE B Chl 16QAM Err Y Tin B Marker 7 750 000 syn 711 19 mA 40 EQ l Liniag E um z T b 4 i 3 1 aiv I M E start O Sym Stop SYY syr Internal overflow may also affect the following measurements ACP Internal overflow may cause glitches in the signal which cause spectral splatter and therefore an increase in ACP Time capture and a spectrogram function are useful tools for analyzing transients and spectrum splatte
50. lter Removes Unwanted Frequency Components and Compensates for sin x x How can you verify non compensated sin x x As with incorrect interpolation not compensating for the sin x x function degrades the modulation quality Since it is a linear error it can be removed by equalization Therefore this kind of error can be easily detected by applying equalization and examining the transmitter s bits to RF frequency response In the case of a transmitter with an analog I Q modulator the frequency response shows a sin x x shape similar to the frequency response for incorrect interpolation Figure 51 In the case of a transmitter with a digital IF the frequency response shows a tilt similar to the frequency response for IF filter tilt Figure 55 With equalization the constellation and EVM improve as for incorrect interpolation see Figure 50 Figure 71 Effects of internal overflow on a the constellation and b the magnitude of the error vector versus time display 57 As in the case of incorrect interpolation non compensated sin x x may also cause an impact on channel bandwidth CCDF curves ACP and code domain power Internal overflow There are several scaling errors associated with digital hardware For instance in a DSP when a mathematical function produces an output that is greater than the largest possible value the result is incorrect For example if we add 0101 to 1101 the result should be 10010 but
51. m measurements such as frequency response group delay or distortion measurements However complex digitally modulated stimulus signals are increasingly used as they may provide more realistic measurement results Stimulus Ideal Source T M l M Baseband 1 0 I Filters T Modulator _IF Filter Amplifier Power Control RF LO gt as a a e Measurements Ideal Receiver Sometimes individual blocks or components cannot be isolated and the measurement can only be made at the final stage of the transmitter Therefore you may be forced to infer the causes of problems from measurements at the antenna port The ideal testing tool is not only able to perform the measurements but also has the flexibility to provide insight about system impairments by analysis of the transmitted signal This application note focuses on transmitter measurements and troubleshooting techniques performed at the antenna port although in practice some of these measurements can also be made at other locations in the transmitter For instance signal quality measurements can be performed on the RF IF or baseband sections of the transmitter 2 2 Measurement Domains Figure 6 Time and frequency domains 10 The transmitted signal can be viewed in different domains The time frequency and modulation domains provide information on different parameters of a signal The ideal test instrument can make measuremen
52. mulus signal BW Bae See Intp du eet e ES Bandwidth CCDE eser Complementary Cumulative Distribution Function CDMA SE ik ab te es bue edi Code Division Multiple Access CAMA anaana n aaaea aa IS 95 standard based CDMA system cdma2000 s sess cdmaOne derivative 3G system CMOS sss Complementary Metal Oxide Semiconductor DAC al oen eene tee e weis Digital to Analog Converter DEGT 42000 dado lt ld e Mus Digital Enhanced Cordless Telephone DSP eee Digital Signal Processor or Digital Signal Processing DVB 6 RAV a ehe Xe Digital Video Broadcast Cable EVM Du SERM e epp ems Error Vector Magnitude NA A DEINDE LTDA ee A Fast Fourier Transform FIR ree SRY Leon Dae dedos Finite Impulse Response EM vieni eMe e Ue d et Mn Frequency Modulation frame obese e In TDMA systems a repeating time interval divided among multiple users see timeslot GSMs nuni PST Global System for Mobile communications GMS Kets 23 205 iiil Heeg E Canes Be eas S Gaussian Minimum Shift Keying IE 2s E ee Ee eiue Intermediate Frequency IQ soo funt Rub elsi RU Doa Hi sc AMAS In phase Quadrature A aay es verat Intersymbol Interference LMDS 23 25 00 ae eat Ba Local Multipoint Distribution System A A as eee es ti eg a arse et Local Oscillator MSB eee eng o La di e e a inv Most Significant Bit NADG e252 Se cia o ie North American Digital Cellular system OQPSK Offset Quadrature Phase Shift Keying ORF Se ein CIN O
53. no other linear errors the frequency response should look similar to the sin x x function that the equalizer is compensating for as shown in Figure 51 Incorrect interpolation can also affect the following measurements 45 Figure 50 Effect of equalization on signal with incorrect interpolation polar diagram and magnitude of the error vector versus time a without equalization and b with equalization Figure 51 Bits to RF frequency response of transmitter with incorrect interpolation TRACE B Chi QPSK Err Tine 1 5 b 3L llb dll div tt HEH He xo Left 40 syn TRACE A Chi QPSK EG Chan Mea TW eet Right 70 syn 6 dB EQ LogMaa div 4 B Center 1 8 GHz Span 3 90625 MHz e Channel bandwidth Incorrect interpolation has an impact on the spectrum shape For instance for a transmitter with analog I Q the spectrum of a signal with incorrect interpolation has a rounded top instead of a flat one Therefore the channel bandwidth measurement may be affected For example Figure 52 shows a 3 dB bandwidth of 665 kHz for a signal with a symbol rate of 1 MHz 46 TRACE A Chl Spectrun A Offset 665 000 Hz 0 044 dB dBn LogMag Figure 52 Spectrum o
54. ode channels It is also important to look at the code domain power levels of the inactive channels which can indicate specific problems in the transmitter as we will see in the last chapter of this application note For instance unwanted in channel spurs raise the code domain noise level Compression can cause mixing of active code channels to produce energy in particular inactive channels 2 3 2 Out of channel measurements In band out of channel measurements are those that measure distortion and interference within the system band but outside of the transmitting frequency channel 2 3 2 1 Adjacent Channel Power Ratio ACPR Whatever the technology used or standard followed ACP measurements are required to ensure that the transmitter is not interfering with adjacent and alternate channels The Adjacent Channel Power Ratio ACPR is usually defined as the ratio of the average power in the adjacent frequency channel to the average power in the transmitted frequency channel For instance in Figure 22 the ACPR across a bandwidth of 1MHz for both the transmitted and adjacent channels is 261 87 dB for the lower adjacent channel and 61 98 dB for the upper adjacent channel The ACPR is often measured at multiple offsets adjacent and alternate channels Figure 22 ACPR measurement 24 Base Ch Freq 1 90000 GHz ACPR fverages 20 PASS Ref 11 38 dBm Spectrum Total Pwr Ref 10 00 dB MaxP 10 3 Center 1 00000 GHz
55. odulated Signals Pete Watridge Wireless Symposium 1999 3 4 5 7 Adaptive Equalization and Modulation Quality Bob Cutler Hewlett Packard Wireless R amp D Seminar 1997 8 Spectrum Analyzer Measurements and Noise HP Application Note 1308 literature number 5966 4008E 7 Related Literature 1 Using Vector Modulation Analysis in the Integration Troubleshooting and Design of Digital RF Communications Systems HP Product Note 89400 8 literature number 5091 8687E 2 Using Error Vector Magnitude Measurements to Analyze and Troubleshoot Vector Modulated Signals HP Product Note 89400 14 literature number 5965 2898E 3 HP 89400 Series Vector Signal Analyzers literature number 5965 8554E 4 HP VSA Series Transmitter Tester literature number 5966 4762E 5 HP 8560 E Series Spectrum Analyzers literature number 5966 3559E 6 HP 8590 E Series Spectrum Analyzers literature number 5963 6908E 7 HP ESA E Series Spectrum Analyzers literature number 5968 3278E DU HEWLETT PACKARD For more information about Hewlett Packard test and measurement products applications and services and for a current sales office listing visit our web site http www tmo hp com You can also contact one of the following centers and ask for a test and measurement sales representative United States Hewlett Packard Company Test and Measurement Call Center P O Box 4026 Englewood CO 801
56. orms To accurately interpret the symbols and recover the digital data at the receiver it is imperative that the transmitter and the receiver have the same symbol rate The symbol clock in the transmitter must be set correctly Symbol rate errors often occur from using the wrong crystal frequency for example if two numbers are swapped in the frequency specification Slight errors in the clock frequency impair the signal slightly but as the frequency error increases the signal can become unusable Therefore it is important to be able to verify errors in the symbol timing How can you verify errors in the symbol rate The effect of symbol rate errors on the different measurements depends on the magnitude of the error If the error is large enough the instrument cannot demodulate the signal correctly and modulation quality measurements are meaningless For instance for a QPSK system with a specified symbol rate of 1 MHz an error of 10 kHz actual symbol rate of 1 010 MHz can cause an unlock condition when looking at the constellation and measuring EVM For a W CDMA system with a specified symbol rate of 4 096MHz an error of 200Hz actual symbol rate of 4 0962 causes an unlock condition in the code domain power measurement Therefore the methods to verify small symbol errors those that do not cause an unlock condition and large symbol errors those that create an unlock condition are different Small errors The best way to verify
57. oubleshoot instability at a particular LO as described in the next chapter 12 2 3 In band measurements 1 See Glossary for definition The measurements required to test digital communications transmitters can be classified as in band and out of band measurements regardless of the technology used and the standard followed In band measurements are measurements performed within the frequency band allocated for the system for example 890 MHz to 960 MHz for GSM In band measurements can be further divided into in channel and out of channel measurements 2 3 1 In channel measurements The definition of channel in digital communications systems depends on the specific technology used Apart from multiplexing in frequency and space geography the common cellular digital communications technologies use either time or code multiplexing In TDMA technologies a channel is defined by a specific frequency and timeslot number in a repeating frame while in CDMA technologies a channel is defined by a specific frequency and code The terms in channel and out of channel refer only to the specific frequency band of interest frequency channel and not to the specific timeslot or code channel within that frequency band 2 3 1 1 Channel bandwidth When testing a transmitter it is usually a good idea to first look at the spectrum of the transmitted signal The spectrum shape can reveal major errors in the design For a transmitter with a
58. owing is a suggested troubleshooting procedure to follow if the transmitter design does not meet the specifications 1 Look at the signal in the frequency domain and verify that its spectrum appears as expected Ensure that its center frequency and bandwidth are correct 2 Perform in band and out of band power measurements channel power ACP check CCDF curve spurious and harmonics 3 In the case of bursted signals perform timing measurements 4 Look at the constellation of the baseband signal 5 Examine error metrics EVM I Q offsets phase error frequency error magnitude error and rho 6 If the phase error is significantly larger than the magnitude error examine I Q phase error versus time Perform phase noise measurements on LOs if accessible 7 If phase error and magnitude error are comparable examine magnitude of the error vector versus time and error vector spectrum 8 Turn the equalizer on and verify that it reduces modulation quality errors and check frequency response and group delay of the transmitter for faulty baseband or IF filtering or other linear distortion problems 28 In these measurements variations from the expected results will help you locate faults in different parts of the transmitter The following sections describe the most common impairments and how to recognize them from their effects on the different measurements Refer to Appendix A for a more detailed troubleshooting procedure Tabl
59. own in Figure 58 The measured symbols preserve the right amplitude but vary in phase around the ideal symbol reference point TRACE A Chi MSK1 Meas Tine 1 5 Const A N 300 M div 1 9607843757 1 96078437557 50 LO instability also impacts the following measurements e Adjacent Channel Power As indicated earlier residual PM or phase noise at far offsets in the RF LO can degrade the ACP see Figure 59 TRACE A Chi Spectrun oe Marker 849 900 000 Hz 84 038 dBH dBn LogMag Figure 59 ACP degraded by High Phase Noise phase noise at ue at Far Offsets far offsets fdiv 190 Low Phase Noise D dr d E Center 850 MHz Span 200 kHz Code domain power LO instability appears in the code domain power display as an increase in the noise level in certain non active code channels The code channels affected depend on the orthogonal coding scheme of the particular system and on the code channel configuration of the signal Figure 60 shows the code domain power display for a signal with phase noise Base Ch Freq 1 80000 GHz PN OfsB x 64 chips Base Ch Freq 1 80000 GHz PN Ofs x 64Ichips Code Domain IS 95A Averages 18 Code Domain 15 95 Awerages 4 Ref 0 00 dB Ref 0 00 dB 5 00 dB Figure 60 Code domain power Sync Esec for signal a without phase noise and Walsh Channel 6
60. p 20 dB dB above Average 31 Figure 28 Polar diagram of compressed signal compared to ideal trajectory Base Ch Freq 1 80000 GHz PN Ofs x 64 chipsl Code Domain Compression may also be detected in other measurements Polar diagram If the high peak levels of the transmitted signal are clipped the signal has a lower overshoot This effect can be seen by comparing the trajectory of the compressed signal to the ideal trajectory in the polar diagram as in Figure 28 Filtering at the receiver causes dispersion in time In practice compression often causes an error in the symbol s after a peak excursion of the signal Therefore EVM may be affected TRACE A Chi QPSK Meas Tine 1 5 Tu Reference A BN Ideal Trajectory 1 0 gt Actual Trajectory 300 Hn div m ad 1 5 2 0270270705 2 02702707052 Code domain power Non linearity in the amplifier also causes an increase in the code domain noise level in CDMA systems Compression causes code domain mixing Therefore energy appears in the non active channels in deterministic ways For instance in Figure 29 for a cdmaOne signal Walsh code 1 mixes with Walsh codes 12 and 32 causing energy to show up on Walsh codes 13 and 33 Also Walsh code 12 mixes with Walsh code 32 to create power on Walsh code 44 Base Ch Freq 1 80000 GHz Code Domain IS 95R PN Ofs x 64 chips fverases 10 o7 IS 95R
61. ponse can be displayed and measured as magnitude phase and group delay Ideally the magnitude of the frequency response should be flat across the frequency band of interest and its phase should be linear over that same frequency band Group delay is a more useful measure of phase distortion It is defined as the derivative of the phase response versus frequency dp dw that is the slope of the phase response If the transmitter does not introduce distortion its phase response is linear and group delay is constant Deviations from constant group delay indicate distortion Figure 19 a Magnitude of the bits to RF frequency response should be flat across the frequency band of interest indicated by b 3 dB bandwidth on signal spectrum Figure 20 Rho 22 EN Unflatness Indicates Linear Distortion Problems 0 3 sum ap Sun 48 427 d6 Center 1 8 GHz q TRACE B Chi Spectrum I 3 dB BW Span 12 5 MHz Or dB LogMag b e 199 Poner 11 893 dBm T Center 1 8 GHz Span 12 5 MHz 2 3 1 7 5 Rho CDMA systems use p rho as one of the modulation quality metrics Rho is measured on signals with a single code channel It is the ratio of correlated power to total power transmitted see Figure 20 The correlated power is computed by removing frequency phase and time offsets and performing a cross correlation between the corrected measu
62. ps Res BH 500 000 kHz Samples 3501 points 200 00 ns Current Data Max Pt 1 02 dBm Min Pt 110 66 dBm Mean Transmit Power 1 19 dBm Burst Width 561 600 ps The best way to troubleshoot burst shaping impairments that affect ACP such as short rise and fall times and frequency drift is by combining the time and the frequency domain Time capture and a spectrogram function that shows how the apparent frequency spectrum varies with time are the ideal tools for that The spectrogram can be used to view a three dimensional picture of what is happening Frequency drifts and high adjacent channel interference due to sharp edges can be easily detected as seen in Figure 74 TRACE A Chi Spectrun Marker o 850 000 000 Hz 29 355 dB High ACP ACP Spectral Splatter Caused by Rise and Fall Times Center 850 zu Span 100 kHz 60 Figure 75 Magnitude error versus time for a signal with amplitude droop 4 Summary The modulation quality of the signal may be affected by most burst shaping errors such as amplitude droop overshoot on power up frequency drift and too short a burst width Those errors have an impact on EVM and the related displays Amplitude droop for instance may be easily detected by looking at the magnitude error versus time display as shown in Figure 75 The measurement algorithm may compensate for amplitude droop In that case the error is given in a separate metric TRACE H Chl PI 4 Mag Er
63. r Code domain power Since EVM is affected internal overflow errors may also increase the code domain power noise floor Equalization does not remove errors caused by internal overflow since these are not linear Figure 72 Burst modulator 58 3 2 11 Burst shaping impairments Baseband 1 Q Filters Modulator IF Filter p 1 Amplifier I f Power Control Burst Modulator Burst Control I I I I I I I I IF LO RF LO I I In TDMA systems the RF power is bursted on and off in well defined time slots such that the frequency channel can be shared with other users It must be ensured that the burst parameters are accurate These parameters include the rise and fall times as well as the burst width The power must be on long enough to transmit the timeslot data During the off time the power must be low enough to be considered off Other potential problems include overshoot on power up frequency drift and amplitude droop In TDMA systems the transmitter output is turned on and off many times per second Instability in frequency and power as the burst is turned on can seriously impair the operation of the system Overshoot on power up can also result in compression at the amplifier which raises the interference in the adjacent frequency channels Frequency drift also causes higher interference levels in the adjacent frequency channels Temperature changes in the components of the burst modulator Fig
64. r 29 3 2 1 Compression Other Channel 7771 Other Channel 1 0 0 1 Baseband 1 Q Filters Modulator IF Filter Other Channel ud Other Channel The power amplifier PA is the final stage prior to transmission Key characteristics of the PA are frequency and amplitude response 1dB compression point and distortion The PA selected must be appropriate for the signal type To avoid compression of the signal the input levels and output section gains in the amplifier must be tightly controlled Compression occurs when the instantaneous power levels are too high driving the amplifier into saturation For instance if the signal peak power is not properly taken into account signal compression can occur This issue is particularly relevant to CDMA systems because the peak to average ratio of the multi code signal changes depending on the channel configuration Mobile station transmitters that use constant amplitude modulation schemes like GSM mobile station transmitters which only carry information on the phase of the signal are more efficient when slightly saturated But in other digitally modulated systems compression causes clipping and distortion which may result in a loss of signal transmission efficiency and cause interference with other channels How can you verify compression The best way to verify that the signal is compressed is to make ACPR and CCDF measurements before and after the amplifier and compare t
65. red signal and the ideal reference If some of the transmitted energy does not correlate this excess power appears as added noise that may interfere with other users on the system A P Power that correlates with ideal Total Power Signal Power Signal Power Error Power The rho measurement indicates the overall modulation performance level of a CDMA transmitter when transmitting a single channel Since uncorrelated power appears as interference poor rho performance affects the capacity of the cell 2 2 3 1 7 6 Code domain power In CDMA systems a signal with multiple code channels can be analyzed in the code domain To analyze the composite waveform each channel is decoded using a code correlation algorithm This algorithm determines the correlation coefficient factor for each code Once the channels are decoded the power in each code channel is determined 2 23 Base Ch Freq 850 000 MHz PN 0fs0 x 64 chips Code Domain IS 95A Averages 7 Figure 21 Code domain power measurement Ref 0 90 dB m ae Walsh Channel Act Set Th 20 00 dB Time Ofs 11950 8 us Pilot Freq Error 0 7 Hz Paging Carrier FT 32 1 dB Sync 7 0 dB Avg AT 7 3 dB Max IT 13 3 dB Avg IT 10 6 dB 48 4 dB 50 4 dB Measuring code domain power as shown in Figure 21 is essential for verifying that the base station is transmitting the correct power in each of the c
66. root raised cosine filter the 3dB bandwidth of the modulated frequency channel should approximate the symbol rate For instance in Figure 8 for a symbol rate of 1 MHz the measured 3dB bandwidth is 1 010 MHz Therefore this measurement can be used to determine gross errors in symbol rate 2 3 1 2 Carrier frequency Frequency errors can result in interference in the adjacent frequency channels They can also cause problems in the carrier recovery process of the receiver The designer must ensure that the transmitter is operating on the correct frequency The carrier frequency should be located in the center of the spectrum for most modulation formats It can be approximated by calculating the center of the 3dB bandwidth For instance in Figure 8 the measured carrier frequency is 850 MHz 13 TRACE B Chi Spectrum A Offset 1 010 000 Hz 0 056 dB B Marker 850 000 000 Hz 26 329 dBn 0 dBn LogMag Figure 8 Carrier frequency and channel bandwidth 10 measurements fdiv 100 dBn Center 850 MHz Span 2 MHz Other common methods for determining the carrier frequency are Measure an unmodulated carrier with a frequency counter Calculate the centroid of the occupied bandwidth measurement see section 2 3 1 4 When performing an occupied bandwidth measurement the testing instrument usually gives an indication of the frequency carrier error as shown in
67. ror R Marker 0 0000 syn 1 4058 x T gt x Keal start O Sym Stop 155 SyH Burst shaping errors are not linear and cannot be removed by applying equalization Conformance measurements are performed to verify that a digital transmitter design meets system requirements If the transmitter does not comply with the specifications the problem must be tracked down to the impaired device or section This application note has presented the most common transmitter measurements and a general methodology for troubleshooting transmitter problems A link between the different transmitter measurements and the most common impairments has also been established With these tools and adequate measurement equipment you will be able to quickly recognize and verify problems in your design 61 Appendix A Detailed 4 1 Incorrect 3dB Bandwidth Troubleshooting Channel Correct Signal Shape Symbol Procedure Bandwidth Rate Error Otherwise CCDF Curves Different Before and After Amplifier Figure 76 Compression Troubleshooting rocedure f P High ACP at incorrect Burst on and off Timing if TDMA System Measurement Otherwise Impairment Spurious and Harmonics Meets Harmonics Spurious Specs Burst Shaping Error Discrete Tones Out of Channel or Out of Band Interfering Tones Amplitude Variations During Burst EVM or Phas
68. se and group delay As noted above equalization compensates for certain signal imperfections in the transmitter transmission path or receiver Equalization removes only linear distortion Linear distortion occurs when the signal passes through one or more linear devices having transfer functions containing amplitude unflatness for example ripple and tilt and or group delay variations over the bandwidth of the signal There can be many sources of linear distortion in a system bandpass filters in the IF improper cable terminations improper baseband filtering non compensated sin x x antenna mismatch signal combiners and multipath signal effects From a modeling standpoint all of the linear distortion mechanisms can be combined and represented by a single transfer function H f When applying equalization the measuring instrument must counteract the effects of the linear distortion To achieve that an equalizer filter whose transfer function is 1 H f is applied over the bandwidth of the signal Once equalization has been applied the inverse transfer function of the equalizer which represents the linear distortion elements of the device under test can be displayed and measured If measured directly at the transmitter s output the inverse transfer function is basically the bits to RF frequency response of the transmitter or the variations from the ideal frequency response caused by non linear distortions 7 The actual frequency res
69. t the baseband IF or RF sections of the transmitter As shown in the last chapter of this application note the different modulation quality displays and tools can help reveal or troubleshoot various problems in the transmitter For instance the I Q constellation can be used to easily identify I Q gain imbalance errors Small symbol rate errors can be easily identified by looking at the magnitude of the error vector versus time display The error vector spectrum can help locate in channel spurious The value of EVM as an indicator of modulation quality can be enhanced by the use of equalization in the measuring instrument Equalization is commonly used in digital communications receivers Although its primary function is to reduce the effects of multipath it also compensates for certain signal imperfections generated in both the transmitter and receiver For this reason it is useful to have an equalizer in the measuring instrument An instrument with an equalizer will better emulate a receiver that is the impairments that the equalizer of the receiver removes are also removed by the measuring instrument Therefore the impairments that have little effect on system performance also minimally impact the measured EVM Figure 16 shows the magnitude of the error vector versus time with and without equalization With equalization the constellation looks much better and the magnitude of the error vector versus time is lower The signal has not chang
70. the IF is implemented digitally TRACE A Chi QPSK Meas Tine 1 5 Const e i l i 300 div 1 9507843757 1 96078437567 2 Quadrature errors If the phase shift between the IF or RF LO signals that mix with the I and Q baseband signal at the modulator is not 90 degrees a quadrature error occurs The constellation of the signal is distorted see Figure 32 which may cause error in the interpretation of the recovered symbols 33 TRACE A Chi QPSK Meas Tine 1 5 Const LN ae 300 div Figure 32 Quadrature error b 1 9607843757 1 96078437567 3 IQ offsets DC offsets may be introduced in the I and Q paths They may be added in the amplifiers in the I and Q paths For digital IF implementations offsets may also occur from rounding errors in the DSP See Figure 33 1 0 Measured Compl Constin Figure 33 1 0 offsets 4 Delays in the I or Q paths When the serial bit stream is encoded into symbols and the bits are split into parallel paths for creation of the I and Q signals it is important that these signals are properly aligned Problems in this process can cause unwanted delays between the I and Q signals Delays can be caused by the modulator or by the previous components in the I or Q paths for example the baseband filter or the DAC For instance if the baseband filters are analog variations in group delay between the filters
71. tion note includes some references to digital communications receivers it does not cover measurements and possible impairments of receivers For more information on digital communications receivers please refer to Testing and Troubleshooting Digital RF Communications Receiver Designs 5 Note The above application notes can be downloaded from the Web at the following URL and printed locally http www tmo hp com tmo Notes English index html Table of Contents 1 Wireless Digital Communications Systems 6 1 1 Digital communications transmitter 2005 6 1 1 1 Analog I Q modulator versus digital IF 7 1 1 2 Other implementations oooooooooooo o 8 1 2 Digital communications receiver 0 000000 8 2 Testing Transmitter Designs eee eere 9 2 1 Measurement model 1 0 ccc eee ees 9 2 2 Measurement domains 0 0000 00 cc eee 10 2 2 1 Tire domal 22s a edge EA 10 2 2 2 Frequency domain 00 000 cee eee 10 2 2 3 Modulation domain 0 000 0 eee eee 11 2 3 In band measurements 0 6 0 6 eee eee 12 2 3 1 In channel measurements 0000000 020s 12 2 3 1 1 Channel bandwidth o o ooo o oooooo oo 12 2 3 1 2 Carrier frequency elles 12 2 3 1 8 Channel power 0 0 0 00 cee eee 13 2 3 1 4 Occupied bandwidth o o oooooo o 14 2 3 1 5 Peak to average power ratio and CCDF curves 14 2
72. tions transmitters work 2 How to test transmitters and what test equipment characteristics are important 3 The common impairments of a transmitter and how to troubleshoot them The following information is also included as reference material e A detailed troubleshooting procedure Appendix A e A table of instrument capabilities Appendix B e A glossary of terms e A list of reference literature The first two chapters in the application note covering topics 1 and 2 are targeted at new R amp D engineers who have a basic knowledge of digital communications systems The third chapter is targeted at R amp D engineers with some experience in testing digital communications transmitter designs For basic information on digital modulation techniques essential background for this application note please refer to Digital Modulation in Communications Systems An Introduction Ml The measurements and problems described apply to most wireless communications systems Some measurements specific to common technologies or standards are also mentioned For more detailed information on CDMA and GSM measurements please refer to Understanding CDMA Measurements for Base Stations and Their Components 2 Understanding GSM Transmitter Measurements for Base Transceiver Stations and Mobile Stations 3l Understanding PDC and NADC Transmitter Measurements Jor Base Transceiver Stations and Mobile Stations A Although this applica
73. transmitter to examine the statistics of the signal and the impact of the different sections on those statistics These measurements can also be performed at the output of the transmitter to compare the statistics to an expected curve CCDF curves are also related to Adjacent Channel Power ACP measurements as we will see later Besides causing higher levels of distortion high peak to average ratios can cause cumulative damage in some components Performing CCDF measurements at different points of the transmitter can help you prevent this damage Peak to average power ratio and CCDF statistic measurements are particularly important in digitally modulated systems because the statistics may vary For instance in CDMA systems the statistics of the signal vary depending on how many code channels and which ones are present at the same time Figure 12 shows the CCDF curves for signals with different code channel configurations The more code channels transmitted the higher the probability of reaching a given peak to average ratio In systems that use constant amplitude modulation schemes such as GSM the peak to average ratio of the signal is relevant if the components for example the power amplifier must carry more than one carrier There is a clear trend toward using multicarrier power amplifiers in base station designs for most digital communications systems See 6 for more information on peak to average power ratio and CCDF curves
74. ts in all three domains Two types of transmitter test instruments are discussed the spectrum analyzer SA and the vector signal analyzer VSA Their measurement capabilities in each domain are described in the following sections of this chapter Refer to Appendix B for a list of spectrum analyzers and vector signal analyzers from Hewlett Packard and their capabilities for measuring and troubleshooting digital communications transmitters 2 2 1 Time domain Traditionally looking at an electrical signal meant using an oscilloscope to view the signal in the time domain However oscilloscopes do not band limit the input signal and have limited dynamic range Vector signal analyzers downconvert the signal to baseband and sample the I and Q components of the signal They can display the signal in various coordinate systems such as amplitude versus time phase versus time I or Q versus time and I Q polar Swept tuned spectrum analyzers can display the signal in the time domain as amplitude envelope of the RF signal versus time Their capability can sometimes be extended to measure I and Q Time domain analysis is especially important in TDMA technologies where the shape and timing of the burst must be measured 2 2 2 Frequency domain Although the time domain provides some information on the RF signal it does not give us the full picture The signal can be further analyzed by looking at its frequency components Figure 6 Both spectrum
75. tter filter Since the ideal frequency response of the Nyquist filter is finite the ideal time response impulse response is infinite However the baseband filter is usually implemented as a digital FIR filter which has a finite impulse response that is the actual time response is a truncated version of the ideal infinite response The filter must be designed so that it does not truncate the ideal response too abruptly Also the filter must include enough of the ideal impulse response to prevent excessive distortion of the frequency response For example Figure 43a shows the ideal time response and frequency response of a root raised cosine filter for an alpha of 0 2 Figure 43b shows the actual time response after a flat time window has been applied Since samples of significant value have been truncated the actual time response is significantly different from the ideal and the frequency response is distorted The time window applied by the actual filter must be appropriate for the specified alpha to avoid too much distortion of the frequency response In this case Figure 43 the window applied is too short in time which increases ACP in the frequency domain 0 5 0 4 Hj 0 2 0 0 02 04 06 08 fi F symbol Frequency Domain l 04 06 f Feymbol 0 8 Figure 44 a Polar diagram and b magnitude of the error vector versus time for incorrect alpha The error vector is large between symbol points
76. ure 18 BTS Ch Freq 935 200 MHz TSC Auto Phase amp Frequency GSM 900 Ref 0 00 Phase Err Ref 0 00 Phase Err w Freq 0 500 LT 1 00 Deg HA R filth WA AREAS m 500 0 mbit 147 5 bit 500 0 mbit 147 5 bit Phase Error 0 19 rms 0 52 pk at bit 143 3 21 07 Hz 61 7 dBc Maximum Freq Error IQ Offset Avg Type 64 40 ps 21 1 See Glossary for definition Significant phase errors can indicate problems in the baseband section of the transmitter The output amplifier in the transmitter can also create distortion that causes unacceptably high phase error for multicarrier signals Significant phase error at the beginning of a burst can indicate that a synthesizer is failing to settle quickly enough In a real system poor phase error reduces the ability of a receiver to correctly demodulate especially with marginal signal conditions This ultimately degrades sensitivity Frequency error is the difference between the specified carrier frequency and the actual carrier frequency A stable frequency error simply indicates that a slightly wrong carrier frequency is being used Unstable frequency errors can indicate short term instability in the LO improper filtering AM PM conversion in the amplifier or wrong modulation index if the transmitter is implemented using an analog frequency modulator See 3 for more information on phase and frequency error and other GSM measurements 2 3 1 7 4 Frequency respon
77. ure 67 RFLO DSP and DAC Transmitter with Analog I O Modulation IF Filter ARE A C ERRAT 1 Combined With 100 1 Baseband Q Reconstruction Filter Upconverter Amplifier Filters Modulator Symbol Q Q Implemented Digitally Power Control RF LO Transmitter with Digital IF Although it is possible to implement digital modulation using all analog hardware it is becoming more and more common to digitize the input signal immediately and stay with digital processing right through to the IF The IF is then converted to analog prior to upconverting to the final frequency for transmission The DACs can be placed at various points depending on the actual implementation as shown in Figure 67 Digital processing and conversion may be subject to problems The most common impairments are non compensated sin x x and internal overflow Figure 68 The DAC impulse response is a pulse which corresponds to a sin x x Function in the frequency domain 55 Non compensated sin x x Although the ideal output of the DAC is a series of delta impulses that represent the different amplitude levels in practice the impulses have a certain width x prior to the reconstruction filter that smoothes the signal A pulse in the time domain translates into a sin x x function in the frequency domain as shown in Figure 68 There must be compensation for the sin x x shape somewhere in the transmitter design Reconstruction
78. ure 72 may cause changes in the amplitude of the signal when the burst is on This artifact is known as amplitude droop and affects the modulation quality of the signal Rise and fall times that are too long affect adjacent timeslots The rise and fall times of the waveform are directly related to the spectrum splatter during the turn on and turn off of the burst This splatter affects the ACP The faster the rise and fall edges of the burst the more spectrum required and therefore the higher the interference in the adjacent frequency channel TDMA and frequency agile systems have much in common When the carrier hops from frequency to frequency the transmitter is still turning on and off many times per second The same burst issues apply Adjacent channel interference is still a concern It is therefore important to characterize the burst shape parameters and associate them with the measurement results to isolate the cause of the problem Figure 73 Burst rise and fall times too long Figure 74 Spectrogram showing high ACP caused by short rise time 59 How can you verify burst shaping impairments The best way to detect most burst shaping errors is by performing the appropriate timing measurements Figure 73 shows an example in which the rise and fall times are too long BTS Ch Freq 935 200 MHz TSC Auto Power vs Time GSM 900 FRIL Ref 10 00 dBm RF Envelope 149 80 ps 0 00
79. utput RF Spectrum PAA LA A A Power Amplifier PD e Se tied Pacific Digital Cellular system PHS a exea tds Personal Handyphone System lg MT Phase Modulation RES ae kn era due i Cela o de E s Radio Frequency SA it fox o meses dee Ve Ue ES gs Spectrum Analyzer SAW ra Logis cx eri meet chess E arenes ti due de Surface Acoustic Wave TDMA eec hr Time Division Multiple Access timeslot In TDMA systems amount of time per frame each mobile has to transmit information see frame i TM ELEME Vector Signal Analyzer Walsh code Orthogonal code used in cdmaOne systems for channelization in the forward link and as orthogonal modulation in the reverse link W CDMA scp i Aue etek as Wideband CDMA 3G system 6 References 1 Digital Modulation in Communications Systems HP Application Note 1298 literature number 5965 7160E 2 Understanding CDMA Measurements for Base Stations and Their Components HP Application Note 1311 literature number 5968 0953E Understanding GSM Transmitter Measurements for Base Transceiver Stations and Mobile Stations HP Application Note 1312 literature number 5968 2320E Understanding PDC and NADC Transmitter Measurements for Base Transceiver Stations and Mobile Stations HP Application Note 1324 literature number 5968 5537E Testing and Troubleshooting Digital RF Communications Receiver Designs HP Application Note 1314 literature number 5968 3579E 6 Power Statistics of Digitally M
80. y to verify AM PM is by isolating the amplifier and measuring its AM PM characteristics using a network analyzer or a signal analyzer with PM demodulation capability If the amplifier cannot be isolated the best way to verify AM PM conversion is by looking at the polar diagram Comparing the actual trajectory of the signal with its ideal trajectory for a few symbols shows higher errors for higher amplitude levels as seen in Figure 66 Because small errors at lower amplitudes may cause relatively large phase errors the correlation between higher amplitudes and large phase errors may not be obvious Filtering at the receiver causes dispersion in time Therefore AM PM conversion may cause an error in the symbol s after a peak excursion of the signal TRACE B Chl 16QAM Ref Tine 300 n div Signal with AM PM Conversion x Ideal Traj ectory 2 0270270705 2 02702707052 54 The code domain power measurement may also be affected by AM PM conversion Since the error is highly correlated with the signal the code domain energy may appear in the non active channels in deterministic ways AM PM conversion is not a linear error and cannot be removed by equalization 3 2 10 DSP and DAC impairments 100 1 Baseband Reconstruction 1 0 Filters Filter Modulator IF Filter Symbol Encoder Implemented Digitally Upconverter Amplifier Power Control Fig

Testing and Troubleshooting Digital RF Communications

Contents

Download Pdf Manuals

Related Search

Related Contents