DCU calibration saga precision required DCU ADC nonlinearity
DCU calibration saga • precision required • DCU ADC (non)linearity • FHIT measurements • expectations • uncertainties • DCUID <-> DETID • Individual versus generic calibration • Impact on monitoring of • Temperatures • Voltages • Leakage current FHIT page: http: //www. fynu. ucl. ac. be/themes/he/cms/activities/tracker/hybrids. html FHIT talk: https: //indico. cern. ch/get. File. py/access? contrib. Id=s 1 t 5&res. Id=0&material. Id=0&conf. Id=a 032140 26. 10. 2011 A. Kaminskiy 1
“Official” DCU calibration Nice summary from Franz 26. 10. 2011 A. Kaminskiy 2
DCU raw data distribution for various channels Produced by Franzzzz 26. 10. 2011 A. Kaminskiy 3
1 thermistor (TIB) DCU NTC conversion Temp C 26. 10. 2011 Gain: nominal and +- 5% ADCoffset=-50 ADCoffset=+50 A. Kaminskiy ADC counts 4
Temp C 26. 10. 2011 2 thermistors TOB, TEC? Gain: nominal and +- 5% ADCoffset=-50 ADCoffset=+50 A. Kaminskiy ADC counts 5
FHIT meaurement 1: ADCcount versus Input. Voltage at 3 different V 250 Voltages, using ITH connection About DCU : FHIT is a setup allowing to perform a DCU calibration. This calibration is possible through the FHIT internal switch matrix which enables the disconnecting of one DCU input (ITH). This input is the only one accessible on the connector. It is supposed to be connected to a CTN and is provided with a DCU internal current source (20 u. A). Through the resistor, the temperature is measured. Via FHIT, known voltages can be injected in this input. Reading DCU response provides the DCU calibration. Exampe from FHIT webpage <result name="DCU_cal_lowinput_m. V" value="9 19 29 39 49 58 68 78 88 98 108 117 127 137 147 "/><result name="DCU_cal_low_VNOM" value="13 17 22 33 50 68 85 103 123 142 161 182 201 221 242 "/><result name="DCU_cal_low_VMAX" value="10 13 19 28 44 61 79 98 117 135 155 175 195 215 236 "/><result name="DCU_cal_low_VMIN" value="19 23 30 45 61 79 97 116 135 154 174 194 214 234 254 "/><result name="DCU_cal_highinput_m. V" value="157 314 471 629 786 943 1101 1258 1415 1573 1730 1887 2045 "/><result name="DCU_cal_high_VNOM" value="263 595 927 1262 1595 1930 2267 2606 2937 3270 3605 3942 189 "/><result name="DCU_cal_high_VMAX" value="257 588 922 1255 1590 1923 2260 2599 2930 3263 3598 3934 173 "/><result name="DCU_cal_high_VMIN" value="274 607 942 1275 1610 1945 2282 2622 2953 3288 3631 4065 2260 "/> What I get from LYON DB DCU_CAL_LOWINPUT_MV 9 19 29 39 48 58 68 78 87 97 107 117 126 136 146 DCU_CAL_LOW_VNOM 144 16 21 27 39 55 72 91 109 129 148 168 187 208 228 DCU_CAL_LOW_VMAX 150 12 17 24 32 48 65 84 102 122 141 161 181 201 221 DCU_CAL_LOW_VMIN 166 24 28 40 54 71 88 107 126 145 164 184 203 224 243 DCU_CAL_HIGHINPUT_MV 156 312 468 624 780 936 1092 1248 1404 1560 1716 1872 2028 DCU_CAL_HIGH_VNOM 249 570 905 1240 1575 1911 2245 2583 2914 3246 3581 3914 4095 DCU_CAL_HIGH_VMAX 242 564 898 1233 1569 1904 2238 2576 2906 3239 3573 3905 4095 DCU_CAL_HIGH_VMIN 265 587 921 1256 1592 1929 2264 2602 2933 3266 3606 3995 4095 26. 10. 2011 A. Kaminskiy 6
What do we know about DCU ADC gain According to Guido Magazzo • ADCoffset is 0 because of autocompensation mechanism of DCU • ADC is not linear below 20 mv A. Messineo et al, 2011 Xavier Rouby, 2003 26. 10. 2011 A. Kaminskiy 7
FHIT meaurement 2: ADCcount for all 8 DCU channels using 3 different V 250 (“power scan”) What I get from LYON DB MOD 3020033001 HYB 30216681601821 DCU_FHIT 1047123 DCU_CORR 5503503 TOOL 311 FIRM rev 20 BOARD rev 22 DATE 21/11/2003 08: 51: 58 ET_V 125_IN_MV 1279 1367 1123 ET_V 250_IN_MV 2500 2695 2207 ch 0 raw 333 356 362 mv 207 206 ch 1 raw 2533 2746 2224 mv 1287 1374 1127 ch 2 raw 2538 2751 2230 mv 1289 1376 1130 ch 3 raw 560 16 17 mv 318 42 36 ch 4 raw 175 171 169 mv 130 117 111 ch 5 raw 3329 3604 2935 mv 1677 1792 1479 ch 6 raw 2361 2374 2343 mv 1202 1192 1186 ch 7 raw 2493 2631 2468 mv 1267 1317 1248 This seems to be exactly what we need if we want to measure V 250 and V 125 at the hybrid. But…… 26. 10. 2011 A. Kaminskiy 8
Concerning DCU ID Q: Are DCUid read out by FHIT correctly? A: NO. As in Arc system, 3 bytes of DCUid are scrambled. But those are scrambled always in the same way. Q. Could we recover the correct DCU_ID? A. Yes. We should just unscramble them. Q. Can we verify that this unscrambling is valid for ALL the hybrirs? A. I checked for TIB/TID, where I do know which module is where. Thanks to the file by A. Rizzi, I have Module. Id <->Det. Id <->Dcu. Id correspondence. So, unscrambled FHIT DCUid corresponds to DCUhard. Id read out by FEC. Q. And for TOB and TEC ? A. You can do this ###### exercise it by yourself ! 26. 10. 2011 A. Kaminskiy 9
26. 10. 2011 A. Kaminskiy 10
26. 10. 2011 A. Kaminskiy 11
Hybrid vs FHIT offset is ~8 m. V ADC Input m. V SLOPEHi 2. 14507+- 0. 00134889 offset. Hi -55. 1956 1. 47072 SLOPEHi_vmax 2. 14406+- 0. 00140184 offset. Hi_vmax -61. 1916 1. 52846 SLOPEHi_vmin 2. 14981+- 0. 00135809 offset. Hi_vmin -41. 3069 1. 48075 Residuals (difference between real DCU counts and linear fit with P 0=0) 26. 10. 2011 A. Kaminskiy 12
Hybrid vs FHIT offset is ~23 m. V ADC Input m. V SLOPEHi 2. 1481+- 0. 00152236 offset. Hi 1. 03512 1. 61834 SLOPEHi_vmax 2. 14631+- 0. 00148776 offset. Hi_vmax -4. 84211 1. 58156 SLOPEHi_vmin 2. 15272+- 0. 00132859 offset. Hi_vmin 23. 6784 1. 41235 Residuals (difference between real DCU counts and linear fit with P 0=0) 26. 10. 2011 A. Kaminskiy 13
Let’s try to understand where this offset comes from… TIB only 26. 10. 2011 A. Kaminskiy 14
Does the offset depend from number if APVs ? 4 APV 6 APV 26. 10. 2011 A. Kaminskiy 15
26. 10. 2011 A. Kaminskiy 16
Power scan data: let’s select only points with 0 < offset <100 26. 10. 2011 A. Kaminskiy 17
Cut "V 250 offset>0 && V 250 offset < 100“ Keep 2693 of 3482 entries 26. 10. 2011 A. Kaminskiy 18
"V 250 offset>0 && V 250 offset < 100" So, selecting the records (measurements) where “short voltage scan” makes sense, We cut away the low gain peak in long DCU calibration curve In other words, This low gain peak is just an artifact of the measurements Since the main peak is quite narrow, let’s set all the ADC gain to 2. 15 and look at V 250 and V 125 maps. Note: since we divide on ADCGain 0, we expect that “low ADCGain 0” modules result in A. lower converted values 26. 10. 2011 Kaminskiy 19
V 250 “Individual” calibration “Global” calibration 26. 10. 2011 Measles is cured !! A. Kaminskiy 20
V 125 “Individual” calibration “Global” calibration 26. 10. 2011 A. Kaminskiy 21
Official ADC Gain What’s about the previous calibration? Was it correct? ADC Gain There is a clear correlation between “official calibration” and what I have got from the measurements. So Det. ID <-> Dcu. Id map was correct. The “low gain peak was just cut and assigned to “mean” value 26. 10. 2011 A. Kaminskiy 22
I 20 calibration (DCU current source used to supply thermistors) seems to be also affected by fake peak in adc. Gain 0. 26. 10. 2011 A. Kaminskiy 23
Kdiv value: all TIB hybrids are higher then TOB/TEC Is it real? Or TIB adaptor boards are bit different ? 26. 10. 2011 A. Kaminskiy TIB_6_stereo_up TIB_6_stereo_down TOB_4_stereo_up TOB_4_stereo_down TOB_4_rphi_up TOB_4_rphi_down TOB_6_rphi_up TIB_6_rphi_down TIB_4_rphi_down TEC_6_up TEC_6_down TEC_4_up I don’t see any reason why for TIB hybrids the voltage divider should work in a different way. Resistance of the both arms are THE SAME, so it doesn’t matter which exactly resistance is. I would say, the systematics come from the adapter boards, which are different for TIB TOB TEC 24
Conclusions: • DCU calibration file (“basic. root”) had some “errors” which could be repaired • DCUid are scrambled, but always (at least for TIB) in the same way, so DCUid could be recovered • DCU calibration made by FHIT suffered from FHIT-Hybrid voltage offset, which shift calibration curves • Voltage shift scales with number of APVs • No evident Voltage. Shift correlation found with • DCUid or its single bits • Setup id (tool id) • Hybrid type (with same number of chips) • Low ADCGain peak corresponds to big offset measured during other electical test; SO: Individual DCU calibration is not needed. It brings just additional uncertainty…. 26. 10. 2011 A. Kaminskiy 25
How many thermistors are actually connected ? • Should be 1 for TIB, 2 for TOB/TEC • There is a list (from A. Venturi) of exceptions (97 modules), where there is 1 thermistor in TOB/TEC • Just applying the list one can cure some number of “hot spots”, but introduce “cold spots” • Intermittent contact in thermistor ? Tsil without correction 26. 10. 2011 Tsil with correction A. Kaminskiy 26
CONCLUSIONS (my own!) • DCU calibration made by FHIT setup ( which is an excellent hybrid tester!) are not precise enough to spot a difference between single DCUs • Anyway since we have a huge statistics, we can select good measurements (still a good statistics) and assign the mean values to all DCUs. • Non-linearity of DCU ADC in the very beginning of its range affects • Leakage current measurements (at present integrated luminosity) • Temperature of the HOT MODULES • I didn’t manage to calibrate the DCU in the beginning of its dynamic range. • We cannot state if DCU ADC calibration curve has an offset (so we put it to zero, which could be wrong) • Uncertainty of temperature measurement is ~few C and it will not be worse when we go cold • Is there some systematics? It would be very useful to make and absolute calibration with real TK, possibly with a big lever arm. • The possibility of use Vband. Gap and Tdcu are not explored enough. To be done…. 26. 10. 2011 A. Kaminskiy 27
Backup slide Distribution if Ileak in STANDBY for non irradiated tracker 26. 10. 2011 A. Kaminskiy 28
Coherent noise during HV ramp up Fake hot cooling loop during HV ramp up Only TSil Tsil 26. 10. 2011 A. Kaminskiy 29
- Slides: 29