CMS Note - HEP
Contents
1. 20 0 50 100 150 200 250 nsec Figure 9 Measured pulse shapes at various settings of VSHA decimal values from the APV User Manual We have taken data with VSHA varying from 0 to 0x64 We have found that in our case increasing it beyond 0x4B does not result in any increase in the seen charge suggesting that the optimal value lies around 0x32 0x40 4 Conclusions The APV6 coupled to a full sized detector seems to perform as expected from our previous experience with the PreMux chips We have illuminated a Forward Milestone detector with 2 MeV electrons obtaining signal to noise ratios of 20 5 to 1 in peak and 12 3 to 1 in deconvolution mode We will now proceed to test irradiated detectors in the same way The parameter settings and their influence on chip behaviour need further systematic testing on many more hybrids so that we can have a better grasp of what the best possible operating conditions are for silicon detectors in the CMS tracker References 1 Characterization of neutron irradiated silicon microstrip detectors Lenzi et al To be published in Proceedings of Como 1998 INTERNATIONAL CONFERENCE ON ADVANCED TECHNOLOGY AND PARTICLE PHYSICS 2 APV6 User Manual M French 3 The APV6 Readout chip for CMS Microstrip detectors M Raymond et al Third Workshop on electronics for LHC experiments October 1997 CERN LHCC 97 60 4 First Testbeam with APV6 Silicon System M2. Syncronous to APV Out En 40 Mhz clock a True trigger within a 3ns time Interface Board does the following Differential Buffering of the APV analog signal gain 6 mV uA Level Addaptation of LVDS clock and trigger signals to the APV Buffering of the Output Enable signal from the APV Figure 1 Block diagram of the DAQ set up We have taken great care in assuring that all APV6 signals trigger reset calibration are correctly in phase with the 40 MHz clock also great care has been placed on the receiving network so that the signals remain clean The detector hybrids with the APV6s have been produced at CERN these hybrids can hold up to four chips and have 100 Ohm resistors on them to terminate clock and signal lines In our set up these resistors have been unsoldered as correct termination and polarization levels are provided by the interface board see appendix a This last board also buffers the analog output and the output enable signals from the APV6s The B source used is Strontium 90 with the use of an electromagnet only the end point electrons 2MeV are selected collimated and sent to the detector The scintillator placed on top of the tested detector provides the particle triggers Finally the pulser allows us to precisely set the correct delay value between arrival of the particle and the triggering of the APV6 Note that only particles arriving inside a 3 ns time window of the clock rising edge are use
3. Equivalent input capacitance 13 5 pF 2 The noise distribution plots show relevant tails which have nothing to do with the way the APV6 works but are due to some shorted strips on the detector that was used for these measurements 4 I2C register settings are as above in table 1 Test beam in addition the three registers dedicated to calibration are set as follows e CLVL 0x3c e CSKW Oxfd e CDRV Oxf7 The first sets the amplitude of the calibration pulse the second sets the skewing between calibration pulse and clock the third which channels receive the pulse In figure 3 we show the results obtained signal and noise both in peak mode and in deconvolution mode with pedestals and common noise subtracted The spread of the signal distribution is due to gain variations between the 128 channels and injection capacitor value differences convoluted with noise The noise distribution plot shows the measured pedestal sigmas for the 128 channels Common mode noise distribution not shown is gaussian with a sigma value of less than four ADC channels The value of the CLVL register was inferred from ref 2 to be roughly equivalent to a MIP signal 375 electrons per count We verified with an external capacitor that this value is correct Unfortunately this is a tricky measurement and we estimate an error of roughly 30 on this verification Another work 3 gives an estimate of 625 electrons per count implying that a MIP signal corre
4. Friedl CMS Tracker Week July 1998 5 Addendum to the June 98 APV6 Testbeam W Adam et al CMS Week December 1998 10 Layout of the interface board Appendix A Al inant azicng 2BRRRRRRRE g ConrE dge 13x2 Lemo 4x diferernide 100rF R13 750 EE R39 750 SLemm oittereraiae CAF LESIO RCADSTRITRACIBUFFER DSN TRI Tracker Front End Readout Intertace 11 Appendix B Sequencer block diagram gt LDs Clock to APV a Trigger to APV _ it ay A DELAY gt Qos ae DAQ Trig in N Cal request Trig pulse lt cal Trig in Example in the case of Cal Trig in 7 x being used pM Reset MAX 7160 7ns gt lt Output enable from APV Serial input RS 232 wos 20 Mhz Conversion clock to ADC Serial Input used to program the delay between DAQ Trig in and Trigger to APV 25 ns steps In calibration mode use Cal Trig in the sequencer will generate a prompt calibration pulse 50 ns width followed after the programmed delay by a trigger pulse The ADC Conversion clock starts on falling edge of the Output enable from APV 12
5. NIN 0 83 peak decon These results are in some agreement with what we saw in calibration mode especially when one considers as stated previously that the decoupling capacitors present on the detector further reduce the signal by 10 and that the error on the absolute charge calibration value In peak mode we have increased the delay by 15 ns where the output reaches the maximum this is in agreement with the expected amplifier response 2 and with what shown in figure 5 In deconvolution mode instead we retain the original delay setting which gives the best results In this mode it is very important to track precisely the starting point 0 to 10 of the pulse which is not affected by the VSHA setting anyway At this point we can use the Landau peak to extract the noise of our system in E N C counts Assuming that a MIP releases 24000 electrons in our detector a 2 MeV electron is a close approximation of a MIP we have e 24000 0 92 22080 Charge effectively transferred through the de coupling capacitor e 22080 20 5 1077 electrons noise in peak mode e 22080 12 3 1795 electrons noise in deconvolution mode To be compared with the expected values of 996 and 1621 electrons for respectively peak and deconvolution mode One of our worries with this high setting of VSHA was that we might see charge and clusters for out of time 25 ns events since as shown in figure 5b the pulse shape does not return to zero af
6. Z 5 2 VSHA 0x7a 196 mV R 200 s s o o Jo VSHA 0 2V i 150 e oe 5 0 100 8 oe L e ee 50 o F 6 e ee e e J 0o o 0 o o o 50 t ganS oo ee a a E g G E T a g e a 0 20 40 60 80 100 120 Delay ns Figure 6 Pulse shape in deconvolution mode with VSHA 0x0 and VSHA Ox7E 3 As one can easily observe from fig 5 the choice made for the VSHA parameter is rather extreme We would like to stress that this does not correspond to the optimal value we would choose for normal operation We anyway wanted to test the chip functionality especially in deconvolution mode even with this extreme setting 7 50 Peak v lue 164 E Mean value 8 8 E E Peak value 8 0 40 E z 2 P E E D A uw 5 F ZO ia Ba i ae 6 source 10 E E Peak mode J CL pe cL id hl UP E A E E 6 8 0 VZ 14 16 Noise ADC Counts so E Peak val E i 60 BopdiE cE 229 E I K 200 E Mean value 10 0 OF E Peak value 9 6 ps TAS E 40 E 150 E n N a 2 E 2 E E 1 c SOF zZ Lu as lu 1 Ag A 6 source E Deconvolution mode 4 10E l 8 source E 25 E Deconvolution mode E LOL eln da n 1 wane 5 Et e A Th e Al 200 400 600 8 C 12 14 16 Charge ADC Counts Noise ADC Counts Figure 7 MIP signal and noise in peak top and deconvolution mode bottom Vdet 250 V VSHA Ox7E e S N peak 20 5 e S N deconvolution 12 3 e SAS 1 39 peak decon e
7. significant amount of charge on them 4 sigma noise cut We then look at the nearest neighbours of these seeds to see whether they also have some charge 2 sigma noise cut We sum the charge of the seed and its neighbours cluster and if the total charge is greater than five times the noise sigma value we record the event Obviously in calibration mode the clustering algorithm is skipped This procedure implies that our charge amplitude scale is always referred to zero The noise values we quote are the charge weighted average of the cluster strip s noise 2 Results 2 1 Calibration mode We first took data using the internal calibration capacitors of the APV6 ellbtetion Peak mode TE E Calibratio gt eak mode E Peak value 211 SEE Peak valu j4 E Mean value 8 6 12 E wn A m wo 47 o 40 Es D OF E Z E a e Eo mm 300 E ui Ez E E Ar E E ase nee i nil M r o KE DANC M A T 100 150 200 250 300 z 10 12 14 16 ADC counts ADC counts 20 CalibrationDesdavelutton med EF Calibration Noise Deconvolution mode Bees ti eat gs ah aL 18 eak value 9 7 Peak value 166 OE Mednvaluest0 4 S o 12 E oO E KE c CE 5 OF Ki z goe 2 E eso a eT o H MM H Mi if 100 150 200 250 300 8 14 16 ADC counts Figure 3 Data obtained using the internal calibration capacitors Signal and strip noise in peak top and deconvolution mode bottom
8. 6 mV Charge ADC counts 200 CR RC 50ns 150 100 L so be eet 120 fi ni L 140 160 180 w9 ii ea Pp tt ea 0 20 40 60 80 100 Delay ns Figure 5 Pulse shape in peak mode with VSHA 0x0 and VSHA 0x7E also shown is the ideal response of a CR_RC shaper peaking at 50 ns constants through the register VSHA assuming that a longer shaping time might be more suited to the actual pulse shape of a silicon detector signal 10 ns long which given the APV6 fast shaping is a far cry from an ideal delta like pulse Changing this parameter can have serious implications concerning how the chip will perform in deconvolution mode Namely we might lose its ability to resolve pulses with a 25 ns resolution In figure 5 we show the actual pulse shapes we have measured using the internal calibration with this parameter set at O nominal value from 2 and at hexadecimal 7E both in peak and in deconvolution mode As one can see changing VSHA can change dramatically the pulse shape In deconvolution mode though the algorithm manages anyway to filter out the long tail We then took data with this setting The results are shown in figure 7 where we show the Landau distributions for the MIP signals after common noise and pedestal subtraction and the distribution of the pedestal widths for those strips that were exposed to the collimated B source Deconvolution Mode v a Ss T
9. Available on CMS information server CMS NOTE 1999 058 The Compact Muon Solenoid Experiment CMS Note Mailing address CMS CERN CH 1211 GENEVA 23 Switzerland April 29 1999 Performance of the APV6 bonded to a full size Forward Milestone silicon detector Simone Busoni Carlo Civinini Raffaello D Alessandro Michela Lenzi Marco Meschini Andrea Piergentili Marco Pieri Dipartimento di Fisica e INFN Sez di Firenze Florence Italy Abstract We describe our test set up used to read out APV6 hybrids bonded to full size silicon detector modules 1 of the type built for the Forward Milestone Choice of APV6 parameters is discussed and results with calibration pulses and MIPs are presented 1 APV6 test set up 1 1 Introduction We have designed our set up so that it can also be used with irradiated detectors this has meant placing the source and part of the electronics inside a climatic chamber Discriminator i Hybrid Silicon Detector Kapton Collimated 6 This part sits inside the ee climatic chamber Magnet i Analog output to ADC F i Clock and trigger from Sequencer i H 12C from VME Output Enable to Sequencer H m 6 idth Teder ati iee 5 Triggers are accepted from Sequencer to Sequencer window Serial in Coarse 25ns programmable trigger delay 40 Mhz APV clock Out True Trigger In APV trigger Out Output Enable In 20 Mhz ADC clock Out
10. d for APV6 triggering For more information on the use of the APV6 see ref 2 3 1 2 Settings 1 2 1 I2C registers The APV6 allows great flexibility of use with the possibility of setting various parameters which control the way the chip responds to input signals We report in Table 1 the APV6 register nominal values and the values used in October 1998 test beam and which we used in our set up during calibration data taking Most of the parameters have only minor variations with the exception of VADJ which sets the APV6 output baseline The value of VSHA is set to zero this sets the shaper timing constant at its shortest possible value 50 ns peaking time REGISTER DEFAULT VALUE REGISTER B TEST VALUE IPRE 111 OX 6F IPRE 120 OX 78 ISHA 88 0X58 ISHA 81 0X51 IPSP 84 OX 54 IPSP 66 0X42 ISFB 43 OX 2B ISFB 41 0X29 VPRE 150 OX 96 VPRE 184 OX B8 VSHA 0 VSHA 0 VADJ 120 OX 78 VADJ 166 OX A6 VCAS 0 VCAS 0 Table 1 Nominal values for the APV6 registers from ref 2 and from the 1998 October test beam 1 2 2 Timing One of the most important parameters for the correct operation of the APV6 is the value of the latency register Below figure 2a 2b there are two timing diagrams for both modes of the APV6 settings Calibration and DAQ During the calibration procedure external delays due to the electronic chain that generates initial triggers are not important since the sequencer provides all the correctly
11. ns per count leads to a 30 worse result than what expected Given the uncertainties on the absolute value of the injected source we then started taking data with the B source system 2 2 DAQ mode with the B source As mentioned before the detector used with these measurements has a depletion voltage of 80 V We started by using the test beam parameters for the I2C registers and with the detector biased at 100 V figure 4 top The particles which hit the detector are 2 MeV electrons The peak value obtained by fitting a Landau distribution to the data is 101 ADC channels Peak mode and 80 ADC channels Deconvolution mode L oan E 100 Peak value 101 80 Peak value 80 k TO Es 80 H Vdet 100 volts S E E 60 E Vdet 100 volts 60 F 50 E L E ist T 5 40 E b source A A B source E Deconvolution mode Se Mi SO E li Peak mode Iz s 20 E 20 E E E 10 E O C iP AEE he ool O EI well Matron bool 200 400 600 200 400 600 Charge ADC Counts Charge ADC Counts 90 ES 100 E Peak value 94 80 E F 70 Ea Peak value 110 BO Vdet 250 volts 60 C n E o g Oe oO E Vdet 250 volts 2 60 Es sie L a E e D S E Z W40 E Ld M E 40 6 source w It Deconvolution mode YY Ls 90 F 6 source A i i E Peak mode 20 10 E L 0 E es epe ie i 0 L na onl Lan h 200 400 600 200 400 600 Charge ADC Counts Charge ADC Counts Figure 4 MIP signal in peak and deconvolution mode Vdet 100 V
12. se a pulser to set an additional delay D1 which is used to fine tune the exact value for the correct sampling of the particle signal 1 ns resolution This has the same function basically of the Skew register used in calibration mode As stated in the introduction particles are accepted only if within 3 ns from the APV6 40 MHz clock rising edge 1 3 Detector characteristics and data processing The detector used see ref 1 is a wedge shaped single sided silicon microstrip detector with 1024 readout strips with pitch varying from 50 u on the short side reaching 70 wu at the end of the module Strips are connected to the front end electronics through decoupling capacitors which are integrated on the detector These detectors have been fully characterized for what concerns impedance measurements between adjacent strips between strips and the backplane and of the decoupling capacitors From these measurements we infer that at a bias voltage of 250 V the input capacitance seen by the APV6 corresponds to 13 5 pF while the decoupling capacitor transfers 91 8 of the detector signal charge to the APV6 We bonded 256 strips to two APVs the data shown is for 128 strips 1 APV only Measured detector full depletion voltage is equal to 80 V Our data is reconstructed with the use of cluster finding algorithms standard to normal silicon detector operation First common noise and pedestals are subtracted then we look for strips channels which have a
13. sponds to a lower setting of CLVL 39 and not 60 as we used The difference between the two values is of the order of 45 which could very well be given the above quoted error and the variations between different chips due to injection capacitor values reference current variations and so on The data in the plots can be summarised as follows e S N peak 28 1 e S N deconvolution 17 1 Seal Saecon 1 27 e Noea Nascon 0 77 Taking from ref 3 the formula for the expected E N C noise as a function of input capacitance e Peak mode 510 36 pF e Deconvolution mode 1000 46 pF we expect an E N C value for our detector of 996 electrons in peak mode and 1621 electrons in deconvolution mode If we use 375 electrons per count 2 then we injected 22500 electrons in the APV6 In this case we have a measured E N C of 800 electrons in peak mode and 1315 electrons in deconvolution On the other hand choosing the 625 electron value from ref 3 will result in an E N C of 1334 electrons in peak mode and 2192 electrons in deconvolution mode Again summarising Peak mode Deconvolution mode Expected noise 996 electrons Expected noise 1621 electrons Measured noise using ref 2 800 electrons Measured noise using ref 2 1315 electrons Measured noise using ref 3 1334 electrons Measured noise using ref 3 2192 electrons So while a value of 375 electrons per count seems to lead to an overly optimistic result the 625 electro
14. ter 25 ns This would then lead to high values of occupancies in the tracker In the event we checked this and the results of the delay scan can be seen in figure 8 E 10 amp as 90 E x E Se 3 on E x E 20 k Bn Q E kia G ag i o 80 z N 70 E k Deconvolution Mode gt aan E JECONVOIULION oge S 50 E x Ci 4 L I o 10 20 Delay ns Dm E aa ais x 0 9 E Ss g 3 p xX 3 0 8 E Fa 2 E T G7 E i E 2 o6 E S n 3 b Sea 7 Deconvolution Mode D E aoa Bose o 03 E 602 E i Zo EA 4 i L pj 290 10 O 10 20 Delay ns Figure 8 Charge values top and relative cluster finding efficiency bottom as a function of time 0 corresponds to the optimal setting in deconvolution mode Vdet 250 V VSHA Ox7E The charge values which correspond to the Landaus peak values drop significantly in agreement with the pulse shape of figure 5b but what is more important the cluster reconstruction efficiency drops to 10 at 25 ns Thus even this high setting of VSHA seems to have little influence on occupancy rates We think that a careful choice of this parameter is a must for everybody interested in the use of the APV6 with silicon detectors A good starting point is given by figure 9 taken from ref 2 160 140 Ty a r E jt E 120 E 100 80 ADC counts Vsha 100 60 Vsha 75 Vsha 50 Vsha 25 Vsha 0 40
15. timed pulses from only one initial external pulse sent to its own calibration input see appendix b As we can see from the diagram the sequencer generates a pulse Cal Request 50 ns long followed after a delay D2 in steps of 25 ns set from the serial input by a trigger pulse 25 ns long After a settable delay Skew register from the falling edge of Cal Request the APV6 receives a a Calibration sequence b DAQ sequence ip 100 ji o i i 5 delta like pulse on its input it then reaches its peak value Latency is calculated from this moment onwards By varying D2 on the sequencer one can choose which cell in the analog pipeline to sample Figure 2 Timing sequence for Calibration mode a DAQ mode b 1 is the time of particle passage 2 photo multiplier output 3 discriminator output 4 NIM logic pulser delay D1 1ns resolution 5 trigger output to the APV6 on the sequencer board 6a signal after integration by the APV6 amplifier 6b trigger to the APV6 on the hybrid after cable and interface board In Calibration mode 1 is replaced by an external pulser All units are in ns diagram not to scale l Nominal values are such as to ensure a reasonable working point for the APV6 close to the design values intended for the chip On the other hand process parameter variations mean that optimal values should be established by measurement 3 In DAQ mode all delays are relevant in addition we u
16. top Vdet 250 V bottom e S peak 100 V 101 S N 13 5 e S deconvolution 100 V 80 S N 8 2 S eat Saccon 1 27 e NIN 0 77 peak decon The noise distributions not shown are in agreement with those obtained in the calibration runs and the Seat Saccon gt NoeaidN secon Values are also consistent However the absolute value of the collected charge is less than half of what was obtained in the calibration runs Even if we consider that due to the presence of de coupling capacitors we lose with these detectors 8 2 of the ionisation charge and that the absolute value of the calibration has an estimated error of 30 there still could be some other effect working against us We thought that one possible explanation could be inefficient charge collection To investigate this possibility we started raising the detector bias voltage up to 250 V corresponding to roughly a factor three over depletion In figure 4 bottom we show the signal and noise in these conditions again with test beam I2C parameters decon e S peak 250 V 110 S N 14 7 e S deconvolution 250 V 94 S N 9 7 S eat Saccon 1 17 e NIN 0 77 peak decon This has a clearly beneficial effect in as much that the detector signal is faster and thus the APV6 collects more charge but still we see less charge than what expected At this point we also started changing the shaper time Peak Mode 300 o VSHA 0 2V e VSHA 0x7a 19
Download Pdf Manuals
Related Search