1. Cosmic tests and Performance of the ATLAS Semi-Conductor Tracker Barrels Bilge Demirköz Oxford University On behalf of the ATLAS SCT Collaboration

2. Overview Description of a barrel module Macro-assembly: Modules  Barrels Data Acquisition System: Calibration mode –Barrel tests, check for common mode noise –Performance 4 Barrels integration  “SCT” SCT + TRT = Inner Detector Barrel Data Acquisition System: Physics mode Cosmics!

3. SCT Barrel Module Each side has: 2 silicon sensors –Hamamatsu –768 instrumented strips –Strip pitch = 80µm Binary Readout chip: 6 ABCD3T ASICs –Discriminator –Pipeline –Data compression logic –Readout Buffer 2 sides glued with 40 mRad stereo angle TPG baseboard with BeO ceramic facings –Thermal management –Up to 10W / module at lifetime dose! 12 cm chan 0 chan 767 40mrad stereo  6cm

4. ATLAS inner detector SCT barrels: 4 layers SCT endcaps: 9 disks Pixels TRT

5. Barrel 3 macro-assembly

6. Hardware Data Flow ROD BOC Formatted Event MODULE BPM TTC SERIAL TTC x48 DATA x96 DORIC VDC DATA CLKCOM DATA Electrical Domain Optical Domain SBC VME Configuration & Commands Histograms & Data Data Path in ATLAS TIM Backplane TTC BOC Configuration ROS Formatted Event SLink Readout Crate On Detector Event Builder TCP CTP

7. Software data flow Configure and define test Run… from ROD  RawData FittingService AnalysisService –finds optimal settings of the ABCD chip –calculates averages for noise, offsets… –cuts defined for identifying “defects” –publishes  TestResults –publishes a subset of the TestResults  (IS)Summaries  COOL (database for offline)

8. Testing and when? Modules are tested after –Production (at module assembly sites) –Macro-assembly (at Oxford) –Reception at CERN –Insertion (at CERN) Check that module performance does not change at different stages Development of “final running” software through all these stages –Also used for endcaps –How to recover errors on modules?

9. Typical Test Sequence Basic Tests –Establish Communication –Optimise Opto settings Digital Tests –Verify Communication –Do trigger counters work? Analogue Tests –Measure Gain, Offset, Noise –Measure Noise Occupancy –Look for Time Structure –Detect excess noise possibly related to high frequency, synchronous triggers Check module supply and sensor currents The Binary Architecture of the ATLAS SCT dictates that analogue information must be extracted by scanning chip thresholds.

10. Histogram Data (from ROD) Channel Fits Calibration Curves Fit Analyse Channel threshold corrections Channel DAC setting Threshold in mV Example Calibration

11. Results = 4 x 10 -5 Equivalent Noise Charge (ENC)

12. Checks for common mode noise The number of hits per event is summed and histogrammed. At higher thresholds, all entries should fall in the 0-3 bin –“Good Chip!” Any “spikes” might suggest the presence of common mode –“Bad Chip?” One event with between 12 - 15 hits Vertical scale – Threshold in mV with reference to 1.0fC – so 20mV is ~1.3fC Except when coincident with other known defects, effects such as this were not reproducible. No significant common mode noise observed! Good Chip! Bad Chip? Plots for Module 20220040200324

13. Double Trigger Noise Verify that the detector readout does not generate noise in subsequent events Send two consecutive triggers separated by a controlled time interval Scan around 132 bco = pipeline depth Pipeline == stores data until trigger (L1A) is received.

14. # of bcos 6 x10 -4 3 x10 -5 # of bcos Barrel 3 LMT09 z-4 Lower side of the module

15. Light leak! One opto-package leaks light onto sensor Effect occurs at bin 136 111010nnnnbbbbbbbb time channels 2 light leaks on Barrel3: occupancy of 6 x10 -4 on a few channels  very small effect in the large picture… Optopackages on other barrels fixed before macro-assembly No lightleaks found!

16. Channel Defect Statistics BarrelTotal Channels Not bonded DeadNot Reachable Part bonded NoisyOtherTotal Defects 358982418035738491460111483 4737280552452561624227841 588473617377025697492301818 6103219238525136401971936495720 Total32440327933885153640131301179862 The Bottom Line:  99.8% working channels! Here add the graph From the module paper Showing number of defects On module

17. Insertion of B3, 20/09/2005

18. SCT inserted into the TRT 17/02/2006 Poster by Heinz Pernegger on “SCT Integration and Testing”

19. Barrel TRT Proportional counters embedded in radiator –4mm diameter straw tube, 30µm wire –1.4m long wires, readout on both ends –Cosmics: Ar:CO 2 = 70:30 73 layers of straws ~ 36 hits per track Total number of straws: ~ 210176 Noise occupancy = 2% in 75nsec Final: Xe:CO 2 :O 2 = 70:27:3 Drift time accuracy ~140µm Measured with efficiency 87%

20. Cosmics setup SCT: 504modules 12RODs, 1TIM, 1Master LTP TRT: 2x6568 chan 9RODs, 3 TTCv, 1 Slave LTP 3 Scintillators 144cmx40cmx2.5cm TDC/QDC measurement Trigger time jitter ~ 0.5nsec ~ 300MeV cutoff for alignment studies View from outside towards Side A 20cm of concrete

21. TDC/ADC measurement Measure time of flight between scintillators charge deposited by the muon for event and momentum selection Measure Relative time of trigger with respect to system clock for phase corrections TDC, Time resolution ~ 35 psec ADC, Charge resolution ~100 fC Decoded online and performed simple time checks Not yet decoded and used in the offline.

22. Physics mode of SCT Set thresholds to optimal 1fC Read 3 time bins instead of 1: Expanded mode –1 time bin = 25 nsec Accept any hit in these three time bins Latency in components: trigger, TIM, ROD, BOC, cable/fibre lengths… ~ 30bcos SCT has a pipeline depth of 132 bcos Need to delay this signal by the right amount TRT: reads every 3.125nsec and 24 time bins (total of 75 nsec).

23. A module sees cosmics! Coincidences between module sides And histogram for each module! On ROD, histogram while scanning through BOC Tx coarse delays (32bcos) noise Coincidences between module sides = cosmics!

24. Tracks through the top sector of the SCT and the TRT

25. Summary ATLAS SCT Barrel Integration is complete! –All 2112 modules tested –99.8% of 3.2 million channels working The Data Acquisition System, SctRodDaq, has been developed –Largest system to date: 1 crate, 14 ROD/BOC pairs, 672 modules –Ongoing work on multi-crate readout –On-ROD monitoring of dataflow –Can “time-in” the readout for physics data taking Combined sector test (SCT + TRT) with Cosmics from May 2006 Offline is looking at the data right now… –Present residuals from tracks (without alignment) are better than the specified building tolerances.

26. Next steps Collect high statistics sample of cosmics muons for alignment studies Offline decoding of the timing (TDC) information –Efficiency estimates –Alignment studies Digitization work –Use the online noise and defects lists from the COOL database in SCT simulation Move to the pit in Summer 2006 Endcaps move into the pit Fall 2006

27. Backup slides

28. TTCvx Trigger and Timing Trigger LTPTTCvi L1AinL1Aout/clocked BC A-channel B-channel OptoLink TIM SCT TRT LTP Busy Orbit ADC TDC Corbo start OR phase delay L1Ain Latency in components: trigger, TIM, ROD, BOC, cable/fibre lengths… ~ 30bcos NIM VME LTP= Local Timing Processor

29. SCT Sector Cooling (3kW of power -- SCT sector only) 800,000 channels in the SCT

30. SCT Module Readout 1 TX fibre = clock+command for each module 2 RX/data links fibre-optic communication opto-package (DORIC+VDC) connection to the module redundant TTC from neighboring module dog-leg

31. Opto-package Chip S11 module Side view of the barrel

32. SCT databases DAQ parameters: –Configuration data (chip settings,…) (frequently updated): 2MB/update –Average noise, offsets(…) for each chip and lists of defects for each module: ??MB/calibration DCS parameters: 100MB/day –Temperatures, currents for each module All in COOL by offline identifier Should be easy to correlate –Noise versus current –Figures given for commissioning of full barrel

33. SctRodDaq Framework -- simplify!!

34. Readout: Redundancy If one side of the module can not be readout from its master chip due to opto-failure, it can be readout through the other side excepting its own master chip. M0S2S1S3E5S4 M8S10S9S11E13S12 Link 0 Link 1

35. Readout: Redundancy If one side of the module can not be readout from its master chip due to opto-failure, it can be readout through the other side excepting its own master chip. M0S2S1S3E5S4 M8S10S9S11E13S12 Link 0 Link 1

36. Readout: Redundancy Modified module connects M8 to E13 and allows all 12 chips to be readout. (similarly for M0 after E5 for Link 0 opto-failure) Prototype has been made and tested. Will be available for Barrel 6 construction M0S2S1S3E5S4 M8S10S9S11E13S12 Link 0 Link 1

37. Correlated Noise?? Occupancy per event in Noise Occupancy Test Binomial Distribution Threshold over 1fC 128 channels max

38. OPE Analysis: LMT21 Z-6 high threshold Threshold over 1fC Channels