T. Ince, Universitaet Bonn ICATPP, 2011/10/03
ATLAS pixel detector
Outline:
Technology
Calibration
Performance
Operation
Summary1
13th International Conference on Advanced Technology and Particle PhysicsOctober 3-7, 2011
Tayfun InceUniversitaet Bonn
on behalf of the ATLAS Collaboration
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Inner detector
2
beam pipe
2 T magnetic field.
Tracking of charged particles with pT > 0.5 GeV (as low as 0.1 GeV) and |η| < 2.5.
Electron id over |η| < 2.0.
Pixel: n-in-n silicon sensors.
SCT: p-in-n silicon sensors.
TRT: polyimide drift tubes interleaved with polypropylene fibers.
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Pixel detector1456 modules in three layers of barrel region. 288 modules in three layers of end-cap regions.
Barrel modules tilted by 20o in Rφ plane against the Lorentz angle to reduce occupancy.
80 million channels.
10 µm resolution in Rφ and 110 µm in z.
Modules operating at about -13 oC temperature.
3
50.5 mm88.5 mm122.5 mm
88.8 mm
149.6 mm SCT
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Pixel module250 µm thick silicon sensor with an active area of 16.4x60.8 mm2.
16 front-ends (FEs) per module bump bonded to the sensor.
Radiation hard up to NIEL 1015 1 MeV neq/cm2 and 50 MRad.
47232 pixels per module. Most with size 50x400 µm2. Special pixels (long and ganged) to cover regions between FEs.
Module controller chip (MCC) to send trigger, commands and configuration data, and to perform event building.
2x80 Mb/s readout in b-layer (innermost), 80 Mb/s in layer 1 and disks, and 40 Mb/s in layer 2.
4
lead-tinor
indiumbumps
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Optolink calibration (I)
Frequency of optolink calibration is driven by replacement of dead Tx plugins on the off-detector side. Done as often as once per week.
On-detector laser power, and off-detector sampling threshold and phase are tuned to settings that avoid data readout errors. One laser power setting per optoboard (i.e. 6 or 7 modules).
5
~1 m ~80 mon-detector off-detector
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Optolink calibration (II)Optical corruption measured in units of bit flips as a function of data link threshold and phase. Optimal point chosen in error-free region (EFR). Boundaries between EFR and error regions are not equally stable, so optimal operating point not in the centre of EFR.
Data (Up) links of b-layer, layer 1 and disks tuned at 80 Mb/s, while layer 2 at 40 Mb/s.
6
tuningpoint
error freeregion
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Threshold calibration (I)Frequency of module level calibration is driven by periods of long-enough (min 8 hours) downtime of the LHC. e.g. technical stop.
Inject # of test charges at target threshold, vary 7-bit DAC per pixel and count generated hits. Optimal value is at 50% efficiency.
If 7-bit DAC per pixel does not produce sufficient range of thresholds for tuning target, use 5-bit DAC per FE first to shift global threshold per FE.
7
thresholdfrom 50%
noise from16.5↔83.5%
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Threshold calibration (II)Threshold tuned to 3500 e. Dispersion is ~40 e.
Pixel noise is ~170 e for normal and long, and ~300 e for ganged.
0.3% of noisy pixels are masked online to reduce noise rate down to ~10-7 hits/pix/evt. After offline masking, rate <10-9 hits/pix/evt.
8
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
ToT calibration (I)Time-over-Threshold (ToT) is the time in which signal is above threshold. In units of bunch crossings (BC), i.e. 25 ns. ToT is proportional to the deposited charge.
Tuning target is 30 BC for a charge of 20k e (most probable for m.i.p.). Resolution ~1 BC.
Inject # of test charges at 20k e, vary 3-bit DAC per pixel to find optimal point resulting closest to target ToT.
In addition, can use two 8-bit DACs per FE to achieve target ToT and optimal tuning.
9
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
ToT calibration (II)Once tuned, inject varying test charges to measure ToT per pixel as a function injected charge.
Parametrize with ToT = p0(p1+Q)/(p2+Q) for each front-end and each pixel type for offline reconstruction.
Improved position resolution is an example advantage of measuring ToT.
10
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Timing and timewalkAfter timing adjustments for trigger delays, different cable lengths and module-by-module variations, dispersion is 0.17 ns.
Low charge hits may end-up in the next BC due to slow rise above threshold. Compensated by hit doubling based on ToT.
11
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Lorentz angleIn a magnetic field, charge carriers drift at an angle w.r.t the normal of electrode surface. Displacement is about 30 µm.
Lorentz angle corresponds to track incident angle resulting in min cluster size. 211.4±0.6 mrad for collisions.
Lorentz angle has a linear dependance to temperature changes with a slope of -0.78±0.18 mrad/K.
12
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Hit to track association and particle-idEfficiency of a track having hits for pixel detector layers crossed is about 99%. Dead modules excluded.
Full efficiency in b-layer due to track selection.
Association inefficiency mostly due to known dead regions (pixels and or FEs) in modules.
Charged particles with m>>me traversing material loses energy mostly via ionization and excitation, described by Bethe-Bloch formula.
Energy loss proportional to collected charge.
13
p
K
π
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Detector status0.1% of pixels have disconnected bumps. Most are from couple of disk modules.
0.04% of pixels are analog dead.10% of them have merged bumps.
96.8% of pixel detector operational. 55/1744 (3.2%) modules and 47/27904 (0.17%) FEs are disabled. 2.1% to 3.2% in 3 years of operation.
14
lead-tinindium
disconnected bumpsintegrated over all modules
Communication26
Optoboard6
Clock3
LV3
HV17
failure types
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Radiation damage monitoring (I)Radiation damage on the pixel detector will result in:
an increase of leakage current,an increase (decrease before type inversion) of voltage needed for full depletion of sensors,and a decrease in charge collection efficiency.
Leakage current per pixel measured using pixel FE.
Increases already measurable on all layers at the module level, while only on b-layer at the pixel level.
15
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Radiation damage monitoring (II)Leakage current per module also monitored via HV power supply with a precision ~80 nA per half-stave (i.e. 6 or 7 modules).
Increases on all layers are within predictions so far.
Before type inversion, decrease of depletion voltage measured via cross-talk method:
if not fully depleted, cross-talk between pixels is much higher.readout a pixel while injecting high enough charge into neighboring pixels to generate a cross-talk hit only when not fully depleted.
16
annealing
radiation
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Pixel operationFor Pixel safety, high voltage (HV) is not turned on until LHC declares STABLE BEAMS.
Preamplifiers are also off before switching on HV to avoid excessive noise which can block data acquisition (DAQ).
Once stable beams declared, apply WARM START procedure which includes checking background rates in beam conditions monitor (BCM) and collimator positions, ramping up the HV and switching on preamplifiers.
Pixel data taking efficiency is 99.9%. Loss dominated by warm start procedure.
17
>1 fb-1 in September 2011
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
SummaryThe ATLAS pixel detector is calibrated and performing well. Regular checks of performance and re-tuning when necessary.
Threshold tuned to 3500 e with a dispersion of 40 e.
ToT is tuned to 30 BC with a resolution of 1 BC.
Timing dispersion of 0.17 ns between modules.
0.1% of pixels with disconnected bumps. 0.04% of pixels are analog dead.
96.8% of the pixel detector is operational.
Radiation effects which are within expectations so far are monitored. LHC delivered ~4 fb-1 corresponding to NIEL 1013 1 MeV neq/cm2. Pixel can stand for O(102) more radiation.
Insertable B‐Layer (IBL) will take us even closer to the interaction point and restore/improve performance. See talk from J. Grosse‐Knetter on Wednesday.
18
T. Ince, Universitaet Bonn ICATPP, 2011/10/03
Questions
19
