the atlas pixel detector - running experience – markus keil – university of geneva on behalf of...

The ATLAS Pixel Detector - Running Experience –

Markus Keil – University of Genevaon behalf of the ATLAS Collaboration

Vertex 2009Putten, Netherlands, 13/09/2009

M. Keil - ATLAS Pixels: Running Experience

Outline

Introduction

Timeline & Detector Status

Calibrations

Cosmic Ray Data Taking

The ATLAS Detector


Length: 44 mHeight: 25 m

Pixel Detector

Requirements:– Position resolution in

rΦ-direction < 15μm– 3 track points for |η| < 2.5– Time resolution < 25 ns– Hit detection efficiency > 97%

Basic Properties:– 1744 Pixel Modules on

three barrel layers and 2 x 3 discs

– 80M readout channels– Innermost layer at 5 cm

• Radiation tolerance 500 kGy / 1015 1 MeVneqcm-2

– Evaporative C3F8 cooling integrated in local support structure → Module temperature below 0○C


The ATLAS Pixel Module

Sensor– 250 μm thick n-on-n Si sensor– 47232 (328 x 144) Pixels– Typical pixel size 50 x 400 μm

(50 x 600 μm pixels in gaps between FE chips)

– Bias voltage 150 – 600 V

Readout– 16 FE Chips with 2880 pixels each– Pulse height measured by means

of Time over Threshold– Zero suppression in the FE chips,

MCC chip builds module event – Data transfer 40 to 160 MHz

depending on layer (occupancy)



Outline

Introduction


Calibrations


Pixel Detector Operation 2008 and 2009

Pixel Detector sign-off was foreseen to happen in a two-week period end of April / beginning of May 2008

– Cooling plant failure May 1st → Stopped until August– August 2008: Evaporative Cooling back, sign-off finished

September – October 2008: Combined cosmic ray data taking, pixel HV off during beam injection tests

November – December 2008: Calibration and Combined cosmic ray data taking

January – May 2009: Cooling plant consolidation Jun 2009: ATLAS Cosmic ray data taking July-August 2009: Detector consolidation


Detector Status

Leak-down measurements indicate a cooling fluid loss of 14 kg/year– Three leaky loops not operated in 2008; this year all loops will be operated

unless a degradation should be observed.– Risk analysis of C3F8 in the N2 atmosphere of the inner detector ongoing

(expected concentration < 0.1%)

Off-detector failures– Optical transmitter plugins in the readout crates (VCSEL arrays) had a failure

rate of 1 per week of operation– All optical transmitter plugins in the readout crates have been exchanged; new

plugins produced with stricter ESD protection requirements

1.6% of the detector cannot be operated due to on-detector failures– Mostly due to problems in the optical links or the HV connections



Outline

Introduction


Calibrations


Optical Data Link Tuning

Parameters to be adjusted:– On-detector laser power– Sampling threshold and phase

Bit error rate measurement for full parameter space to find the best parameter settings

Difficulty: one common laser power setting for 6/7 modules

In 2008: 97% of the detector successfully tuned

Now: Improved algorithms → Almost all links can be tuned automatically


Threshold and Noise

Threshold setting: 4000 e Threshold tuned pixel by pixel, threshold dispersion after tuning ~ 40 e Noise approx. 200 e for most pixels, slightly higher for special pixels Threshold / Noise approx. 25 for normal pixels


Interchip Region


Time over Threshold


FE Chips provide Time over Threshold information for each hit– Nearly linear dependence on deposited charge

Pixel-by-pixel tuning; chosen tuning: 30 BC for 20 ke (corresponding to the signal of 1 mip) Calibration by means of test charge injection Charge measurement with ToT in cosmics data taking

– Landau peak at 18300 (Simulation 19000 e): Confirms ToT Calibration


Outline

Introduction


Calibrations


ATLAS Cosmic Ray Data Taking

Smooth integration into ATLAS data taking beginning of September 2008 High voltage off during LHC beam injection (no stable beam) to protect FE chips

→ no beam-splash events taken

Timing in with ATLAS → first cosmic ray track seen September 14 400k tracks taken with and without magnetic field 96% of the modules operated (3 leaking cooling loops conservatively switched off) Between day 35 and day 70: in-situ calibration of detector 250k tracks taken in 2009


Cosmic Ray Tracks


• Track with 8 pixel hits on track (2 x 2 hits in module overlap regions)

• Red: hits on track• Green: unassociated hits (noise)

• Noise occupancy: ~ 10-10 hits/pixel/BC

Alignment

Alignment of pixel barrel modules from cosmics data – Beam data needed for endcap alignment

Alignment not yet perfect due to limited statistics, but large improvement w.r.t. nominal geometry and good starting point for alignment with beam:

– Precision direction: 128 μm → 24 μm beam direction: 282 μm → 131 μm


Efficiency and Noise Occupancy

After alignment measured efficiency is >99.7% for active modules

Noise occupancy after masking of noisy pixels: ~ 10-10

– Fraction of masked pixels: 10-4


Timing

Each hit has to be assigned to the correct bunch crossing (25 ns)

Module clocks have to be precisely aligned with the bunch crossing clock

In cosmics data taking: readout of 8 consecutive bunch crossings (plot shows hit time w.r.t. beginning of readout window)

Correction of propagation delays:– First step: time alignment of readout crates

(scope measurements)– Second step: time alignment of modules

(using cable length data) Remaining effects:

– Trigger jitter– Random phase of cosmics– Timewalk; “in-time” for less than 5400 e

Plan to start data taking with 5 BC readout, later reduce readout window to 1 BC


Ext

erna

l Del

ay

= c

onst

- t

imew

alk

Lorentz Angle Measurement

Cluster size vs. Track angle with and without magnetic field → Measurement of the Lorentz angle

Measured value close to expected value (225 mrad) Theoretically expected dependence on mobility can be nicely seen when

including modules of different temperature– Measured: (-0.78 ± 0.18) mrad/K, expected: -0.74 mrad/K


Summary

Pixel detector signed off and commissioned in very short time in 2008 Calibration measurements show detector performance as expected

– Noise approximately 200 e, Threshold dispersion approx. 40 e– Time over threshold allows for reliable charge measurement

Cosmic ray data taking with approximately 96% of the detector very useful for understanding of the detector

– Good progress in alignment– Efficiency after alignment > 99.5%– Noise occupancy approximately 10-10 hits/pixel/BC– Useful input to detector simulation

Current status and next steps:– Calibration program has been rerun at lower temperature– Combined cosmic ray run Sept./Oct.– The Pixel Detector is ready for beam with ≥ 98% working modules!


Backup


Number of Masked Pixels



The Frontend Chip

One frontend chip contains 2880 pixel cells organised in 18 columns with 160 pixels each (400 x 50 mm)

Each pixel cell in the matrix contains preamplifier, discriminator and readout logic, which transfers hits to buffers at the bottom of the chip

Peripheral region contains hit buffers, logic for trigger coincidence and data serialisation and programmable DACs for the currents and voltages needed for the operation of the chip.

Hit transfer to EOC is done column pair wise, whereas pixel configuration is done with a 2880 bit long shift register that connects all pixels


The Pixel Cell


Preamplifier and Discriminator Signal Shapes

Time over threshold (length of discriminator signal) depends on– Deposited charge– Discriminator threshold– Feedback current

Information of the ToT (in units of 25 ns) is read out together with the hit information → possibility to measure the deposited charge

Optical Links


the atlas pixel detector - running experience – markus keil – university of geneva on behalf of...

Documents

atlas pixel detector

keil atlas pixels

pixel detector sign

detector statusleak

detector failuresmostly

detector consolidationm

atlas detectorm

pixel detector operation