(non-)Destructive high-rate tests on silicon strip modules

1
(non-)Destructive high-rate tests on silicon
strip modules

• Emulating LHC beam incidents using the PS booster
  and measuring the effect on a LHCb Velo silicon
  strip module
• Lars Eklund, on the behalf of the LHCb
  Collaboration

2
Outline

• Introduction
• Motivation and previous publications
• The participants
• The PS booster and the LHCb/VELO module
• The measurements
• Observables and program
• The surprise
• Results and interpretations
• Summary

3
Motivation (1)

• The LHC
• Stored beam energy 102 - 103 times larger than
  any previous accelerator
• New machine, limited operational experience
• The LHCb Velo
• Very close to the beam silicon sensors _at_ 7-30 mm
  distance (moving!)
• Located next to the injection line TI8
• Designed and built but operation procedures can
  be changed
• LV HV on/off at injection?
• Feedback to the machine
• Intensity limit at injection
• Currently H/W 1011 protons and F/W 1010

4
Motivation (2)

• Possible beam incidents
• Injection failures
• incomplete or unsynchronized kicker fire gt
  mostly Alice LHCb
• wrong magnet settings in transfer line gt mostly
  Alice LHCb
• wrong magnet settings in the LHC gt everybody
• Circulating beam failures (mostly caught by
  collimators)
• magnet failure / mishap gt everybody
• RF failure gt everybody
• collimator failure / mishap gt everybody
• Extraction failures
• Under-kick, unsynchronized beam dump gt mostly
  CMS

5
Previous studies

• Atlas silicon strip sensors LASER (2 types)
• IEEE Trans. Nucl. Sci. NS47 (2000) 1902
• Voltage across AC coupling vs. RRC CRC
• Atlas silicon strip 1064 nm LASER (1 W)
• NIM A 541 (2005) 15-20
• Beam spot 8 µm, 10 ns pulse, 109 MIP equivalent
• Damage HV bias gt 200 V _at_ 109 MIP (on one strip)
• CMS silicon strip 24 GeV protons (CERN/PS)
• NIM A 518 (2004) 328-330
• Beam spot 10x3 cm2, 42 ns bunch, 2 bunches of
  7x1010 protons
• No damage
• Atlas pixel 24 GeV protons (CERN/PS)
• NIM A 565 (2006) 50
• Beam spot 6x3 cm2, 42 ns bunch, 8 bunches of
  1x1011 protons
• No damage

6
The PS booster
Resto2
Main bldg
7
The beam line

• Proton beam with 1.4 GeV energy
• Intensity 2x109 9x1012 p
• Beam spot 5 mm (max 4x1013 p/cm2)
• Bunch length 200 ns
• Cf. tests in the PS max 3x1010 p/cm2
• Compare with LHC
• Pilot bunch _at_ injection 2x109 protons (450 GeV)
• 300 µm beam spot
• 0.4 ns bunch length
• Full luminosity (L1034) SPS injection train
• 288 bunches of 1011 protons
• 4x1013 protons/cm2/bunch

8
The set-up

• Module mounted close to the beam dump
• Back-splash gives non-negligible dose
• Rough estimate of dose 1013 neq 1 kGy (very
  preliminary)
• Small scale experiment

9
The victim

• LHCb/Velo spare from production
• Double sided (R Phi sensors)
• 2048 AC coupled n-on-n strips / side
• 16 FE chips (IBM 0.25 µm)

• Mounted in the beam line
• Cooled to 1 C (LV on)
• Florescent screen to view the beam
• Insert/retract from beam line
• Remote control and read-out

10
Electrical model static case
CDET 1 nF/2048 ch. RDET 1-100 MO/2048
ch. CAC 250 nF/2048 ch. Rbias 1 kO x 2048
ch. CRC 10 nF RRC 5 kO CFB 400 fF
(per ch.) CG 10 pF (per ch.) CLV 32 x
100nF
11
The measurement sequence - observables

• Intensity steps 2x109, 2x1010, 2x1011, 2x1012
  9x1012
• Each step LV/HV off, LV on/HV off, LV on/HV 150
  V LV on/HV 300V
• Each beam shot follows the same pattern
• A set of standard measurements
• I/V of both sensors
• Noise pedestal data
• Test pulse data at 1.5, 0 and -150 V (for some
  shots)
• Insert the module, acquire during the shot
• 14 consecutive triggers of front-end data
• Voltage on hybrid GND and sensor bias via
  oscilloscope
• Beam spot image via a a camera
• Repeat the same set of measurements
• Shots on two sensor positions
• Shots on five front-end chips (only LV on/off
  matters)

• No measurable damage up to
• 9x1012 _at_ 300V bias on the sensor
• 2x1011 (LV on) on the FE chips

12
Beam images
Combined R-F sensor front-end data
Beam line camera on scintillating screen
13
I/V curves

• I/V curves in-situ between each shot
• Superimpose temperature corrected I/V curves
• Small increase probably due to accumulated dose
• Rough estimate between first and last curve
  3x1012 neq 200 Gy
• Work in progress
• Correlate with radiation monitoring data

14
Thermal image No hot-spots
The majority of the shots hit this area
15
Noise Pedestals

• Noise pedestals measured in-situ between each
  shot
• Plots show date taken towards the end of the
  program
• No change visible
• Detailed analysis is in progress

16
Test pulse response post-zap

• Test pulse response
• booster in-situ after a few shots at 2x109
• lab lab measurement after the full program
• Gain difference due to different analogue
  drivers/receivers
• Bad channels identical to production QA

17
Post-mortem why did it survive?
vivum

• Deposited energy (in 300 µm Si)
• 9x1012 x 24 k MIPs x 3.6 eV 1.2 Joule / 200 ns
• Temperature increase in 1 cm2 Si 2.5 C
• Maximum SPS injection train (288x1011) 4 Joule /
  10 µs
• Local energy store the RC filter
• 10 nF _at_ 300V gt 0.5 mJ
• Absorption volume critical
• Massive ionisation in biased silicon
• QRC(300V) 3 µC
• Deposited charge _at_ 2x109 7.5 µC
• Possible transient damage
• Current through front-end
• AC coupling diode
• Voltage on front-end input
• Fast HV ramp-down

18
Voltage across the sensor vs. time

• Oscilloscope measurements
• Hybrid GND
• Backplane
• 1 sample / ns
• Ground reference arbitrary
• Huge ground bounce
• Large pick-up
• Plot Vbackplane-VhybridGND
• Two distinct features
• Sharp rising edge (50 ns)
• Slow charge-up

19
The first 50 ns
6 GV/s
2.5 GV/s
2 GV/s
20
Electrical model the first 50 ns

• VIN IZ/N x RIN
• N is large ( 2048)
• RIN is small (Os)

Ramping 300 to 0 V in 50 ns seems to be OK!
21
Shots on the FE chips

• 56 shots on the FE chips 2x109 2x1011
• No destructive latch-up
• Design rules include structures to prevent
  latch-up
• Seems to be effective!
• SEU analysis in progress none observed so far
• Requires large energy deposited in small volume
• Nuclear reactions necessary
• Cross-section very low
• Triple-redundant registers corrected every 2 ns

22
Summary

• The PS booster provided beam to emulate LHC beam
  incidents
• 200 ns shots, 2x109 to 91012 protons
• A VELO strip module was subject to a large number
  of shots
• Two positions on the sensor, five FE chips
• Survived 9x1012 protons on sensor with 300 V bias
• Survived 2x1011 protons on the FE chip
• No visible change in performance
• I/V curves, noise, pedestals, thermal imaging,
• Saving graces
• The whole sensor responds as a unit
• Large area sensor many channels
• CAC gtgt CRC (CDET)
• Protection diodes on the FE inputs
• Triple-redundant registers in FE chips
• Analysis measurement still in progress

23
Back-up slides
24
Total number of shots
Shots on the sensor (position 12)
Intensity LV off HV off LV on HV off LV on HV 150V LV on HV 300V
2109 1 2 293 2
21010 1 1 1 1
21011 1 1 1 1
21012 1 1 1 1
91012 2 2 5 5
Shots on the front-end chips
Intensity Beetle 4 Beetle 4 Beetle 5 Beetle 5 Beetle 6 Beetle 6 Beetle 7 Beetle 7
Intensity LV on LV off LV on LV off LV on LV off LV on LV off
2109 - - 2 4 3 3 3 6
21010 3 3 5 3 3 3 6 6
21011 - 3 - - - - - -
25
Beam size seen by the F-sensor
Response to beam during initial 25 ns of beam
rising edge in f detector
26
Fitting rising edge of all shots

• Termination of HV monitoring signal was improved
  during the program
• Rising edge not affected by termination
• 150 V Shots 3-5 24 _at_ 2e9, shot 10 _at_ 2e10 and
  shot 14 _at_ 2e11 are less than 1 GV/s
• 300V 9e12 Shots 34, 42, 44 are greater than 5
  GV/s
• Weak correlation with intensity voltage
• Large shot-to-shot variation

27
Re-charge of HV

• Average time constants
• t 6.8 µs _at_ 2e9 150V
• t 13 µs _at_ 9e12 150V
• t 10 µs _at_ 9e12 300V
• Need spice simulation to understand recovery
  times
• Re-charge depend on intensity
• Some long term (10µs) process in the sensor?

28
Decay-time of all available wave forms

• Falling edge clearly affected by the termination
• Not possible to compare the two data-sets
• De-convolution of impulse response maybe possible