(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.

(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

14 September, 2009

L. Eklund, Vertex 2009 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

14 September, 2009


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 @ 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

14 September, 2009


Motivation (2)

Possible beam incidents• Injection failures:

– incomplete or unsynchronized kicker fire => mostly Alice & LHCb

– wrong magnet settings in transfer line => mostly Alice & LHCb– wrong magnet settings in the LHC => everybody

• Circulating beam failures: (mostly caught by collimators)– magnet failure / mishap => everybody– RF failure => everybody– collimator failure / mishap => everybody

• Extraction failures:– Under-kick, unsynchronized beam dump => mostly CMS

14 September, 2009


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 > 200 V @ 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

14 September, 2009


The PS booster

RestoResto22Main Main bldgbldg

14 September, 2009


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 @ injection: 2x109 protons (450 GeV)

– 300 µm beam spot– 0.4 ns bunch length

• Full luminosity (L=1034) SPS injection train– 288 bunches of 1011 protons– 4x1013 protons/cm2/bunch

14 September, 2009


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

14 September, 2009


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

Electrical model – static case

Al

SiO2

n

p+

n+

CDETRDET

CACRbias

CRC

QRC

CRC

RRCRRC

HV bias (-300V)

HV return (GND)

QRC

RC filter

GND bonds (16x5)

pre-amp

CFB

Vfp

CG

protection diodes

bond wires

FE inputs (2048 channels)

VDD bonds (16x4)

LV (GND)

LV (VDD)

CLV

CDET = 1 nF/2048 ch. RDET = 1-100 MΩ/2048 ch.CAC = 250 nF/2048 ch.Rbias = 1 kΩ x 2048 ch.CRC = 10 nFRRC = 5 kΩCFB = 400 fF (per ch.)CG = 10 pF (per ch.)CLV = 32 x 100nF

10 MΩ

GND probe

HV probe

Osc. GND

22 nF

1 kΩ

10 MΩ

10 pF

10 pF

14 September, 2009


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 m

easu

rable

dam

age

up to

• 9x1

012

@ 3

00V b

ias o

n th

e se

nsor

• 2x1

011

(LV o

n) o

n th

e FE ch

ips

14 September, 2009


Beam images

Beam line camera on scintillating screen Combined R-Φ sensor front-end data

14 September, 2009


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 September, 2009


Thermal image: No hot-spots

The majority of the shots hit this area

14 September, 2009


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

14 September, 2009


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

14 September, 2009


Post-mortem – why did it survive?

• 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 @ 300V => 0.5 mJ– Absorption volume critical

• Massive ionisation in biased silicon

– QRC(300V) = 3 µC

– Deposited charge @ 2x109: 7.5 µC • Possible transient damage

– Current through front-end– AC coupling diode– Voltage on front-end input– Fast HV ramp-down

vivum

HV bias reduced to 0 V

14 September, 2009


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-uptime [µs]

14 September, 2009

L. Eklund, Vertex 2009 19time [µs]

The first 50 ns …

6 GV/s

2 GV/s

2.5 GV/s

14 September, 2009


Electrical model – the first 50 ns …

Al

SiO2

n

p+

n+

CDETIZ

CACRbias

CRC = 10 nF

QRC = 3 µC

RRC

RC filter

GND bonds (16x5)

pre-amp

CG

protection diodes

bond wires

FE inputs (2048 channels) VDD bonds (16x4)

CLV

IZ/2048

VAC

IGND = IZ/80

VIN

IZ

RIN

Adt

dVCI RCRCZ 50

Current during the discharge

Divided between 2048 inputs and 80 GND bonds

QRC transferred to CAC

VVVAC

RC

CC

biasAC 12

VIN = IZ/N x RIN

• N is large (~ 2048)• RIN is small (~Ωs)

Ramping 300 to 0 V in 50 ns seems to be OK!

Rbias reduced to ~100kΩ/2048 via punch-through mechanism Still to large to play a role

14 September, 2009


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

14 September, 2009


Summary

• The PS booster provided beam to emulate LHC beam incidents– 200 ns shots, 2x109 to 9*1012 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 >> CRC (+CDET)– Protection diodes on the FE inputs– Triple-redundant registers in FE chips

• Analysis & measurement still in progress

14 September, 2009


Back-up slides

14 September, 2009


Total number of shots

Intensity

LV offHV off

LV onHV off

LV onHV

150V

LV onHV

300V

2*109 1 2 29+3 2

2*1010 1 1 1 1

2*1011 1 1 1 1

2*1012 1 1 1 1

9*1012 2 2 5 5

Shots on the sensor (position 1+2)

Intensity

Beetle 4 Beetle 5 Beetle 6 Beetle 7

LV on

LV off

LV on

LV off

LV on

LV off

LV on

LV off

2*109 - - 2 4 3 3 3 6

2*1010 3 3 5 3 3 3 6 6

2*1011 - 3 - - - - - -

Shots on the front-end chips

B2

B6

B0

B3

B4

B7

B5

B1

63 shots on the sensor

56 shots on the FE chips

14 September, 2009


Beam size – seen by the Φ-sensor

“FWHM” of beam~80 strips of 70 m pitch~5.5 mm

Response to beam during initial 25 ns of beam rising edge in detector

14 September, 2009


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 @ 2e9, shot 10 @

2e10 and shot 14 @ 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

14 September, 2009


Re-charge of HV …

• Average time constants– τ = 6.8 μs @ 2e9 & 150V– τ = 13 μs @ 9e12 & 150V– τ = 10 μs @ 9e12 & 300V

• Need spice simulation to understand recovery times

• Re-charge depend on intensity– Some long term (10μs)

process in the sensor?

14 September, 2009


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

(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.

Documents

beam failures

beam line proton beam

beam insertretract

p beam spot

bunch slide

lhc beam incidents

beam spot 10x3

beam spot 6x3