1. The Effect of Highly Ionising Particles on the APV25 Readout Chip
R. Bainbridge on behalf of the CMS Tracker collaboration
Colmar, 11th September 2002

2. Highly Ionising events
Interactions between incident particles and silicon sensors can produce Highly Ionising Particles (i.e, recoiling nuclei and/or nuclear fragments)
HIP events result in signals equivalent of up to 1000 minimum ionising particles, which can saturate the FE electronics and result in signal loss
Estimate of +Si interaction rate in the CMS Tracker:
(+p)  200 mb  (+Si)  2.6 barn (E = 300 MeV)
 mean free path (+ in Si) = 7.7 cm
 rate  5  10-3 per BX per plane per % occupancy
Rate at which these interactions saturate the FE electronics?…
Duration of this saturated state?… Resulting inefficiency?…

3. First observation of the HIP effect
X5 test beam (Oct 2001) saw 6 TOB modules exposed to 120 GeV pion beam with 25 ns bunch structure
APV analogue data exhibit large (truncated) signals and suppressed output in all other channels (i.e, negative common mode shifts)
A sufficiently large signal will fully suppress the output of all 128 channels  APV insensitive to MIP signals
“Deadtime” is the period for which an APV is insensitive to MIP signals
Raw analogue output from 6 modules
(4 APV25s per module)
Disabled APV + activity downstream

4. Origin of problem Inverter stage powering scheme:
All 128 inverter FETs draw from V250 via external hybrid resistor, RINV
On-chip CM subtraction (VR  VCM)
Large signals drive down output on all other channels
Inverter FETs seeing HIP signal draw largest current possible
 VR pulled down
 output of all channels suppressed
Output remains suppressed until signal dissipates
Possible solution:
Reduce RINV from 100  to 50 
 more current available at inverter stage (reduces effect of large signals pulling down VR)

5. Identifying HIP events
Large signals result in negative tail in CM distribution
Peak at -150 is a consequence of limited APV dynamic range
APV still sensitive to MIP signals for intermediate CM shifts
Only fully suppressed baselines result in deadtime
HIP events that result in deadtime are identified by:
CM  & Qclu  500
CMMAX
-140
500
Common mode distribution [ADC counts]
Cluster charge distribution for CM  -150 [ADC counts]

6. HIP rate measurement (X5)
“HIP rate” = rate at which +Si interactions result in deadtime
(not rate of +Si interactions!)
HIP rate [per pion per plane] = NHIPs
Ntriggers  Nplanes  Multiplicity
Beam
Ntriggers
Nplanes
Multiplicity
NHIPs
Rate ( stat. error)
Pion
247500 6
1.9
1051
(3.7  0.1) · 10-4
288077 5
2.0
1142
(4.0  0.1) · 10-4
Muon
225000 7
(3.3  1.2) · 10-6

7. X5 beam structure
25 ns bunch structure, with 924 RF buckets per SPS orbit (23.1 s)
Pions delivered to X5 area in trains of ~60 consecutive RF buckets
1 train per SPS orbit  trains spaced by empty gaps of ~22 s
Triggers were distributed uniformly throughout train (for these particular run conditions / trigger requirements)
SPS orbit = 23.1 s
(924 RF buckets)
1 train per
SPS orbit X5
Train length  60  25 ns
Distribution of triggers in trains

8. HIP rate measurement (X5)
Distribution of HIP events in particle trains
Alternative HIP rate measurement:
Normalise distribution of “HIP events” to trigger distribution to obtain:
“probability of trigger containing HIP event” as function of trigger position in train
Probability of trigger containing HIP event is constant: ~4  10-3
 4  10-4 per pion per plane
Probability of trigger containing HIP event as function of trigger position

9. Identifying disabled APVs
APVs experiencing deadtime can also be identified:
If a trigger shortly follows a HIP event, then an APV may be read out when in a saturated state
No signal and pedestal structure are observed in APVs experiencing deadtime
 criteria for selecting disabled APVs:
CM  -140 ADC counts,
spread  10 ADC counts
Spread = ADCMAX – ADCMIN
Spread [ADC counts]
Peak at ~70 ADC counts and tail due to MIP signals
Disabled APVs

10. Deadtime measurement (X5)
Normalise distribution of “disabled APVs” to trigger distribution to obtain:
“probability of trigger containing disabled APV” as function of trigger position in train
Probability of disabled APV is small at start of train as cannot be HIP events in preceding RF buckets
Probability saturates later in train due to finite deadtime
“Risetime” provides measure of average deadtime: ~300 ns
Distribution of disabled APVs in trains
Deadtime ~300 ns

11. Laboratory deadtime measurements
Measure sensitivity of APV25 to normal signals after simulated “HIP”
Inject & measure amplitude of MIP signal
Sweep injection time of “HIP” signal t
Latency
MIP signal
Trigger on MIP signal
Vary HIP injection time
Deadtime as a function of HIP signal (Edep)
Response to MIP signal after HIP
Deadtime
Measured deadtime  350 ns
Threshold Edep ~10 MeV required before deadtime is observed

12. Simulation* of pion-Si interactions
Cumulative energy deposition spectrum
Differential energy deposition spectrum
X5 data provide no information on signal magnitude and spatial distribution  simulation required
Predicts Edep up to ~100 MeV
Probability of Edep  10 MeV is ~10-3
Predicts comparable HIP rates for X5 and CMS
Only inelastic collisions contribute significantly to P(Edep  1 MeV)
 negligible HIP rate with muons
* CMS NOTE 2002/011 (M. Huhtinen)

13. PSI beam test Aims for dedicated HIP study:
Verification of previous HIP rate measurements at X5
Observation of full APV25 recovery after HIPs to allow direct deadtime measurement
Rate / deadtime measurements with reduced RINV values
Hardware/trigger requirements for PSI
Trigger logic providing particle-vetoes before triggers (clean measurements)
Trigger card providing multiple triggers (every 75 ns)
+ APV25 multi-mode operation (peak mode readout, 3 frames per trigger)
 data every 25 ns for 750 ns after initial trigger
6 TOB + 3 TIB + 3 TEC modules exposed to 300 MeV pion beam

14. HIP rate measurement (PSI)
Criteria for selecting HIPs as for X5
Rates in agreement with simulation and X5 results
Observe factor ~0.6 between rates for 500 m and 320 m sensors
Slightly reduced HIP rate for lower values of RINV
Type RINV [] HIP rate
320 m thick sensors:
TIB (2.3 ± 0.1) 10-4
TIB 100 (3.7 ± 0.1) 10-4
500 m thick sensors:
TEC 100 (6.2 ± 0.1) 10-4
TOB (5.5 ± 0.1) 10-4
TOB (5.8 ± 0.2) 10-4
TOB 100 (6.2 ± 0.1) 10-4
Preliminary

15. Example of APV25 recovery
Peak
(raw data)
Top: raw analogue data (peak mode)
Bottom: numerical deconvolution of raw data
350 ns
325 ns
375 ns
275 ns
200 ns
225 ns
250 ns
400 ns
300 ns
425 ns
575 ns
600 ns
625 ns
650 ns
550 ns
525 ns
450 ns
475 ns
500 ns
175 ns
125 ns
25 ns
0 ns
150 ns
50 ns
75 ns
100 ns
300
t = ns: HIP event
t = 125 ns: Baselines begin to recover
t = 200 ns: Baseline flat in deconvolution mode
Digital zero
t = 275 ns: Nominal baseline position in decon mode
Continued slow recovery in peak mode
100
Deconvolution
(numerical calc)
200
t = 650 ns: Baseline still recovering in peak mode
Slow time-varying signals filtered out by deconvolution algorithm
 improved baseline behaviour (~flat and shorter recovery period)
Nominal baseline pos.

16. Baseline recovery
Average baseline position as function of time after HIP
(peak mode)
Curves grouped according to module type and resistor value
Modules equipped with lower RINV values recover first
Expect quicker recovery and no overshoot in decon mode
Time [ns]
Baseline position
Nominal position
Fully saturated
Disabled APVs?

17. Inefficiency due to HIP effect
“Hit reconstruction efficiency in hipped APV” as function of time after HIP event
Time [ns]
1.0
0.8
0.6
0.4
0.2
0.0
Efficiency
-60 < CM < -30
-90 < CM < -60
-30 < CM < 0
-30 < CM < 0
Simulation provides probability of a given energy deposition, P(E)
Deadtime dependence on energy, (E), provided by lab measurements
Resulting inefficiency (per plane per % occupancy) is sub-% level
Less than inefficiency due to unbonded or noisy strips
Using PSI data, perform efficiency scans for various CM shifts and convert into “effective deadtime”, ’(CM)
Calculate probability of observing CM shift, P’(CM), from PSI data
Inefficiency again sub-% level
Preliminary

18. Conclusions
Large signals, resulting from rare interactions between incident hadrons and silicon sensors, can saturate the FE electronics and result in inefficiencies (per plane per % occupancy) at sub-% level
Encouraging that we are investigating effects in the readout chain that affect the performance of the Tracker at 10-3 level