Transcript Document

Lab measurements and simulations of hips
• previously presented APV measurements* assumed signal divided equally between 7 channels
-> significant deadtime predictions for CMS
• but relative absence of –ve saturated baseline events (no signal) in test beam data
either beam test analysis biased (true for results presented previously)
or 7 channel model pessimistic (probably also true)
=> worth investigating effects of different hip signal distributions
OUTLINE
Introduction
Simulations (SPICE)
New deadtime measurements for hip signals on one/two channels
Hit loss rate predictions for new deadtime measurements
Summary
*http://cmsdoc.cern.ch/Tracker/managment/Agenda_GTM/GM_01_12/Mark_CMShipstalk.ppt
[email protected]
CMS Tracker Week, January 2002
1
saturated signals
in 4 strips in
this example
Introduction
X5 hip event
~ 8 mip
range
• X5 hip event shows up as saturated signals in
several channels
• APV output range only ~ 8 mip (0.7 MeV) so no
information on actual signal size in saturated channels
• First measurements on APV modelled hip charge shared equally between 7 channels
(choice simply governed by number of chans available on test setup)
• Recoil nucleus should have short range (e.g. < 43mm for E < 100 MeV) but true situation more complicated
• V. large signals on one/two channels still give > 0.7 MeV signals on neighbours due to inter-channel
capacitance
[email protected]
CMS Tracker Week, January 2002
2
SPICE simulations
• motivation: can’t see what’s going on inside chip otherwise
R (on hybrid
1/chip)
sensor
vCM
APV
preamp
s.f.
• model: 128 channels with nearest neighbour interstrip
capacitance (10pF) and AC coupling to APV I/P
• preamp o/p (after s.f.) linear to ~ 50 mips (4.5 MeV)
inverter
• inverter O/P linear to ~35 mips (3.2 MeV)
vi
vo = -vi + vCM
1.0
1.0
source follower O/P
0.8
0.6
10
10 mip
steps
0.4
0.2
0.6
0.4
0.2
80 mip
0.0
0.0
0.0 0.5 1.0 1.5
time
[email protected]
inverter O/P
0.8
Volts
Volts
hip
signal
• signals > ~ 50 mips on a single channel cause that
channels inverter to draw max current
-> significant voltage disturbance on vCM
CMS Tracker Week, January 2002
-6
2.5x10
0.0 0.5 1.0 1.5
time
-6
2.5x10
3
Simulations (2)
1.0
1.30
source follower O/P
0.6
nearest neighbours
0.4
0.2
-6
2.5x10
inverter O/P
0.8
Volts
preamp
R=100
R=50
s.f.
inverter
vi
vo = -vi + vCM
1.10
0.0 0.5 1.0 1.5
time
R=50
R=100
0.6
0.4
0.2
vCM
1.20
1.15
hip channel
0.0
1.0
vCM
1.25
Volts
Volts
0.8
R (on hybrid
1/chip)
non-hip channel
0.0 0.5 1.0 1.5
time
[email protected]
-6
2.5x10
0.0 0.5 1.0 1.5
time
-6
2.5x10
results here for 200 mip (18 MeV) signal on one channel only
• saturated signal in hip channel
• big signal in nearest neighbours (~25 mip), shorter duration
• combination -> transient disturbance vCM on R
• vCM disturbance couples to inverter O/Ps of all channels
• reduced value of R reduces effect
• “spikey” behaviour of vCM interesting, could be decoupled
CMS Tracker Week, January 2002
4
Lab measurements - “improved” setup for charge injection
10pF
previous
this study
7 APV I/Ps
see hip charge
shared equally
10pF
10pF
10pF
hip charge injected on
one or two channels
other channels see
signal due to interstrip
capacitance
10pF
10pF
ADC units
111 mips
500 mips
1111 mips
200
200
200
160
160
160
120
120
120
80
80
80
40
40
40
0
0
[email protected]
• Inter-channel capacitance -> signal sharing and saturated
signals in several channels
• i.e. localised hip signal still shows results consistent
with beam data
0
0
10 MeV
• These results for hip charge injection on one channel only
5
45 MeV
100 MeV
CMS Tracker Week, January 2002
5
Deadtime measurement technique
Inject and measure amplitude (in APV O/P frame)
of normal size signal
300
ADC units
250
sweep injection time of hip signal
200
normal signal disappears during period when
hip signal causing baseline saturation for all channels
150
100
unplug normal signal and repeat to get baseline
50
0
0
200
400
600
800
1000
100
50
subtract baseline measurement from measurement
with signal -> result gives deadtime = period
during which the chip is insensitive to signals
all measurements here in deconvolution mode
0
0
200
400
600
time [nsec]
800
1000
inject normal signal
latency
vary injection time of hip signal
[email protected]
trigger on
normal signal
CMS Tracker Week, January 2002
t
6
Deadtime measurements
• hip signal confined to 1 channel only
R = 100W
ADC units
10 MeV
100 MeV
0
R = 50W
hip signal size
[300 mm Si Mips]
hip signal size
[300 mm Si Mips]
111
111
140
140
175
175
222
222
279
279
350
350
442
442
500
500
630
630
794
794
1111
1111
250
500
time [nsec]
[email protected]
0
• Deadtime dependence on hip signal size
characterised by a threshold and then
rising to a saturated level
• Main difference when R -> 50
is increase in energy threshold required to produce
deadtime
•Deadtime saturation level
~125 ns R=100W
~100 ns R=50W
(5 bunch crossings)
(4 bunch crossings)
250
500
time [nsec]
CMS Tracker Week, January 2002
7
Deadtime measurements
• hip signal shared between 2 channels
R = 100W
R = 50W
56
• Threshold energy for onset of deadtime
significantly worse than for signal on
one channel case
70
88
ADC units
10 MeV
100 MeV
0
111
111
140
140
175
175
222
222
279
279
350
350
442
442
500
500
630
630
794
794
1111
1111
• Significant improvement in threshold energy
and deadtime duration when R -> 50W
• Deadtime saturation level
~300 ns R=100W
(12 bunch crossings)
~100 ns R=50W
(4 bunch crossings)
250
500 0
250
500
time [nsec]
time [nsec]
[email protected]
CMS Tracker Week, January 2002
8
vCM measurements vs. simulation
simulation
1.25
1.3
1.20
1.2
Volts
Volts
measurement
1.15
1.1
1.10
1.0
1.05
0.9
R=100W, C=0.1mF
R=100W
0
time [msec/div.]
time
3x10
-6
• Voltage measured (with scope probe) on inverter supply resistor
-> some similarity between measurement and simulation
• Decoupling inverter supply effective at removing spike
• Effect on deadtime worth investigating
[email protected]
CMS Tracker Week, January 2002
9
Deadtime measurements – effect of decoupling inverter supply
R = 100W
R = 50W
R = 100W + 0.1mF
56
R = 50W + 0.1mF
• Results here for signal
shared between 2 chans
70
ADC units
88
0
111
111
140
140
175
175
222
222
279
279
350
350
442
442
500
500
630
630
630
794
794
794
794
1111
1111
1111
1111
111
111
140
140
175
175
222
222
279
279
350
350
442
442
500
500
630
• Effect of decoupling
“spike” on inverter
supply quite dramatic
for R=100W case,
• Less so in 50W case,
but still some
improvement
100 MeV
0
250
500
250
5000
time [nsec]
time [nsec]
CMS Tracker Week, January 2002
10
250
500 0
250
500
time [nsec]
time [nsec]
[email protected]
10 MeV
Deadtime measurements – comparison with previous result
R = 100W
1600
• Results here for 100W inverter supply
resistor (existing situation)
Deadtime [nsec]
1400
1200
1000
signal shared equally between 7 channels
signal on 2 chans with interchannel capacitance
signal on 1 chan with interchanel capacitance
800
600
400
• 7 channel case shows smooth rise (up to ~ 60
bunch crossings at high energies)
• one/two channel + inter-channel capacitance
model show big reduction in saturation level
over 7-channel equal sharing model (5 – 12 bunches)
200
0
0
400
800
1200
Hip energy [ Mips in 300mm Si]
111 MeV
[email protected]
• but deadtime starts to appear sooner in 2 chan case
and hit loss calculation sensitive to this threshold
1111 MeV
CMS Tracker Week, January 2002
11
Deadtime measurements – effect of decoupling and/or reducing R
R = 100W + 0.1mF
R = 100W
400
signal on 2 chans
signal on 1 chan
R = 50W + 0.1mF
R = 50W
signal on 2 chans
signal on 1 chan
signal on 2 chans
signal on 1 chan
signal on 2 chans
signal on 1 chan
Deadtime [nsec]
300
200
100
0
0
400
800
1200
Energy [ Mips in 300mm Si]
0
400
800
1200
Energy [ Mips in 300mm Si]
0
400
800
1200
Energy [ Mips in 300mm Si]
0
400
800
1200
Energy [ Mips in 300mm Si]
• Deadtime dependence on whether hip signal on 1 or 2 channels significant only in R=100W case
• Decoupling and/or reducing R -> substantial improvement in deadtime
• Can parameterise deadtime and use to predict deadtime in CMS
but already obvious that R-> 50W and/or adding decoupling will give significant improvement
[email protected]
CMS Tracker Week, January 2002
12
Hit loss rate predictions - method
deadtime(E)
Prob.(E)
Prob. of missing hit (E)
X5
CMS
1 MeV 10 MeV 100 MeV
10 MeV
100 MeV
Prob. of missing hit (E) = Prob.(E)*[deadtime(E)/25ns]*128*occupancy
Total probability of hit loss per layer =
SE Prob.(E)
(note: above plots taken from previous talk (7 – channel case))
[email protected]
CMS Tracker Week, January 2002
13
Hit loss rate predictions – for CMS (G.H.)
Total probability of hit loss
(per 300(500)mm layer, per % occupancy)
R=100W
R=50W
signal shared equally between 7 chans
(previous result adjusted for latest simulations (M.H.)
0.33 (0.76) %
0.20 (0.49) %
signal shared equally between 2 chans
(+ inter-channel capacitance)
0.34 (0.65) %
0.023 (0.053) %
• 7 chan vs. 2 chan: No significant difference in hit loss prob. for R = 100
- reduction in deadtime at high hip energy compensated for by lower threshold for deadtime onset
• 7 channel results:
R: 100 -> 50 gives ~ 40% reduction in hit loss probability
but in 2 channel case get better than order of magnitude reduction
- presumably reduction in vCM transient much more effective for charge distribution produced
by 2 chan. + inter-chan capacitance model
• not calculated exhaustively but other variants (1 chan only and/or decoupling) will give results
similar to 2 chan/50W case
[email protected]
CMS Tracker Week, January 2002
14
Threshold hip energy required for saturated baseline
• APV lab measurements
7 – channel equal sharing
1 chan + inter-channel capacitance
2 chan + inter-channel capacitance
9 – 18 Mev
13 – 16 MeV
6 – 8 Mev
• simple linear CM assumption
depends on analogue O/P baseline position
if ¼ to ½ output range
25 – 50 MeV
• discrepancy => saturated baseline threshold = non-linear function of hip energy
• actual hip energy required to saturate baseline depends on details of charge distribution
[email protected]
CMS Tracker Week, January 2002
15
Summary
• Modelling hip signal as large charge deposited in one or two channels
-> saturated signals in more channels if inter-channel capacitance included
• Resulting hit loss rate prediction (2 chan.) similar to previous 7 chan equal sharing measurements
but order of magnitude improvement if R -> 50W
(relative hit loss rate could go from 0.3% -> 0.02% per 300 mm layer per % occ.)
• One/two channel results here suggest deadtime resulting from hip events could be in
range 5 – 12 bunch crossings for existing inverter power scheme
-> some evidence for this in existing X5 beam data.
• Accurate determination of hit loss rate in CMS depends on:
how well hip spectrum known (magnitude and rate)
hip energy distribution between channels (will vary from event to event)
[email protected]
CMS Tracker Week, January 2002
16