CARIOCA

1
CARIOCA

Werner Riegler, CERN November 24th, 2003, LHCb
week
Discussion of the final Prototype results Plans
for CARIOCA / ASDQ decision
2
CARIOCA
TDR ASDQ is our baseline solution CARIOCA
is our preferred solution caveat we cannot
afford ASDQ the
3
CARIOCA

• CARIOCA is a responsibility of the CERN LHCb muon
  group.
• Francis Anghinolfi and Pierre Jarron are our
  advisors from the MIC group.

4
CARIOCA Block Diagram
Signal tail cancellation 2x pole/zero,
t01.5ns, topology from ASDQ
Preamp tail cancellation 1x pole/zero, topology
from ATLAS MDT
Topology from ATLAS MDT
Preamp
LVDS, standard cell
topology from ATLAS MDT prototype
5
Manpower
Walter continues in Cagliari
6
Submissions
7
CARIOCA

• We had a very useful review in February
• We got very useful suggestions in order to
  increase stability (coupling).
• Francis Anghinolfi got involved in order to help
  us ironing out some of the problems in the
  preamp.

8
CARIOCA10

• We received CARIOCA10 on September 15th.
• Test board designed by Davide (Cagliari) and
  produced at CERN.
• 17 CARIOCA boards were equipped ? 34 chips.
• Tests were started October 1st.
• All test results can be found on
  http//home.cern.ch/riegler

9
CARIOCA10

• 8 channels
• pos/neg switch
• Test pulse even/odd
• 8 individual thresholds
• Can be switched to a single threshold
• Analog output of channel 8

10
CARIOCA10
3x4mm chip 82 pins ? 25 pins on each side
11
CARIOCA10

• Traditionally one does extensive LAB tests before
  putting the chip on the chamber.
• Because our last testbeam period in T11 was
  October 22nd to Nov 11th , lab test are not yet
  finished
• There is no way we could have advanced further up
  to now
• CARIOCA10 was tested on M3R3 (4boards), GEM (6
  boards) and we fully equipped a CERN M3R1
  chamber.
• We found a nice way for high rate tests in GIF
  without having beam this is also ongoing.
• Results are preliminary

12
CARIOCA10 test board

• We wanted the results quickly,
• we dont have the final package
• We did an optimum and worst case package

• Optimum Package (no package)
• Chip bulk is glued to the board
• gound with conductive Epoxi,
• Wire bonds are very short

• Worst Case Package
• Chip bulk is insulated from the board
• gound
• Wire bonds are very long

13
Sensitivity (discriminator)
On CARIOCA10, sensitivity was doubled in order
to decrease minumum detectable charge (4fC?2fC)
for GEM application. Maximum threshold is
300mV (limited by discriminator). Sensitivity
decreases by factor 2 from 0 to 220pF.
14
Sensitivity variations
Channel to channel variations are smaller than
chip to chip variations
15
Sensitivity Variations, 0pF
Pos 16.0mV/fC, 0.56mV/fC r.m.s, i.e. 3.54.
Pos 14.5mV/fC, 0.62mV/fC r.m.s, i.e. 4.31
? package causes a decrease of 9 Neg
14.7mV/fC , 0.56mV/fC r.m.s., i.e. 3.8 Neg
13.1mV/fC, 0.56mV/fC i.e. 4.3 ? package
causes a decrease of 11.
16
Sensitivity Variations,0pF
Subtracting average per chip and scaling by
Sqrt(8/7) Pos 16.0mV/fC, 0.34mV/fC r.m.s,
i.e. 2.15. Pos 14.5mV/fC, 0.34mV/fC r.m.s,
i.e. 2.36 Neg 14.7mV/fC , 0.40mV/fC
r.m.s., i.e. 2.7 Neg 13.1mV/fC, 0.26mV/fC
i.e. 2.0
17
Sensitivity Variations
Sensitivity is 16(14.7) mV/fC for the positive
(negative) amplifier. Sensitivity variations are
lt5 r.m.s. The DIALOG DACs have 2.44mV LSB I.e.
0.16 fC _at_ 0pF and 0.32 fC _at_ 220pF
18
Extrapol. Minumum Detectable Charge
2.4 fC, 0.37 fC r.m.s. 2.4 fC, 0.24fC
r.m.s
Minumum detectable charge is correlated with the
sensitivity, I.e. the reason for this Limit is a
minimum voltage pulse at the Discriminator input
in order to make it fire.
19
Offsets

• Offsets were measured on 272 channels by
  recording the threshold value that inverts the
  discriminator output.
• One DTV sets the threshold for all 8 channels.

20
Offsets
Channel to channel variations are smaller than
chip to chip variations
21
Offsets
795.6mV, 9.9mV r.m.s. The threshold DACs on
the DIALOG chip have a range of 625mV to 1250mV
in 8 bits i.e. bins of 2.44mV. This is
perfectly compatible with this kind of offset
spread.
22
Offsets
Subtracting the average offset for each chip and
multiplying by sqrt(8/7) gives an rms of 4.54mV.
This is the true channel to channel
variation. It corresponds to 0.3fC at 0pF and
0.6fC at 220pF
23
The DTV applies the differential threshold
voltage to the discriminator.
24
DTV itself has an offset of about 7.5mV r.m.s
25
(No Transcript)
26
OffsetsSensitivity

• The channel to channel variation of the
  sensitivity is lt5.
• The channel to channel offset variation is around
  5mV r.m.s.
• Together with the DTV the channel to channel
  offset variation is 10mV r.m.s.
• Both variations become irrelevant when we use
  individual thresholds.

27
Noise
Neg 224042e-/pF                          At
0/100/200pF we can use Pos 188045e-/pF
threshold of 1.5/5/10 fC.  
28
Power Consumption
Power consumption is 43.3/46.6 mW/channel for
the positive/negative amplifier. On on board
(16 channels) the CARIOCA consumes
0.75W. DIALOG Voltage drop from regulator .
29
Chamber Test in T11

• M3R1 module 1 chamber (double cathode readout)
• Uniformity of this chamber was measured with
  CARIOCA9 for the CERN PRR.
• Crosstalk for single/double cathode readout was
  evaluated for this chamber with CARIOCA9.

30
N3 N4 N5 N6
N7 N8
S1S2, no package S3S4, package
Beam goes into the drawing
8,9 7,10 6,11 5,12 4,13
3,14 2,15 1,16
4,13 3,14 2,15 1,16 8,9
7,10 6,11 5,12
P9 P10 P11 P12
P14 P16 P13 P15
8,9 7,10 6,11 5,12 4,13
3,14 2,15 1,16
4,13 3,14 2,15 1,16 8,9
7,10 6,11 5,12
5,12 6,11 7,10 8,9 1,16
2,15 3,14 4,13
1,16 2,15 3,14 4,13 5,12
6,11 7,10 8,9
master test
5,12 6,11 7,10 8,9 1,16
2,15 3,14 4,13
1,16 2,15 3,14 4,13 5,12
6,11 7,10 8,9
Gas
HV
31
Chamber test in T11
32
Chamber test T11
Dialog -1
33
Chamber test T11
Offsets are corrected by 194 individual
thresholds. This will finally be done by DIALOG

34
Chamber test T11
All outputs were connected to the LVDS-ECL
converter with our final shielded twisted pair
cables.
35
Chamber Test in T11

• We used 45mV threshold (?6-7fC) on all 196
  channels.
• All channels had lt50Hz dark count rate.
• Excellent stability without dummy capacitor and
  without shielding !

36
Symmetric Termination
Due to the large detector capacitance the
frontend is extremely sensitive to ground
noise (Cdet100pF, 50?V fires the 5fC
threshold). With symmetric termination the chip
becomes immune to this effect. Penalty larger
noise ! Up to CARIOCA8 we needed this dummy
capacitor since the discriminator firing was
causing a large pulse on the chip ground. For
CARIOCA9/10, many measures were taken in order to
reduce this coupling, especially disconnection
of substrate contacts in transistors of the
digital part. With the final prototype things
work perfectly fine without the dummy capacitor,
but we still have this option !
37
45mV threshold on all Pads - Cathode Pad numbers
Capacitance 112 108 98 88
Threshold 7.6,7.4 6.9,7.0
6.7,6.6 5.9,6.2 fC
7.3,6.6 7.3,6.2 6.6,6.8 6.5,6.5
fC Noise 1.3,1.3
1.3,1.2 1.1,1.3 0.6,1.1 fC
1.3,1.3 1.3,1.1 1.1,1.1 1.1,1.2 fC
HV
38
45mV threshold on all pads wire pad numbers
Wire Pad Capacitances 26.5-28.5pF Thresholds
6.2, 7.4 fC Noise 0.67, 0.69 fC
HV
39
Cathode Efficiency
95 ? 2.42kV 99 ? 2.54kV
95 ? 2.45kV 99 ? 2.56kV
40
Wire Efficiency
95 ? 2.43kV 99 ? 2.55kV
95 ? 2.4kV 99 ? 2.5kV
41
Detector Capacitance

• The cathode pad capacitance in the entire muon
  system will not exceed ?120pF, so with the M3R1
  chamber we have already tested ? the largest
  cathode capacitances !
• We will however have wire pad chambers with
  capacitance up to ? 220 pF (R4) while the M3R1
  chamber has only 30pF wire pad capacitances.
• Since we dont have a wire pad chamber we
  measured the efficiency by adding capacitors to
  the wire pad.

42
On Chamber Wire pad Noise
packaged and non packaged chip
43
Wire Pad Efficiency for different Capacitances
nonpackage side
package side
44
Efficiency

• 2.5kV is a good working point that gives gt95
  efficiency on the double gap (gt99 on the quad
  gap).
• 2.65kV is a good working point that gives gt99
  efficiency on the double gap.

45
Crosstalk
Crosstalk Probability of firing the Neighboring
pad (infinite time window) Plot presented at
the PRR Measured with CARIOCA9 on the
M1R3 Prototype on Pad Position P9,7/10. We
decided to use doubel cathode readout since we
can survive with 10 crosstalk. In M2M3R1R2 the
trigger granularity is given by the wire pads,
not the cathode Pads. Crosstalk only
increases the rate.
46
Crosstalk
CARIOCA9, single, thr 4.7fC, position
P9,7/10 CARIOCA9, double, thr 6.8fC, positionP9,
7/10 CARIOCA10, double, thr 7fC, position
P11,7/10,no package (S1S2) CARIOCA10, double,
thr 7fC, position P11, 7/10, package (S3S4)
CARIOCA10, double, thr7fC, position P13, 3/14,
no package CARIOCA10, double, thr 6.7fC,position
P13,3/14,package (S3S4) ? ??????????????
47
Crosstalk
The preamp input stage was actually changed for
CARIOCA10 in order to improve the signal tail at
large capacitances (phase margin). The design
value was 50 ? since from simulations we know
that this is a good value (ASDQ used 25 ?).
48
Crosstalk Fraction

• Injecting a delta signal in one pad finds a
  signal on a neighbour pad.
• We call the ratio of the two pulse heights the
  crosstalk fraction.

49
Crosstalk Fraction
?1.7 ?1.4?1.5 1.8 ?1.5 ?1.4?
?1.7 ?1.5?1.4 1.7 ?1.5 ?1.4?
CARIOCA9
CARIOCA10
?2.1 ?1.7?1.6 2.2 ?1.6 ?1.7?
?2.2 ?1.6?1.6 2.1 ?1.6 ?1.7?
HV
50
Crosstalk Fraction

• The crosstalk fraction of the M3R1 chamber using
  CARIOCA10 is 1.6-2.2.
• It is 10-30 larger than for CARIOCA9.
• This is a small increase and the 2.2 crosstalk
  fraction is well within our specifications.
• Some time ago we found that we have gt95
  efficiency if our threshold is at lt30 of the
  average signal and gt99 efficiency of our
  threshold is lt20 of the average signal (1.5mm
  pitch).
• With a crosstalk fraction of 20 and 2.2
  crosstalk fraction there is no way to have such a
  large crosstalk !

51
Crosstalk
Simulated Pulse Height Spectrum MEDIAN is at
50. Crosstalk is defined as the probability That
a neighbor pad fires. This depends on Gas Gain
and threshold. Crosstalk Fraction is defined as
the fraction of Pulse height on a neighbor
pad. This is defined by the pad-pad capacitance
and Can be measured in the lab.
52
Crosstalk
Threshold (fraction of MEDIAN) 1 2 3
4 5 6 7 8 9 10
11 12 . . . . 20
53
Threshold Calibration
At the point where the hitefficiency is 50, the
threshold is at the MEDIAN Pulse Height. The
voltage where the double gap efficiency is 95
marks the beginning of our plateau. The voltage
where the double gap shows 99 efficiency is
difficult to find. Therefore we define it as the
voltage Where the single gap efficiency
exceeds 90. Knowing the gas gain curve allows
To define the threshold in terms of Fraction
of the MEDIAN signal. Easy to obtain !
54
Gas Gain
No space charge effects up to 2.75kV Gas gain
doubles for ?V of 106V
55
Crosstalk
gt95 efficiency if the threshold is at lt15 of
the median signal. gt99 efficiency if the
threshold is lt7 of the median signal. At
2.65kV ? threshold is at 5 of the
median signal At 2.75kV ? threshold is
at 2 of the median signal
!!!! The large crosstalk is real ! The double
cathode Readout and the change from 1.5mm to 2mm
pitch brought us to the edge of the
specifications !
56
High Rate Tests
Inefficiency due to signal pileup. Since the
muon trigger uses a 5 out of 5 coincidence, each
of the 5 stations has to be gt99
efficient. Therefore the signal width is a
crucial number. It is not only determined by
the electronics, there is a detector intrinsic
Dead time due to arrival of the electrons.
Since we use an OR of two frontend channels
per station, the rate of correlated hits is the
crucial number. For uncorrelated hits, we
still have 99 efficiency per station even if
one frontend (double gap) has only 90
efficiency. Out goal is a dead time of
lt50-60ns. In addition to the deadtime
(geometrical) we have of course some baseline
fluctuations
1.5mm pitch, Arrival time of the last electron
is 25ns.
57
High Rate Tests
Positive Amplifier
Negative Amplifier
Am241 is definitely a worst case background
signal (60keV gamma)
58
Charge/Hit at GIF
Cs 137, 662keV gammas Dividing the total chamber
current by the count rate at 7fC threshold.
The MIP charge is calculated by assuming 100e-/cm
and a measured gain curve, It is not a very
reliable number ..
59
TDR Numbers assuming correlations from LHCb
2000-089
Worst case behind the Calorimeter 870kHz
Cathode, 1150kHz Wires

Station 1 doesnt even work on paper
60
High Rate Tests at GIF
In the experiment we will have high energy muons
in presence of photon (electron)
background. The ideal situation is the muon beam
at GIF. We didnt have time to do this test
next chance only may next year. There is
another way of testing the high rate behaviour
I.e. signal pileup and baseline fluctuations
S-curve in presence of the background particles.

61
High Rate Tests at GIF
Chamber was positioned very close to the
source. Threshold set to 7fC like in the T11
testbeam.
At out working Point of 2.5kV we find exactly
the maximum rates expected in the experiment
(behind Calo) 1255kHz wires 920kHz
cathodes Rate increases with HV because of the
Compton spectrum The steep increase on the
Cathodes is due to the crosstalk
62
S-Curve at High Rate
Inject a signal delta signal on the pad and count
the coincidence of the Chamber output signal
with a correlated 20ns gate. With source off one
gets the standard S-curve. With source on one
gets all the information on rate, efficiency and
baseline fluctuations.
Efficiency
Baseline
Derivative gives the noisebaseline Fluctuation.
Rate
63
S-Curves at GIF
0, 2.5, 2.65 kV
0, 2.5, 2.65 kV
64
Efficiency
Efficiency
gt99 efficiency at low rate
About 4 geometric Efficiency loss at 2.65 kV
and 1.5MHz ! Compatible with lt50ns deadtime. !
65
Noise
Derivative of the S-curve gives the Baseline
Probability (NoiseBaseline fluctuations)
Noise increases slightly with the rate. Has
to be evaluated more carefully
0, 2.5, 2.65 kV
66
Conclusions

• Up to now, CARIOCA10 works according to
  specifications.
• Still missing Analog shapes, test pulse feature,
  input resistance, radiation tests, input
  protection, large pulse baseline recovery, .
• DIALOG will arrive ?Feb1st 2004 ? There is still
  enough time for CARIOCA tests.
• In case we will find a problem on CARIOCA10 we
  will have to consider a submission in Q1 of 2004
  which would shift our milestones but would not
  kill us.
• Francis Anghinolfi agreed to do the design
  changes in case it is necessary.
• The DIALOG still contains the ASDQ features, I.e.
  we are still free to chose ..
• We have to understand M1 and all background rates
  much better
• We should by no means exceed 1MHz rate/fronted.