AFEI shortcomings and solutions

1
AFEI shortcomings and solutions

• Juan Estrada
• 4/13/2005

• Introduction
• Problems seen in AFEI
• pedestal variation
• SVX saturation
• discriminator to ADC crosstalk
• discriminator pedestal shift
• Conclusion

2
AFE and CFT
8 photons
VLPC
9o K
(x512)
50 fC
Central Fiber Tracker cylinder
DISCR
ADC
Discriminator output every 396 nsec for L1
Amplitude signal readout for L3 and offline
AFE
3
Occupancy in the detector
D.Lam
(online thresholds set to 1 occupancy between
stores) Things will get very difficult for
tracking at high luminosities.
4
Projected Occupancies
D.Lam
5
Light Yield

• These data show LY v. f for h0
• Average for each supersector
• Current worst case LY 8 pe (singlet)
• This is a lower limit due to SVX sat. ADC
  dynamic range
• Expect Axial LY Stereo LY
• Axial Gain gt Stereo Gain
• SVX Sat. Effects smaller in Stereo

O.Boeriu
6
VLPC Luminosity Dependent Effects
0, 10, 20, 30, 40 occupancy (expect close to
40 on inner layers at highest RunII L)

• As VLPC occupancy (or luminosity) goes, up QE and
  gain drop

Gain
Quantum Efficiency
20 drop
10 drop
in addition Fiber Radiation Damage induced
light yield loss (10-20)
7
CFT electronics and the VLPC(introduction to the
problem)
AFE module (analog front end)
DISCR
discr. output every 396 nsec for L1
if the event is accepted at L2 send the amplitude
of the signal to L3
ADC
Digitized signal to L3
1
64
L2 accept
SVXIIe (pipeline ADC)
AFEI is the combination of 2 custom chips
(SVXIIe and SIFT). Issues very small signals
transfered from SIFT to SVX while the
discriminators are firing (baseline shift)
Analog signal (gain 0.5)
(1..16)
Discrim to L1
SIFT(1..4)
(1..16)
(1..16)
8
Known problems in AFEI
9
Tick to Tick
Mean pedestal (ADC)
data from the detector
20 ADC/pe
The pedestal moves up to 1 pe for the high gain
pixels, this effect is much more significant in
the stereo layers This occurs because we reset
the front end once per superbunch.
10
AFEII proto Results
xing 5, 8, 11 , 14 , 17 , 20 all 64ch from
module 1

• Tick to tick variation in pedestals.
• Reset is identical every xing, so there should
  be no tick to tick variations and none is
  observed.

AFEII result
Tick to tick variations eliminated!
11
Channel to Channel variation
Spread with RMS0.5 pe is typical for AFEI. With
AFEI we can set only 1 zero supression threshold
per SVX chip (64 channels).
12
AFEII proto Results

• Chan to chan variation in pedestals.
• RAW ADC info (10bit) is converted inside the FPGA
  into 8bit SVX like data pedestal corrected and
  zero suppressed. So there should be none.

xing 5, all 64ch from module 1
AFEII result
Chan to chan variations eliminated!
13
SVX Saturation (1)
Event (photons produced)
Reset at the beginning of the superbunch.
crossing clock
1 crossing396 ns
?Q1
The voltage at the input amplifiers goes up, and
it can hit the maximum. After this point, the
electronics can not detect any more photons until
after the next reset (during the next gap).
?Q0
14
SVX saturation (2)
The problem was studied in the 4 cassette test
stand during summer 2004 (DØ Note 4495, P.
Hasiakos).

• At 30 occupancy (12 crossings)
• 4 Hits integrated in pipeline
• Typical integrated charge on front end 40 pe
• 40 drop in Signal for hits in triggered
  events, on average

15
AFEII proto Results

• Biggest concern is SVX saturation
• Test inject huge LED pulse, measure small pulse
  in the same superbunch

3pe in xing 20 -gt readout gt 100pe in xing 5
AFEII result
Saturation problem solved!
16
Discriminator-ADC crosstalkstudies in the test
stand
Discriminator to SVX crosstalk Discriminator
firing in AFEI produce extra noise and shift of
the analog readout.
0 discr occ.
Some pedestals shift to higher values (fakes).
32 discr occ.
Some pedestal shift to lower values
(inefficiency).
this is an event by event effect, what matters is
the discriminator occupancy for the event you
triggered on. Very difficult to solve. Analog
information compromised.
DØ Note 4500
17
Discriminator-ADC crosstalkconfirmation in DØ
data
Special runs were taken to study the effect seen
in the test stand using real data.
Ped shift.
SVX
Typical channel
Ped shift.
occupancy
SVX
This analysis was done by Avdhesh Chandra from
TIFR.
Chann. number
18
Discriminator-ADC crosstalkconfirmation in DØ
data (2)

• Using special runs without zero suppression, we
  were able to see the pedestal shift at 30
  occupancy for about 11K channels.
• The effect seen in the test stand is confirmed,
  the pedestal shift as a function of occupancy in
  the event has been verified.

• The magnitude of the effect (from gaussian fit)
  mean -11, sigma 10
• 1 pe is something between 10 to 20 ADC counts.
• The tails reach very far, but the 3 pe shift seen
  in the test stand is not typical things like 12
  pe shift are common.

19
AFEII proto Results

• Discr to ADC xtalk is not there is AFEII. The
  position of the pe peaks are always in the same
  place, and do not depend on the discriminator
  occupancy (compare with slide 16).

AFEII result
TriP results
discr. on.
Discr to ADC xtalk eliminated!
20
Discriminator pedestal shift (1)
Two discriminator scans with a fixed amplitude
LED signal. In one case (RED) the LED is pulsed
once per turn, in the other (BLUE) the LED is
pulsed in every beam crossing. When we expect the
discriminator to be firing at every crossing at
25 occupancy, we see only 15 occupancy when
firing on every crossing. This is equivalent to
0.5 pe shift.
No effect in the same crossing. Threshold scan at
2 different amplitudes.
21
Discriminator pedestal shift (2)
This effect depends on the zero-bias occupancy,
and it is smaller than what we have seen for the
Discriminator-SVX crosstalk. However, it is part
of our L1 trigger 0.6 pe could be a lot.
occupancy
We use a 0.6 pe (instead of the 0.5 in the
previous slide) to take into account the gain
reduction in the VLPC. As the occupancy goes up,
so does our threshold. Efficiency per hit to the
power of eight gives the tracking efficiency for
8/8.
L40E30
L200E30
22
Conclusion

• The current AFE has served well, but exhibits
  several features which limit performance
• Pedestal variation
• SVX saturation
• Discriminator to ADC crosstalk
• Discriminator pedestal shift
• Several of these undesirable features become
  increasingly important and the luminosity
  increases.
• We understand the origin of these problems and
  the proposed electronics addresses and solves
  these problems by design.