Freiburg Kapton Forward Hybrid (Kapton3)
* Hyrid/Module
Performance with no Detector
* Hybrid/Module
Performance with Detectors
Introduction
A Freiburg Forward Hybrid (Kapton 3) populated with 12 x ABCD2T
is currently under evaluation here at Liverpool, the intention is to construct
an ATLAS Forward Module (the readout chips will be wire bonded to 12cm
Silicon Detectors) which will then be used in a 'Forward Module System
Test' based at CERN. Performance of the module will be evaluated with and
without the Silicon Detectors bonded to readout.
The DAQ (DRAFT+SEQSI) is linked to a modified Melbourne Support Card
(Opto Hybrid) V2.0 for pseudo optical operation, this includes a BPM encoding
circuit (for the encoding of Clock and Command) with Kapton interconnects
between the support card and hybrid. The Doric and VDC are driven electrically
not optically.
Hybrid/Module Performance
with no detector
Operating Conditions
Vdd = 4.00V (measured on hybrid)
Vcc = 3.50V (measured on hybrid)
Bias Current = 267 uA
Shaper Current = 30uA
Strobe Delay = 28
Edge detect is enabled
Compression mode is Edge (01X)
The module is organised so that the top and bottom of the hybrid are
readout as individual planes into the DRAFT (due to constraints with the
DRAFT, the DAQ is restricted to only a single plane readout at anyone time),
their nomenclature is as follows:
Top |
M0 |
S1 |
S2 |
S3 |
S4 |
E5 |
Bottom |
M8 |
S9 |
S10 |
S11 |
S12 |
E13 |
Chip Trimming
Without trimming, Trimdac's are set to 0, the Vmean values for Qinj = 2.0fC,
are as follows:
Vmean after trimming for Qinj = 1.0fC and 2.0fC, the algorithm adopted
aims to trim the maximum number of channels with minimum spread:
Untrimmed and Trimmed Response (Qinj = 2.0fC)
Channel |
Vmean (mV) |
Vrms (mV) |
Channel |
Vmean (mV) |
Vrms (mV) |
Untrimmed Channels |
0 - 384 |
179.52 |
56.08 |
0 - 384 |
190.44 |
6.01 |
4 |
385 -767 |
193.39 |
51.93 |
385 - 767 |
190.97 |
6.38 |
8 |
768 - 1152 |
167.54 |
57.14 |
768 - 1152 |
190.37 |
8.45 |
28 |
1153 - 1536 |
167.86 |
49.91 |
1153 - 1536 |
189.37 |
13.59 |
25 |
As can be seen the action of trimming reduces considerably the spread
of the Vmean, from 57.14mV to 13.59mV (worst case RMS for both). Unfortunately
with the present trimming algorithm there is a trade-off between obtaining
the maximum number of trimmed channels against being able to keep the value
of Vmean uniform and consistent for the nominal value of injected charge.
Hybrid/Chip Characteristics after Trimming
Threshold Scans
Note that the horizontal axis (Threshold) is shown in DAC steps (2.5mV/step).
There is no evidence of oscillation at low threshold.
Gain and Input Noise channel by channel
-
Top and Bottom Planes,
Gain, Input Noise and Threshold Uniformity for Qinj = 2.0fC
Hybrid/Module Performance
with W21+W22 Detectors (Prelimanary)
Hybrid is bonded to W21+W22 pair detectors, strip length (approx) 65mm
and 54mm respectively, which gives a combined strip length of 119mm. The
initial bonding of the readout to detectors takes the following format:
-
Chips 0 - 3 (channels 0-512) are bonded to W21+W22 detectors
-
Chip 4 (channels 513-640) is bonded to W21 detector only
-
Chip 5 (channels 640-767) is not bonded to any detectors
-
Chips 6 - 11 (channels 769-1536) are bonded to W21+W22 detectors
Important: AGND and DGND must be linked together, there are bond pads
adjacent to each chip to allow this, for optimum performance.
Operating Conditions (as above)
Vdet = +100V
Idet = 4uA
Module is housed within a Faraday cage, the Faraday cage is linked to
AGND on the support card.
Module Characteristics after Trimming
-
Top and Bottom
Planes, Gain, Input Noise and Threshold Uniformity for Qinj = 2.0fC (bonding
pattern as above).
-
Gain, Input Noise and Threshold Uniformity for Qinj = 2.0fC (module fully
bonded to both W21+W22 detectors) for Top and Bottom
Planes.
Observations
The module does not oscillate!
The results clearly show that there is a large number of bad channels,
especially on the bottom plane. The primary cause of this problem is the
chip to fan-in bonding where neighbouring channels have been found to be
short-circuited to each other on the fan-ins. This is due to the close
packing of the bond pads on the fan-in and hence when a wire bond is made
there is a possibility of it short-circuiting to the neighbouring channel
(as is the case for a large number of channels). As of present this problem
cannot be rectified.
Return to my homepage
For comments etc. please contact Ashley
Greenall
Last modified: 28/07/00