Agenda for SCT links meeting
10:00 Friday, September 13
room 1-1-025
10:05 Discussion of boundary between links and ROD - R. Kowalewski
10:30 Electrical properties of PIN diodes - T. Weidberg
10:50 LED/VCSEL radiation studies at Bern - J. Beringer
11:30 Status of DORIC and LED driver design - D. White
12:05 Updated costing - R. Kowalewski
12:20 Discussion of participation in TSP - T. Weidberg
13:00 Status of STP noise test setup - R. Kowalewski
13:15 lunch
***********************************************************************
The meeting began with the discussion of where the boundary between
the responsibilities of the ROD and those of the SCT links lies. This
discussion was motivated by some questions Alex Grillo had circulated
to the SCT links and ROD groups about two months ago. Since Alex was
present he summarized his concerns and goals regarding the structure
of the SCT electronics work. He sees the need for a contact person
for each bit of electronics in the SCT. These contacts are clearly
defined for the SCT links and RODs. He felt that there were several
items, however, which were hard to categorize at present, given the
uncertainties in what the hardware (especially on the ROD end) will
look like. These are the interface to the ATLAS TTC and DCS systems
and the encoding of the SCT command protocol. He forsees dealing with
these items separately from the links and the ROD until such time as
it becomes clearer what the hardware dictates, at which point the
organizational structure would be modified. He underlined that there
was no problem with groups working on the SCT links to also work on
other areas of SCT electronics. There was some discussion of where
the items listed above logically belong. While no conclusions were
reached, Alex was able to explain his approach to the electronics
organization and members of the links group were able to make their
various viewpoints known to him, which were the purposes of the
discussion.
* * *
Tony Weidberg showed some results on the performance of the PIN diodes
in the GEC package. Measurements of rise time versus bias voltage showed
the following:
Bias (V) 10-90% rise time (ns)
5 15
10 8
15 6.5
These measurements overestimate the true rise time somewhat due to
amplifier oscillations; more careful measurements need to be done.
The slow rise time at reasonable bias voltages is thought to be due to
carrier mobility. The rise time needs to be improved upon in future
versions, which essentially means that GEC has to purchase faster PINs
for use in their package. There are no new GEC packages currently on
order. Some discussion followed on the need to get faster PINs and to
perform irradiation studies on PIN diodes. It was not immediately
clear who would be able to test PINs in large numbers (the Bern
emitter tester is not set up to study PINs), and when faster PINs
would be found. The PINs currently in the GEC packages (there are a
handful available) can be irradiated as a first step; Birmingham has
expressed interest in doing this.
Action item: Birmingham to test PINs in GEC packages. Bob and Tony to
follow up on procuring faster PINs in larger numbers and identifying a
means of testing them.
* * *
Jurg Beringer gave a report on irradiation studies of VCSELs and LEDs
carried out at Bern. Since the June SCT links meeting Bern has
obtained 2 16-fold arrays of low-threshold (2mA) VCSELs and 3 16-fold
arrays of high-threshold (10mA) VCSELs from Sandia Labs, and 320 LEDs
from GEC. One array of each type of VCSEL and 56 GEC LEDs were
irradiated at the CERN PS to a fluence of about 2*10^14 p/cm^2.
The VCSELs produce far more light than LEDs (factors of 5 or more) at
a given current. The effect of radiation damage on the VCSELs is to
increase the threshold current, from 2mA to about 5mA for the low
threshold VCSELs. After annealing with 3mA for 2.5 days, the low
threshold VCSELs recover almost fully. The behaviour of the high
threshold VCSELs depends on whether or not they were pulsed (at 18mA)
during irradiation. Those pulsed showed a degradation in threshold
current from 10mA to 13mA, while the thresholds of those off during
irradiation went from 10mA to >20mA. In neither case was there
significant annealing after 2.5 days at 15mA current. The low
threshold VCSELs are the only ones under serious consideration due to
their low power consumption. The initial tests on radiation hardness
are therefore quite encouraging.
Problems were experienced with the VCSELs when trying to couple their
light into multimode fibers; significant amounts of noise were
observed and the signal amplitude at the receiving end of the fiber
was quite sensitive to details of fiber routing. Emmanuel Monnier
commented that similar problems were seen by the Marseille group in
coupling fibers to VCSELs, but that these problems could be dealt with
by bringing the fiber to within 50 microns of the VCSEL and using long
fibers (70m in their case). The length of fiber coupled to the VCSEL
in the SCT would be only a few meters to the first connector. The
coupling question needs further study. If VCSELs are chosen as the
emitters for SCT links a package needs to be developed.
The GEC LEDs have a much larger spread in light output for a given
current, and require a larger drive current (about 20mA) than the ABB
Hafo LEDs (10mA). The spread would be smaller if the LEDs had been
pre-tested by GEC. 56 GEC LEDs were exposed to 1.8 - 2.6 *10^14
p/cm^2. All showed large decreases in light output (down to 2% of
pre-irradiation levels). Annealing currents of 20mA to 50mA were
applied over a period of 9 days. While the best device recovered to
22% of initial light output, several remained at about 5%. Contrary
to expectations, the best device had been off during irradiation, and
the LEDs pulsed with 18mA during irradiation showed poorer behaviour.
These results are not encouraging. The annealing studies will
continue, at currents of 30mA, 40mA, and 50mA.
In a separate run, 56 ABB Hafo LEDs were irradiated with 8*10^14
n/cm^2. The LEDs were not pulsed during the irradiation. The
relative light output immediately after irradiation was <5%. With a
10mA annealing current the light output on one of the LEDs recovered
to 20% after 8.5 days and was still rising. Another LED recovered to
40% after a few hours of annealing at 30mA and remained steady
thereafter. Some anomalies were observed, however; a LED initially
annealed with 10mA up to 58% after a few days was then biassed with
30mA and showed a DROP in light output down to 35%.
It was pointed out that the variation in LED damage would require
individual tuning of driver currents.
Jurg concluded that
* VCSELs appear to be radhard and that reliability studies will begin
when a large number of devices is received (late September).
Coupling of VCSEL light into fibers needs further studies.
* Initial results on GEC LEDs under proton irradiation indicate they
are less radhard than ABB Hafo LEDs and require higher annealing
currents; however, the annealing is not yet complete.
* Neutron irradiation of ABB Hafo LEDs give similar results to proton
irradiation
* Long term studies will be done operating LEDs/VCSELs for months at
higher temperature.
* * *
Dave White showed the status of the DORIC and led driver (LDC) design.
A project specification is available on the SCT electronics pages on
WWW. The amplifier and comparator stages were shown, as well as a
block schematic for the whole chip. The large stray capacitance of
the PIN in the GEC package has made the design more difficult but Dave
feels that he has this under control. He has taken care to filter
supply lines as well. The simulation indicates the chip reaches stable
operation within 500ns after the clock is started. The physical size of
the chip will fit easily into the space reserved for it (2.0mm x 2.5mm)
on the optical hybrid board proposed for use with the GEC package..
The DORIC will be passive, and will use the 50% point of the leading
and falling edges to define the extracted clock (the use of leading
or falling edge to define the clock changes each time an odd number of
ones is sent in the biphase mark scheme). The driver circuitry off
detector must therefore be able to adjust these 50% points, accounting
for differences in rise and fall times for each channel. This could
only be avoided by having a phase-lock loop in the DORIC, which the
procurers of the chip refuse to allow.
The layout is proceeding. The chip will be designed in AMS 0.8 micron
BICMOS, but use only bipolar npn transistors, and is hoped to be
radiation tolerant. This will be tested once the chip is available.
The submission will be in late October, and the chip should be
available for testing in early 1997.
The design of the LDC is less well advanced. Submission should also
take place in October.
* * *
Bob presented an updated costing based on the same assumptions as
shown at the June meeting, but using the updated fiber prices as shown
by John Dowell in the June meeting, and including alternate quotes for
radhard fibers from Oxford. The numbers appear on the SCT link WWW
pages. The conclusion is that an all-electrical link is a factor of 2
or 3 cheaper than an all optical link. Hybrid electrical-optical
solutions, which are no longer being actively considered, are
intermediate in cost. The cost per module (without contingency) for
complete twisted pair links is approximately CHF 100, while for a full
optical system it is in the range CHF 225-340.
* * *
Tony presented a list of reasons for providing realistic low-mass
links for the TSP of summer 1997. There is a desire on the part of
the rest of the INDET community and of the calorimeters to avoid
excessive material in the TSP. In addition, this is an opportunity
for the SCT links group to gain experience working with links in a
real experimental environment. Steiner Stapnes was present and said
he expects to have low mass modules without heavy support cards for
the 97 TSP, and that he sees low mass links as being very desirable.
He said a working ABC was a critical item, but was possible. The 1997
TSP will probably consist of about a dozen modules.
At this point the desirability of providing links for the 97 TSP was
discussed. Most present agreed that we would gain useful experience
in doing so. There was a consensus that the main aim of this group is
to evaluate the link technologies and make an informed choice based on
radiation studies, noise tests, cost, etc., and that the TSP effort
should not compromise these studies. The timescale of the 97 TSP is
such that the links must be ready by June to allow time for testing in
conjunction with the modules and DAQ at H8 prior to beam. Given this
short timescale the components needed for the links must be obtained
soon. Any links made for the 97 TSP are therefore unlikely to be
identical to what will ultimately be used in Atlas. The final link
technology would presumably be used for the 98 TSP. Bob emphasized
that the 97 TSP will therefore not be the test to decide between STP
and optics.
The suggestion that a full set of optical links and a full set of STP
links be produced was well received. Tony then listed the bits needed
for an optical link which could be built in time for the 97 TSP:
* LEDs and PIN in the GEC package (available)
* DORIC and the LED driver chip, LDC (D. White design, submission Oct 96)
* an opto-hybrid board for DORIC, LDC, GEC package (Dave White)
* VME clock and control driver - adapt existing board (Roy Wastie)
* Fujikura rad-hard fibers (available)
* data receiver - sensitive dc-coupled preamp from P. Jarron's group
available (rise time 7ns). Separate discriminator needed.
* interface to bias card for data and TTC (Maurice Goodrick).
Tony is employed by CERN for the next year and will request
engineering help from CERN. In addition to these items the question
of whether one uses connectors in the TSP and whether one uses bare
fiber or ribbon cable are to be decided. Several people felt that
putting in realistic connectors and realistically packaged cables was
important. Tony will further investigate what resources are needed to
build these links. STP links can also be produced, consisting of fine
STP cables, a repeater, and standard gauge cables. Bob will look into
what is needed in detail.
Action item: Tony and Bob will report back to the group with more
detailed information about what is needed for TSP links.
* * *
Bob presented the proposed test setup for fine STP cables. The test
setup was designed by Philippe Farthouat and Peter Lichard of the TRT
group, and generalizes an existing test setup for bit error rate
studies on single low-mass STP cables to 4 bundles of 20 signal cables
plus 4 dummy cables each. The block diagram of the circuit can be
seen in a brief document which will be attached to the copies of the
transparencies. Provision is made for testing bit error rate, cross
talk, and radiated noise, and for generating both ac and dc common
mode signals between the data generation and comparator ends of the
circuit. The board will be fabricated at CERN and be available in
October.
* * *
These minutes can be found from web page
http://www.cern.ch/Atlas/GROUPS/INNER_DETECTOR/sctnew/Electronics/links/