Agenda for SCT links meeting
14:00 Wednesday, December 18
room 40-SS-B01
20' John Wilson - irradiation studies of LEDs and PINs.
20' Juerg Beringer - LED/VCSEL irradiation studies.
15' Stefan Marconi - STP noise tests - status and first results.
10' Tony Weidberg - Status of DORIC/LDC.
15' Tony Weidberg - Plans for realistic data links in the SCT (for TSP).
*********************************************************************
John Wilson presented work done by Stephen Oglesby on irradiation
studies of 3 GEC packages (consisting of 2 "GEC" LEDs and one PIN
diode). The packages were irradiated to a fluence of 6*10^13 p/cm^2;
the fluence was limited to avoid ruining the devices, as they are
needed for the 1997 TSP optical links.
LEDs:
Data were shown for 4 LEDs. Pre-irradiation light output for the LEDs
varied from 1 to 9 microwatts of optical power with a 10mA drive
current. The devices were biased with 10mA during irradiation, and
held at -7 degrees C. Post-irradiation values dropped to about 10% of
initial light output. After removal from the PS they were kept at
room temperature and annealed at 25 mA for 8 days. After annealing
the LEDs recovered to 15-30% of initial output, with the better
performance associated with the LEDs with higher optical power.
Extrapolating these results to a fluence of 2*10^14 p/cm^2 one expects
a degradation before annealing to about 2% of initial light output,
consistent with the Bern results.
PINs:
3 PIN diodes were also studied. They were irradiated with a -10V bias
at -7 degrees C. The response of the pin to light from an ABB-Hafo
LED (which was not in the beam) was measured. A power meter was used
to check the stability of the ABB-Hafo LED. The responsivity of the
LED in A/W was determined from a linear fit to the PIN current versus
incident optical power. The measurements of current versus power
showed deviations from linearity which need further study. The slopes
of these linear fits were then studied as a function of time/fluence.
The 3 devices showed rather different behaviours. More study is
needed. However, one can set a lower limit on the degradation of
responsivity after annealing at 0.2 A/W, which is 40% of the expected
initial value. Unfortunately, the initial values for these PINs were
not measured due to lack of time before the irradiation run.
The Birmingham group proposes a neutron irradiation of these devices
at the Dynamitron in January.
* * *
Juerg Beringer presented a large volume of data on irradiation of GEC
PINs and VCSELs from Sandia, a U.S. government laboratory. These devices
were measured using the measuring machine at Bern, which has been
previously described. The irradiation of the devices described below
took place at room temperature.
LEDs:
There were 88 GEC LEDs irradiated under proton fluences of 2 -
4*10^14/cm^2, and 64 under neutron fluences of 3 - 10*10^14/cm^2.
Half of the devices were pulsed at 20mA during irradiation, the other
half were unbiased. The GEC LEDs irradiated at 2*10^14 showed
initial degradation of RLO (relative light output) to about 2%.
Annealing studies were carried out at a variety of currents. The
optimum annealing current seems to be 50mA; LEDs annealed at this
current recovered to an RLO of 20+-5%. Average RLO values for LEDs
annealed at 20, 30, and 40 mA were 4%, 5%, and 9%, respectively. A
degradation in LED performance was seen when currents of 80-100 mA
were used. Devices irradiated at 4*10^14 showed almost no light
after irradiation, and almost no annealing so far at 50mA; measurements
are continuuing.
VCSELs:
VCSELs in 20-fold arrays (of which 2*8 can be operated in the Bern
scanning machine) were obtained from Sandia. There were 4 different
production runs, giving VCSELs with threshold currents of 0.5 mA, 2mA
(2 runs), and 10mA. There were 144 devices irradiated under protons
and 112 under neutrons, with fluences as for the LEDs described above.
Half of the VCSELs were pulsed (at 2 or 4 mA) during irradiation, the
other half were left unbiased. The effect of irradiation is to raise
the threshold current. The pulsed VCSELs showed only slight
degradation, even after 4*10^14 p/cm^2. Those unpulsed showed rapid
annealing to nearly full RLO. The problem of coupling the VCSEL light
to a fiber has been studied. Initial results were poor; the amount of
light coupled to the fiber varied strongly with details of fiber
routing. Two effects have been found: the intensity of the light
emitted by the VCSEL varies with angle/position, and this pattern
varies with drive current. The total light emitted, however, seems to
scale with current; it is therefore critical to couple all the light
into the fiber. The solution found at Bern is to bring the VCSEL into
contact with the fiber (after aligning to 20 microns or better). A
second problem seen involves back reflections into the VCSEL cavity.
These can be controlled with a filter. The performance with these
modifications is good. It is clear that packaging developed for the
use of VCSELs must take these effects into account.
Lifetime tests:
Two GEC LEDs (at 2.2*10^14) and one ABB Hafo LED (at 7.4*10^13)
died during or after irradiation. Long term studies of LEDs and
VCSELs have begun. The devices are held at 50 degrees C and DC
biased. The number of devices in the study are 144 ABB-Hafo LEDs,
56 GEC LEDs, and 8 VCSELs (of the 2mA threshold variety). After
roughly one month of test no degradation has been observed. The
aging acceleration factor has yet to be calculated, but is perhaps
of the order of 100.
Information on the irradiation studies in Bern can be found at
http://www.lhep.unibe.ch/atlas/ledlife/welcome.html
* * *
Stefan Marconi presented the status of the twisted pair option. The
PC board for performing noise and transmission tests was designed by
Peter Lichard of the TRT group, and was delivered one week ago.
Random data are generated at half the clock frequency (i.e. a 40 MHz
clock leads to 40 Mb/s of generated data). The data are sent through
LVDS drivers over 10 meters of stp cable, received, and compared to
the (delayed) data sent across the board. There are 4 bundles of 24
stp cables forseen, one of which was fabricated in time for the
meeting. The board has a provision for introducing AC and DC common
mode signals between the driver and receiver circuits.
After some initial debugging the board works correctly at 40 Mb/s, and
should work at 80 Mb/s with better termination of the on-board signals
and better delays. An initial bit error rate test at 40 Mb/s yielded
BER<10^-11. Studying scope traces of the received data suggest that
80 Mb/s transmission will be feasible. Tests to be performed starting
in January include BER, testing noise on unused cables in the bundle,
studying the noise radiated to SCT and TRT electronics (one will use
existing DAQ setups at CERN for these tests), and testing of common
mode effects.
Arno Beer has been working on the cable on connector specifications.
The connector chosen by the TRT has 66 pins, providing for 27 twisted
pairs, a common shield, and 6 low voltage power cables. The connector
is 4.5cm wide and 3.5cm long. The miniature twisted pair cables will
also undergo modification to improve the contact between the cable
shield and the drain wire, and to add an overall plastic coating to
avoid the unwravelling of the shield.
Pictures and further details on the noise tests, cables and connectors
can be found on the web:
http://wwwcn.cern.ch/m/marconi/www/public/stp/mainpage.html
* * *
Tony Weidberg gave the status of the DORIC and LDC (led driver) chips.
The chips will be submitted to the AMS 0.8micron BICMOS process (but
only npn transistors will be used) on January 22. The processing will
take 12 weeks, giving chips in mid April.
PIN diodes with faster rise times have been ordered. The PINs will
have a lower responsivity (roughly 2/3 of previous PINs) but should
have a 10-90% rise time of <1 ns at -5V bias. Several GEC packages
with 3 of these PINs (instead of the normal 2 LEDs and 1 PIN) have
been ordered, and should be ready for tests in February.
Five optical links will be provided for the 97 TSP program. They will
consist of (existing) GEC packages of LEDs and PINs, multimode fiber,
DORIC, LDC, a clock and control driver (to multiplex the signals going
down to the module) and a low noise transimpedance amplifier (obtained
from Pierre Jarron) with a separate discriminator on the counting room
end. A simple interface board will be necessary near the module, both
for connecting the low-mass power cable to 50-way cables and for mounting
the optical components.
A separate set of twisted pair links will also be provided for the TSP.
* * *
These minutes can be found from web page
http://www.cern.ch/Atlas/GROUPS/INNER_DETECTOR/sctnew/Electronics/links/