1. L1 MUON BARREL VCSEL’s OPTICAL LINK
Developed by ATLAS Napoli’s Group
500 links from on detector L1 muon barrel trigger to USA15
Unidirectional link on multimode fibre driven by 850 nm VCSEL
Based on G-Link HDMP-1022/4 and HDMP-1032°/4° chipset
Use of Agilent (now Avago) HFBR-5912 optical transceiver
MHz clock, 800 Mbit/sec serial stream on fiber

2. L1 MUON BARREL VCSEL FAILURE
We started “discovering” VCSEL failures at the begining of 2009
Up tp now we found 23 broken links
Failure rate was a bit higher then one per month, and decreased
a little after the decision taken in 2010 to switch off the link
when not in use
In all broken links same problem: no output light in the transmitter
12 broken links peggy boards were repalced with spare ones.
On other 8 links we have replaced the VCSEL (simple
but delicate operation). Remaining 3 links not yet repaired.
- We have sent 3 broken VCSEL parts to AVAGO for investigation

3. AVAGO INVESTIGATION RESULTS
-Most likely the braking is due to electro-static discharges
Most probably the damage had happened during the production in the
“assembly/test processes”
From 2005 on the company has constantly improved the production line
At the end of 2009 a zener diode was introduced to improve the protection
of HFBR-5912E against discharges up to a factor 20
Our lasers were produced in 2003 and 2004; this means that the lasers are
sensible to ESD’s
- In 2009 and 2010 we have replaced our broken lasers with some spare parts bought in the past and produced in 2007 (mid way between 2004 and 2009)

4. L1 MUON BARREL LINK NEXT ACTIONS
Napoli's group (A. Aloisio and V. Izzo) have measured the optical  spectrum of
all HFBR E used in the apparatus. The measure was taken at the end
of the fiber.
The results of measurements and the preliminary conclusions are illustrated in
the next slides
In the meantime INFN has financed 50 new VCSEL’s to be bought from 2010
AVAGO production (the upgraded one)
Exchanging a link means exchanging a PAD box. This is a boring job, although it is
not too much difficult (only few PAD’s are in a very difficult access position )
In addition INFN has also financed in a conditional way (“sub-judice”) 150
extra VCSEL’s, in case this will be really necessary.

5. the largest majority have “regular” spectra centered around 850 nm (typical variation between nm and 854 nm), max value -20 dbm and width (at – 50 dbm) between 1.5 and 2 nm

6. “fringed” spectra, more irregular respect to the previous ones
“fringed” spectra, more irregular respect to the previous ones. Max value -20dbm, centered
at a lowest value (between 842 nm and 844 nm) and width (at – 50 dbm) a little higher

7. “narrow” spectra similar to Liquid Argon broken links
“narrow” spectra similar to Liquid Argon broken links. More than 15% of the total. It is not clear
if they have problems. The LAr links were cut off from one side; here, with some approximation,
we could say that some of them are cut off. But, with some optimism, we could consider
these spectra similar to those of category 1), regardless their 1 ns width at – 50dm

8. PRELIMINARY CONCLUSIONS
In the literature apparently doesn’t exist any link between spectra of optical power and damage from ESD; and since we don’t have more information, the only reasonable thing to do is to wait for some breakings and to consult the database to verify wich kind of spectra had the broken links.
We could also repeat the measurements in 6 or 12 months from now to verify if it exists some sort of “evolution “of spectra.