Reliability Tests of New
VCSELs at
An
interim report
Maurice
Goodrick, John Hill & Richard Shaw of the Cavendish Lab,
Introduction:
An
initial batch of 50 TX PlugIns was delivered in October 2008: to distinguish them
from the earlier production, and because they have serial numbers starting at T2000,
we refer to them as T2K TXs. They have been extensively tested for performance in
the Cambridge QA system and in the ATLAS experiment.
A
remaining and crucial issue is the long-term reliability of these devices, and
it was decided that 34 of them should be subjected to a long-term lifetime test
at their design current.
The
test system was assembled from spare ATLAS SCT data acquisition modules within
9U (ATLAS-like) VME crate, using available fibre ribbons, controlled and
monitored by a customised version of the sct-daq software.
The
test was started just before the Christmas break, and has been running
continuously since.
System Description:
The
TXs under test are installed in 9 Back-Of-Crate modules (BOCs) in slots 5-12
and slot 14. There is space for 36 TXs, so the 2 spare locations will filled
with 2 old-series TXs, each of which is known to have a dead stream. All these
BOCs are Rev B, and are functionally nearly identical to the Rev C
BOCs installed on the experiment.
Each
BOC has a companion ROD in the front-of-crate location, and these provide the
control over BOC functions.
There
is a TIM in slot 13 to provide clocks.
A
single board computer (SBC) sits in slot 1, and this runs the sct-daq software
under Linux.
The
BOCs at slots 6, 9 and 12 are also populated with 4 RX PlugIns each: these
provide the means of monitoring the light delivered down the loop-back
fibre ribbons. As we have only 12 suitable ribbons, these have to be moved to
monitor 12 TXs at a time. This is done using 3 arrangements:
o Right:
- the 4 TXs in slot 5 are looped-back to the RXs at slot 6
- the 4 TXs in slot 8 are looped-back to the RXs at slot 9
- the 4 TXs in slot 11 are looped-back to the RXs at slot 12
o Mid:
- the 4 TXs in slot 6 are looped-back to the RXs at slot 6
- the 4 TXs in slot 9 are looped-back to the RXs at slot 9
- the 4 TXs in slot 12 are looped-back to the RXs at slot 12
o Left:
- the 4 TXs in slot 7 are looped-back to the RXs at slot 6
- the 4 TXs in slot 10 are looped-back to the RXs at slot 9
- the 4 TXs in slot 14 are looped-back to the RXs at slot 12
This arrangement and the TX and fibre ribbon serial numbers are shown in: Soak_Config
Control Software and Monitoring Procedure:
The
TxTest2 program is run to monitor and record the delivered optical power from
each of the lasers (as an average PIN current) in the particular arrangement
(Right, Mid or Left). The laser current setting used is always 10mA (DAC value
162). Note that the lasers are turned off and on during this test, but is
always left at the 10mA setting on termination.
As
the re-arrangement of the ribbons needs operator intervention, the program is
invoked manually. The standard procedure, which is repeated approximately every
other day, is:
o
With
the ribbon arrangement as left after the last test (Right or Left), run
TxTest2, save output
o
Move
ribbons to Mid arrangement, run TxTest2, save output
o
Move
ribbons to remaining arrangement (Left or Right), run TxTest2, save output
o
Import
the data into the Soak Excel worksheets to make the information readily
accessible
Cumulative Hours and Run information:
The
hours of running and the timing of the monitoring are shown in: Soak_Hours
The
test has been running continuously since 16th Dec 2008, and it is
planned to keep it going for some months.
General Observations:
When
used with GRIN fibre ribbons, the T2K TXs show a coupling that falls off along
the array of VCSELs: thus laser 11 delivers typically half the optical power of
laser 0. This seems to be explained by a mis-alignment during assembly, and is
expected to be corrected for future production. This is clearly visible in the
plots of current as a function of laser number: Soak_Charts
The
fibre ribbon connectors can be difficult to fully insert, and this often causes
a significant drop in the coupled power. This is characterised by an apparent
fall in the total power delivered by the 12 lasers feeding the ribbon. We look
for this, and re-insert the offending connector; residual effects are probably
the largest uncertainty in the measurements.
Results to Date:
The
Trend
plots show whether any slow fall off or sudden drop in coupled power has
occurred: Soak_Trends
Stream
11 (of [0:11]) of TX T2211 failed between the 9th and the 12th
of January. This corresponded to between 577 and 646 hours of operation.
The
forward diode drop of the lasers of this TX were subsequently measured using a
Fluke 77 DMM, and the dead laser was seen to have a significantly lower Vf.
This is consistent with ESD-induced failure seen on other failed lasers. The
measured values can be seen in Soak_Diode.
There
is some evidence that some lasers increase their output power with time, as has
been reported by the Pixel collaboration. In our case the effect is not
entirely clear. Mostly any increase is common to all the lasers of a TX, and
could be explained by a gradual improvement in the extent to which the SMC
connector engages with the guide pins. In a few instances an isolated laser
shows an increase in output. The Trend plots show these shifts.
_ooOoo_