Manfred Krammer <krammer@hephy.oeaw.ac.at>


Short time test - immediately after bonding

Using the modules built in Vienna we have done several tests and
measurements to check their functionality. We have done simple electronic
tests, laser tests and source tests. After comparing the results I believe
that the following 3 measurements are sufficient to prove the
functionality of the chips, check the bonds and determine the number of
bad strips or channels.

1. Readout all chips and record noise vs. strip
        See if all chips are responding and identify noisy channels

2. Use internal calibration to check the gain of all channels (loop
        through all channels)
        The precision is probably not better than 10%, but this is
	sufficient to
        identify bad channels (channels without amplification)

3. Perform a back plane calibration on both sensors
        Find missing bonds, one can distinguish between a missing bond
        between the 2 detectors and a missing bond between detector and
        chip by recording the pulse height.


The noise measurements and the gain scans were routinely done on both the
milestone module (8xAPV6) and on the Vienna-APV25 Module (3xAPV25) before,
during and after the last PSI test beam. Using our set up it took only a
couple of minuts and helped us to identify channels with problems.
The back plane calibration was tested on the milestone module. After we
have intentionally removed some bonds one could clearly recognise them in
the data. Again this measurement is very fast and can easily implemented
in any setup.
If there is interest somebody from Vienna could show some plots during one
of the next meetings.

A few words to the test system:
We use our compact test beam readout system to do these tests. The system
consists essentially of a PC and a VME crate equipped with the standard
modules as far as available.
I believe we shouldn't put too much emphasis on equalising the systems in
all CMS labs. We should rather define exactly what we want to measure. Any
system which is able to control the APV25 and the auxiliary electronics
and reads the quantities described above will do the job.


Burn in test

Naively I would guess that the tests described above are also sufficient
for the burn in test. In addition one should certainly monitor the
detector currents.
These set of measurements done before, after and maybe during some
temperature cycles would reveal all electrical failures.

At this point I would like to support the recent mail of Alan in which he
requests a special burn in for the hybrids. Also my opinion is that the
most likely failures will occur on the hybrids due to their complexity. As
we have foreseen a very sophisticated test scheme for the sensors the
probability for a sensor failure during burn in should be very small.
If the hybrid fails during burn in we have to go through a very difficult
and time consuming repair procedure or even throw away very valuable
items.

My proposal for the testing strategy looks the following:
        Testing of sensors as foreseen (longtime test if necessary)
        Test and burn in of hybrids
        Short time test of all modules after production (see above)
        Burn in of substructures (e.g. petals) using final cooling system