Testing procedure:
-----------------
Each WFD card was first visually inspected and the resistances between
the power lines and ground were measured.
The cards were then subjected to a "smoke test" wherein they were powered
in a setup in which we could measure their current draws.
After this, the heat-sinks and daughter-cards were mounted, the
cards were plugged in the VME crate and the following tests were carried
out (A few tests were added to the ones discussed at the Corpus Christi
meeting):
(1) RAM test (A) in which patterns 0x0000, 0x5555, 0xAAAA, 0xFFFF were
written into all the memory locations for a channel and verified by
reading out
(B): a unique number was written into each memory location
and read out.
(2) Threshold Setting test: The four inputs to a channel had their
thresholds set at different values. The thresholds were measured with a
DVM on the daughter-card pins.
(3) Threshold monitoring test: Data was taken with no input and zero
suppression (ZS) off. The discriminator bits were tested to see which
inputs fired. The number of times an input discriminator fired was
measured as the thresholds were incremented. The 50% and 90% threshold
cutoff values were noted.
(4) ASIC and ZS off test: Data was taken with thresholds set at maximum,
ZS off and no input. We checked that all discriminator bits were
0 and the timeword decremented by 4.
(5) Rollover Timeword test: The clock frequency was varied over the range
180 - 230 MHz. Data was taken with no input, thresholds set high and ZS on. The
timewords were checked for 10 data-taking trials. The highest frequency
at which all 10 trials had correct timewords was noted.
(6) Data taking tests:
(A) Noise measurement: Data was taken with no input and no thresholds
set. The rms deviation of the FADC counts was calculated.
(B) Timeword testing: The input was a 1.5 V negative pulse, 70 ns
wide and with a period of 650 us. We checked the data for correct discriminator
patterns and the presence of two rollover timewords in between pulses.
(C) FADC count test: The input was a 5.4 V pp, 900 ns ramp signal
with. The data was plotted to check for consistency.
Observations:
------------
Quite a few problems were noticed at the visual inspection stage. Incorrect
resistances on the -2 V and +5 V lines were traced to shorts between adjacent
pins on the ASIC and were fixed. Some cards had diodes, transistors or
electrolytic caps in backwards and these were corrected. A few cards had
reversed or wrong chips and they were sent back to TI. (There were 5 such
cards). Apart from this, solder bridges between bent pins and vias, unsoldered
pins etc. were noticed at this stage and were touched up.
On many channels we noticed that the highest frequency at which the rollover
timeword is correct (HGF) is below 200 MHz.
# of channels
HGF (181-190 MHz) 2
HGF (191-200 MHz) 18
HGF (201-210 MHz) 28
HGF (above 210 MHz) 36
On 8 channels, the HGF was only a few MHz above 200 and on these cards, the
rollover timeword seen in the data-taking (timeword) test was occasionally.
incorrect.
All the new cards fared much better in the noise measurements compared to
the old prototype board. With our setup, we used to see about 6 counts pp
noise on the old prototype. On the new WFD cards, the noise is about 2 counts
pp. If we read out 5000 bytes of data for no input, the rms deviation in
FADC counts is less than 1 for all channels tested.
Here, we shall define a "good" card as one that passes all the above-mentioned
tests on all channels (except that it may show incorrect rollover timewords at
200 MHz.)
Subtracting the 5 cards sent back to TI, we found 22 "good" cards out of 34.
The 12 "bad" cards have an assortment of problems which are described in more
detail below. There were 6 cards with problems on only one channel.
The most common problem was a failure in the RAM tests. This happens on 8
of the bad cards. The other problematic cards include one in which zero
suppression cannot be set, one which cannot be STOPped, and two cards
which have data bytes occasionally low (by 8 or 16 counts.) Some of
these problems have been traced to problems on the ASIC or in a RAM
chip or swapped PALs.
On the whole, the WFD card testing is proceeding smoothly.
- Ashutosh Sanzgiri