In the last couple of days, we have learned quite a lot about what will be
required in order to completely fix our problem of losing waveforms from events
with large-amplitude and wide pulses. As various improved "fixes" have been
implemented, the next layer of the onion has been revealed twice now. We
think we may finally be near the core. We now understand that there are at
least three and maybe four different problems which contribute to our loss of
large-charge waveforms:
1. Positive overshoot resulting from the capacitive coupling out of the
fanouts. For big signals, this fires the positive discriminator in the
WFD.
2. Negative undershoot resulting from the fact that large negative signals
are truncated by a diode going into the discriminator circuitry on the
WFD daughter card. The function of the truncation is to reduce cross-talk
between WFD channels... a good idea which probably shouldn't be toyed
with. The fact that the negative signals are truncated but the (smaller
amplitude but wider) positive signals are not results in a net positive
charge deposition on the capacitor which couples to the input of the
discriminator. This discharges with negative undershoot which causes the
negative discriminator to fire.
3. An increased rate of small amplitude real signals (at the level of single
to few pe's) from the PMT's for up to about 200 microseconds following
a large amplitude pulse. There appear to be two possible forms of this
which could indicate two different mechanisms. First, there are an
increased number of real pmt pulses which seem to have a structure
suggestive of afterpulsing (due to Cs and Sb?), ie. there is a broad
peak in the arrival time of such pulses starting tens of microseconds
following the initial event and ending after about 200 microseconds for
sufficiently big pulses. I haven't gone through some time analysis
that shows that such a very long time is reasonable. We have verified
that this is definitely a feature of the PMT and not some external
light source. The second observed form occurs only for *very*
large pulses (saturated output for 20+ microseconds for instance) which
seems to actually induce a type of "exponential decay" of rate of real
pulses which has been observed to last for at least 1 millisecond. We do
not yet understand if this is a feature of the phototube or might be
due to an external light source (such as the scintillator).
The first two problems are the most extreme in causing loss of waveforms for
large-charge pulses. Fortunately, we do think that we have now identified a
reasonable approach to fixing this for all waveforms of interest in MACRO. The
fix involves increasing the series coupling capacitor out of the fanout to 1 mF
(electrolytic capacitor) and decreasing the series coupling capacitor in the
discriminator circuitry to give a time constant of about 5 us. The
combination of these two fixes will take care of all
overshoot/undershoot problems imaginable for MACRO data. (We still
need to verify this for a complete grid of amplitudes and widths of
pulses but the first looks indicate this is indeed a complete fix).
The increase in real pulses following a large pulse is an intrinsic problem
with which we will have to learn to live. Fortunately, it appears that there
remains no danger to any expectations for reasonable slow monopole pulses due
to this source of extra digitization. Rather, only much larger pulses as might
result from nuclearites should be lost due to this source of noise. This
suggests an approach to keep MACRO sensitive to a full range of signals.
If we stop the digitization time at 100 us for *very* large observed charge
and keep it at 1 ms digitization time otherwise, we should see everything of
interest. We should be able to implement this fairly easily using the
extra output of the CSPAM fanin and a bit of logic. This will be investigated.