> >
> > Capacitor in the Fanouts
> > ========================
> > Work-bench tests we've performed during the last two weeks have indicated
> > strange behaviour of the fanout channel that was modified with the 1mF cap.
> > This involves reduced pulse amplitude and 2mv p-p oscillations; none of
> > them were observed before. We believe both have to do with the soldering
> > of the cap and not with a systematic effect. As a cross check we plan to
> > check ASAP the modified channels currently in MACRO.
> The soldering!? Then why can't you just retouch the solder and get
> rid of the problem? I don't understand this suggestion at all. Is it
> just the usual grounding problems instead?
Of course we redid the soldering since the first moment we suspected
the capacitor's unstable behavior being attributed to that!!
I hope it's clear to everybody how hard it may become when it comes
to look at 4mV signals through our fanouts and understand what is going on
at the level of fine details. There are many factors that come into play.
We have paid close attention in making sure where we are plugging the
power of the scope, how fast we drive the pulser, how realistically we can
reproduce an spe withe the pulser and all the little details that may put
the measurement in question.
Today we modified another fanout channel in the test setup in order to have
another point of comparison. In all the tests we've performed in the test bench
the modified channel had at least twice p-p ground (>=2mV) noise compared to
the unmodified channels. In addition, the spe signal becomes "blurred" compared
to the spe's out of the unmodified fanout channel. "Blurring" implies
spreading in both pulse height and width, most probably a result of signal
dispersion. For an spe-like signal generated by the pulser (4mV/13ns FWHM,
as seen on an unmodified fanout channel) the modified channel was
giving out an spe distribution with mean amplitude 4mV and spread ~0.5mV
while the width had a spread of ~4ns. Furthermore, at higher pulse
amplitudes (e.g. 70mV) we've observed a maximum attention of ~4% compared
to the unmodified channel. Notice that this attenuation is most probably
present also at 4mV but remains unobserved since it is buried under
the spread of the pulse amplitudes.
In order to understand if this is a feature of the 1mF caps we are using
or a general feature, we stopped the run today in order to make a similar
measurement on the 4 channels in the detector that are already modified
(3B01-0,3B02-0,3B03-0,3B04-0). All of the above have been also observed.
> >
> > Capacitor in the Disc. Circuitry & SPE's Disc. Efficiency
> > =========================================================
> > We performed a number of test-bench measurements over a grid of pulse
> > width/pulse amplitude values in order to fine tune this capacitor value.
> > In all of these measurements the LED/PMT set up in the dark box was used
> > to create real-like (with p.e. stats) PMT pulses. In the table below you
> > may find the MINIMUM PMT pulse height below which we were starting losing
> > parts (on the WFD discriminator) of the input pulse. The PMT signals were
> > observed on the digital scope and the *mean* measured the pulse is reported.
> >
> > 0.05uF 0.1uF 0.2uF 1uF (original design)
> > -----------------------------------------------------
> > 5usec 30mV 15mV 15mV 12mV
> > 10usec 70mV 25mV 18mV 13mV
> > 20usec >200mV 50mV 18mV 15mV
> > 30usec >200mV 100mV 45mV 20mV
> >
> > Notice that interruptions in the WFD discriminator signal does not necessarily
> > imply complete loss of input signal (due to signal pre/post-sampling).
> > The situation is significantly improved with the 0.1uF capacitor. Larger
> > capacitor values (>=0.2uF) can do even better BUT they increase the time over
> > which the negative discriminator stays "on" to values close to our hardware
> > limit (~100usec). If we really want to increase this capacitor we have to go
> > to >>1uF (40uF?) capacitors that will do stay "on" longer BUT the overall
> > overshoot amplitude will never exceed threshold.
> >
> > It was rather a surprise for us the fact that a 30mV (mean) PMT pulse
> > followed by the (0.01+0.05)*125 uFohm=7.5usec discriminator RC circuit
> > couldn't stay above threshold for its entire 5usec duration. This led
> > us to the question of EFFICIENCY in discriminating spe's by the WFD
> > discrimination circuitry. We have performed multiple tests that demonstrated
> > that at our nominal 4mV/spe gain, spe's are not discriminated with 100%
> > efficiency. It is true that spe's are described with distributions
> > both in height and width. Our measurement show that over the entire
> > spe spectrum, 100% efficiency is achivied around 5mV pulses.
> > At 4mV the efficiency is at about 80% and at 3mV about 60%.
> >
> Is this with the 1 mF capcitor installed or not? This is a *crucial*
> point since that is nominally the thing which really made people like
> me think that electrolytics wouldn't work in the first place. With the
> one that I looked at when I was in Italy, I could see no attenuation
> to fast pulses... but maybe the one you are using here is creating
> attenuation (dispersion)?
The table showing the minimum average pulse height for a given pulse
width in order to have uninterrupted digitization of the PMT signal
was taken with the 1mF capacitor installed in the fanout. THIS IS THE
SAME CHANNEL YOU WERE ALSO LOOKING AT DOUG WHEN YOU WERE IN ITALY.
> > both in height and width. Our measurement show that over the entire
> > spe spectrum, 100% efficiency is achivied around 5mV pulses.
> > At 4mV the efficiency is at about 80% and at 3mV about 60%.
There have been at least 4 independent ways we've used to study the SPE
discrimination efficiency by the WFDs. The numbers quoted aboved were obtained
by firing the portable pulser at a fixed rate (~7.5kHz) and forcing
the WFD stop/log of data. Several pulse heights/widths were exercized
on AN UNMODIFIED WFD/FANOUT channel in MACRO. Based on the pulser rate,
we were expecting to find 8 "reference" pulses in every 1msec WFD frame.
Another study --that did not actually yield a solid efficiency number-- involved
firing the far LED in a tank so that to *marginally* (to the extend that this
is possible) trigger the TOHM. By checking the WFDs offline, we could see how
many counts the LI should had found given a WFD. There were several examples
where this was less compared to the LI's nominal threshold. The caveat in
this measurement is that it is rather cyclic: you use TOHM to check the WFD
and you've used the WFD to check the TOHM. Anyways, if you believe in the
11 count threshold for the LI, then you may find many waveforms where
TOHM fires marginally that do not justify the trigger. For the above two
kinds of studies we've taken actually MACRO RUN data; this will allow us
to look at the more carefully in the future.
The third study of the problem was done directly on the WFD card (by-passing
the fanout). Chris injected spe-like pulses over a grid of widths/heights.
He then used a scaler to find what fraction of the injected spe's were
actually creating a discriminator output. This study yielded that our
discrimination efficiency doesn't turn on before 3mV while it rises to 100%
*very* fast between 3.5mV and 4.5mV (Chris, you might want to forward
to the list that e-mail of yours...).
The last study refers to the original setup of the capacitor in the RC
preceding the discriminator fine tuning where the effect was noticed.
I hope the numbers make more sense now.
--Erik