> From: Erik Katsavounidis <Erik.Katsavounidis@lngs.infn.it>
>
> The fix you are proposing (reducing the time constant in the RC
> circuit before the discriminator) is very appealing. It is
> definitely cost-free and really fast to implement. However, I have
> my reservations as to the extend that it will affect our ability to
> record monopole waveforms. I understand that eventually, for slow
> monopoles, the p.e. statistics will likely work in our
> favor. However, probing a search where nothing can be excluded, we
> may need to allow as many options available as possible.
>
> Waveform Data Loss
> ==================
> The key issue here is what fraction of the waveform will be lost for
> *slow* monopoles of *low* pulse height. In figure 6 you plot the
> fraction of signal recorded versus pulse width AT THE CRITICAL PULSE
> HEIGHT; from this plot it is concluced that NO LOSS will be seen for
> pulse widths up to ~8 microsecs. However, the critical pulse height
> that is assumed here is maybe TOO OPTIMISTIC for a monopole at the
> 3-10microsec range and we already know that this effect (of data
> loss) is a LOW pulse height effect. For large pulses (like the
> CRITICAL PULSE HEIGHTs you assume) the pulse exp. decay with the
> (discriminator) RC time constant will keep the pulse height above
> threshold for most of the time, thus leading to no significat data
> loss.
>
> For that, I would suggest drawing figure 6 NOT at the critical pulse
> height BUT AT THE PULSE HEIGHT EXPECTED FOR A MONOPOLE OF A GIVEN
> BETA.
>
> Ioannis and myself modified one WFD channel according to the
> proposed fix and measured the data loss effect for some long pulses:
>
> --------------------------------------------------------------------------------
> #|Input pulse |Disciminator ON ==>DATA written |Discriminator OFF==>DATA LOST|
> #|Height Width |From /To (usec) & From /To (usec)|From /To (usec) |
> -|-------------|---------------------------------|-----------------|-----------|
> 1|200mV, 5usec |T=0 /T=5 |-- | NO |
> 2|100mV, 5usec |T=0 /T=4.5 & T=5.0/T>8.0 |T=4.5/T=5.0 | YES (10%) |
> 3| 30mV, 5usec |T=0 /T=3.2 & T=5.0/T>8.0 |T=3.2/T=5.0 | YES (36%) |
> 4| 15mV, 5usec |T=0 /T=2.5 & T=5.0/T=7.3 |T=2.5/T=5.0 | YES (50%) |
> 5| 8mV, 5usec |T=0 /T=1.7 & T=5.0/T=6.5 |T=1.7/T=5.0 | YES (66%) |
> 6| 4mV, 5usec |T=0 /T=1.0 & T=5.0/T=5.7 |T=1.0/T=5.0 | YES (80%) |
> 7| 8mV, 8usec |T=0 /T=1.7 & T=8.0/T=9.7 |T=1.7/T=8.0 | YES (79%) |
> --------------------------------------------------------------------------------
>
> Given that our expected slow monopole signatures will be at best the
> ones of test pulses #4,5,6,7 of the above table, we see that we will
> lose more than 50% of the monopole waveform for that width-pulse
> regime. Needless to say, for a (e.g.) 30usec wide pulse, the % of
> data lost will be close to 100 since the "ON" time does not depend
> on the pulse width but on the pulse height only.
>
> Effect of Waveform Data Loss to Data Analyses
> =============================================
> The question is how significant this data loss is and how it can be
> (if) recovered in a way that it will assist monopole data
> analyses. In page 4 of the memo it is stated that "we would still
> record the beginning of a wide monopole pulse and catch the exit
> time from the trailing edge, providing adequate information for
> making a beta calculation." The plot (figure 3) that follows,
> displays how this is realized. This plot leaves me a bit
> perplexed. The region marked as "LOST DATA" appears to be a bit
> after the "middle" (in the time axis) of the input pulse but
> *clearly* before its trailing edge. In this way, it suggests that
> part of the "~flat" (actually exp. decaying) region following the
> "LOST DATA" zone together with the trailing edge of the input pulse
> will be eventually digitized. My expectation is that the "LOST
> DATA" region extends ALL the way through the end of the input pulse
> leaving the digitization of the trailing edge in question as it will
> rely solely on the 15nsec pre-sampling of the digitization that will
> start sharply after it. In this scheme, what will DEFINITELY be
> digitized is any signal that will be made available to the FADC
> SHARPLY following the trailing edge of the input pulse. This is
> displayed in the above table under the "Discriminator ON" column
> with the time interval indicated immediately following the "&"
> symbol. This post-pulse signal may include any possible positive
> overshoot, ground noise, radioactivity decays or anything else you
> can imagine. If resulting from the original input pulse, this
> post-pulse signal is subject to an ADDITIONAL SHAPING coming from
> another RC circuit (R=1K/C=6.9uF) that preceeds the FADC, thus
> making any analysis of the post-pulse signal even harder.
>
> My first guess in the above scheme is that the waveform analysis for
> the monopole searches will become QUITE COMPLICATED. It might become
> very tricky to associate the beginning of any monopole-like pulse to
> its end. The catalysis signature will make things rather worse as
> singals separated in time of the order of microseconds may as well
> be confused.
>
> Effect on Slow Massive Nuclearites
> ==================================
> The critical pulse height figures implied by the proposed fix do
> allow slow nuclearite searches with ionization powers up to hundreds
> of Imin. However, there are still nuclearite masses (i.e. dI/dx)
> that will find the WFDs unable to record their resulted
> waveform. Maybe we should comment on that in this memo. A quick
> calculation shows that nuclearites at the hundreds of ng mass range
> will be fully covered.
>
> Effect on the Monopole Data in the Present Configuration
> ========================================================
> It is concluded in the memo that with the current WFD configuration
> we start becoming insensitive to monopoles at about beta=3x10^-3 and
> above; this might be too optimistic. I recall Doug estimating this
> to be at about beta~5x10^-4 and above, mostly coming from a
> measurement of his, showing that it cuts off at about 1Vx1usec
> signals. Ioannis and me later repeated this measurement on the WFD
> directly, finding practically the same limit. In the memo, the cut
> off pulse height for 1usec pulses is listed at 3V; for 0.3usec
> pulses this becomes 7V. For these two pulse-widths our work bench
> measurements gave 1V and 5.5V respectively. Is this an effect of the
> 100ns rise (fall) times you introduced to your pulser signals? Even
> so, the difference in the integrated charge between 0.3/1.0/2.0
> microsecs is a bit strange to me. I am expecting a monotonic
> function of the integrated charge with a step at about 0.5V, the cut
> off voltage of the zener-diode.
>
> BTW, in the summary figures 4 and 6, I had problems justifying the
> details of the bottom and top horizontal axes. For example in fig. 4
> a ~5usec pulse width is associated to ~10^-4 beta, something that I
> don't think is compatible with what is assinged as beta for pulse
> width of 0.3usec. Moreover, in fig. 6, a rather different pathlenth
> is suggested based on the product of width*beta. Aren't both plots
> drawn assuming 20cm pathlength?
>
> Alternative Idea(s)??
> =====================
> On Feb. 3, 1997, Ioannis sent to all of you an alternative fix that
> involves the replacement of the capacitors in the two RC circuits
> that the signal goes through before the discrimination: one in the
> BU fanout and another one in the WFD daughtercard. It actually
> points to the reverse direction; it suggests increasing both of the
> capacitor values. The proper choice of capacitor values will
> guarantee that any positive overshoot or negative undershoot
> accompanying the PMT pulse will remain bounded in the
> [-2.5mV,+2.5mV] region, thus never resulting to any data loss. It
> will actually improve the quality of the digitized signal as it will
> get rid off any exponential decay currently present in long PMT
> pulses. Our work-bench tests were NOT performed with what
> eventually can be used in a fix implementation. Capacitors of the
> right (physical) size and value (~1000uF and ~100uF) are available
> (found in product catalogs) BUT THEY WERE NOT available to us at the
> time of test. We invite you to test it on your work-bench. Cost and
> implementation are within reach (cost of <$2,000, implementation
> ~1week with 4 persons).
>
> The nice feature of this idea is that it has a straighforward
> "acceptance" protocol: all PMT pulses in the [0-10]V x
> [0-N]microseconds (where N>=17 depending on the choice of the caps)
> will be recorded. Anything lasting more than that will be lost,
> inversely proportional to its amplitude.
>
> Bottom Line
> ===========
> I believe we all agree that the proposed fix might come with data
> loss under some circumstances (long pulses-low ampls.). This might
> actually be a feature of *any* fix that anyone can come up with. The
> question is which has the least effect on our physics
> analyses. Although simplicity and straightforwardness of the fix is
> a major consideration, it should not be the only factor. If the
> analyses issues related to the proposed fix are agreed to be
> insignificant, then we can go ahead and implement it immediately.
>
> That's all for the moment; waiting for your comments,
>
> Erik (&Ioannis) Katsavounidis