Rich> I might suggest a minor modification to the monitor to ignore uVAX3 until
Rich> you trust the calibration constants.
BTW, I just added something to the monitor (SENTINEL) to allow a
microvax to drop out more cleanly: previously, SENTINEL would
continually try to relaunch an RCD process if the microvax was not
responding. Also mini-acq data would get accepted by the RCD process
and would have to be filtered out by the detached analysis process.
I have now improved this: now SENTINEL checks a file to see if a microvax is
supposed to be present or not, and does not try to relaunch the
process for an absent microvax. In principle, things should work
effectively the same as before, but this way is more efficient and
avoids the annoying mini-acq run number problem.
> Just forwarded a mail Rich sent the other day about how he'd deal with monitor
> calibrations. This is more of an issue now than I'd thought, as it turns out
> that laser calibration runs haven't been processed in ages, since they are
> tough to do and the upmu people don't much care about pedestals and timewalk.
> Unfortuantely, those are exactly the constants that the GC monitor is in need
> of.
>
I am uncomfortable with Rich's calibration method, which I have
mentioned before. Rich's constants do have the advantage that the
worst boxes are smoothed over, but the physical assumption that the
method is based on (that rates will be equal for a given plane at the
same energy threshold), although roughly valid, is not true enough to
give the required precision for setting energy scales. The absolute
energy scale is also not well determined with this method. The
differential energy spectra obtained with Rich's constants have a
filled-in valley which is not present in either the Phrase
differential spectrum or in the spectra obtained using ERP standard or
corrected calibration constants.
I think the best thing to use is the adjusted ERP calibration
constants of my offline analysis, as described at the last 2
collaboration meetings (and detailed in my thesis). The ERP and
PHRASE energy scales can actually be fit beautifully to one another
with a linear fit (box by box); this method also eliminates the few
wackily calibrated boxes, makes ERP and PHRASE rates roughly
consistent with one another, and tags ERP hits with a reliable
absolute energy scale at the low end. Installing these constants for
the monitor has been on my to-do list forever; I'll notch it up in
priority.
Although not ideal, I don't think it's such a disaster to have rates
which vary from tank to tank, even quite a bit. What is important for
a GC search is whether the *signal* looks uniform, not whether the
background rates are perfectly uniform (and in fact the background
rate is not really physically uniform from box to box! Imposing this
artificially is not really the right thing to do.) The background energy
spectrum is *much* more rapidly varying in the turn-on regime than the
signal spectrum is (numbers off the top of my head: 10-15% or so
variation in number of signal hits from 7 to 10 MeV, factors of ~10's
in bg rate from 7 to 10 MeV). Therefore large background rate
variations from box to box don't necessarily imply large signal
position nonuniformities (and some of the bg rate variation may well
be real, and not just due to slightly different energy thresholds).
So long as the true effective energy thresholds are <~ 10-12 MeV (and
based on my offline analysis I know this is true for virtually all
boxes, even for uncorrected constants), we lose very little signal.
So long as the overall background rates are not too high (<100 mHz),
the multiplicity thresholds won't be driven up beyond the point that
we'll miss a GC in the galaxy. Right now overall background rates are
well below this, even with the hot boxes. So as things stand now, we
aren't in too horrible shape. Of course I am not suggesting we don't
work on improving the calibrations; I just think there's no need to
panic. (Note that there are hot boxes at 10 MeV in the Phrase too,
although not as bad as box 134.)
Here is what I propose: we use current calibs for now, turning off any
boxes (e.g. 134) which are really wild (easy to do in the database on
VXMACB by setting the gains to zero, and has been done in the past for
egregious boxes). When we get the new calibrations, I'll adjust them
as per my offline analysis and install them (or in fact I can do this
even before we get the new constants; this job goes on my queue with
high priority.)
Kate.