Log entry time 17:55:01 on April 16, 2010

Entry number 317712

I've spent a little more time on the GEM analysis to try and make sense
of the hits we see.  For the most part, the signals seem to be mostly
reasonable with the exception of a few rogue channels.  I had to make a
few changes to get the clustering working properly:

1)  The gassiplex channel to strip mapping needed to be modified for half
the planes (it still needs to be thoroughly, systematically checked).
2)  Software pedestals implemented
3)  Event-by-event pedestal noise suppression added

For those of you playing at home, the run I'm showing below is 27356,
with a few cuts on cluster size and minimum ADC value.  The software I've
used here is a modified version of Ole's TreeSearch-GEM library, which
I'll give to him for "production" on the adaq machines once I check a few
more things.  The replayed file can be found on the adaq machines as:

/adaqfs/home/adaq/happexsp/prex/prex_gem_L_spr_27356.root

The script I used to make most of these plots is here:

/adaqfs/home/adaq/riordan/prex/plotvdc3.C

Additionally, Alex and I believe that the INFN chamber below that doesn't
show a signal may have had voltage that was too low.  

The event-by-event noise is a more crude version of what Guido/Evaristo
suggested, but I think it works for now.  The noise subtracted pedestals
look pretty clean, fig 1.  The noise is roughly the same size as the
noise-subtracted pedestal, fig 2.  Additionally, the UVA and INFN planes
seem to have some correlation, fig 3.  The INFN chambers show high
inter-chamber correlation, fig 4.

I think the clustering may need a little work.  For many events we see
some nice clean signals, fig 5.  However, for many events we'll see more
complicated event structures, such as fig 6.  I think these multiple
messy peaks together are related to the same track, since they show some
correlation with the projected VDC tracks, fig 7, fig 8.  I think if we
let the tracking select out a single one of these multiple peaks the
apparent spatial resolution will improve.  (These could also be a mapping
problem, I need to check more closely).

What was very nice is the strong cluster ADC sum correlation between the
orthogonal planes in the same chamber, fig 9, 10.  The second UVA chamber
probably needs some work and HV adjustment (the hit efficiency on the y
plane is definitely much lower), but you can clearly see the other two of
the chambers giving a strong peak.  I'm hoping the stuff around the peak
is stuff that shouldn't be associated with it.  Believing that, some
estimates give the XY misassociation probability of ~0.3-0.4, which I
understand is close to what was seen before.

We do have three working chambers, so it may be worthwhile to try and
start doing tracking with them using the data we have.  I believe the
tracking efficiency at this point will be pretty low, but can be improved
with work.  One of the critical things to do right now is develop HV
optimization criteria, which will probably be dependent on the tracking
performance.

Figure 1 - Corrected pedestal

Figure 2 - Pedestal Noise

Figure 3 - UVA/INFN Noise Correlation

Figure 4 - INFN Noise Correlation

Figure 5 - Clean Event

Figure 6 - Messy Event

Figure 7 - GEM hit pos vs proj VDC pos

Figure 8 - GEM hit pos, proj VDC pos diff

Figure 9 - GEM ADC Correlations

Figure 10 - GEM ADC Differences