[d2n-analysis-talk] S2m Raw L and R TDC Times -- with a (hopefully useful) primer on thinking about DAQ timing

[Brad Sawatzky]

On Fri, 19 Nov 2010, David Flay wrote:

> I've gone back to looking at the raw S2m TDCs, and upon taking a
> closer look at these on log scale, it's clear that there is a shoulder
> structure to the immediate right of each self-timing peak in the right
> TDCs (right = red, left = blue).  The same may be said of the left
> TDCs.  The main peak of the left TDCs also have a significant width to
> them.

Interesting plots.

You need to cut out the EDTM pulses -- I don't see that in your cutlist.

A tail/shoulder at longer times on the L pmts could be driven by
time-walk.  Note that _all_ structure outside the self-timing peak
around bin 1950 in the S2m R pmt spectra must either:
  1) be noise and/or secondary particles -- ie. that paddle can not have
     generated the trigger, or
  2) if you know that paddle did generate the trigger (say, because none
     of the other paddles have any hits), then the wrong hit from the
     multi-hit is TDC being stored in L.s2.rt[], or
  3) this is not really a T3 trigger and something else is generating
     the trigger timing (ie. DL.evtypebits is being constructed
     incorrectly).

The nice sharp spikes around bin 1950 in the red histos are the events
where that paddle (R pmt) carried the trigger timing.

Those are the peaks that should be aligned to a common bin in the
corrected histos.  Note that you need to align the single-PMT spike, not
the average ([R+L]/2) time for S2m.  The aligned Right-PMT_S2m time
"L.s2.rt_c[x]", where x is the paddle that generated the trigger, should
be your reference time when you align s1 later, not the average paddle
time.

Note that the width of the blue peak around bin 1800 should be driven by
the time it takes for light to propagate across the bar to the L pmt.
If the LHRS acceptance is uniformly illuminated it should have a flat
top (ie. a rectangular 'peak' with a width that is equal to
  index_refraction*bar_length/c
not a Gaussian like in your plots.  Perhaps this is an e-P elastics run?
(The width will also be smeared by the timing resolution of the TDC:
0.5ns/bin for 1877s.)

Note that the averaged paddle time "(R+L)/2" does _not_ remove the
propagation time broadening in S2m since the R PMT is the one that
generates the start time at the TDC.  In principle, you could apply a
propagation time correction to both the right and left PMT timing based
on where the tracking says the particle intersected with the bar and
remove the propagation time effect.  In practice, at least for us, it's
not worth the effort.

Because the position-correlated time offsets are not removed from
S2m_ave[] (and can not be removed without using tracking information),
you are stuck with them in S1 as well.  Because of this, I recommend you
ignore all timing information from the Left PMTs in both S1 and S2m.
You should still require that both PMTs see a hit within a reasonable
time window (this requires that both PMTs on the bar see light within a
physically relevant time window event, which will suppress noise).

I sketched out some math on the timing calculations.  If you haven't
seen this kind of thing before, it really helps when you're trying to
make sense of what you're seeing.  See the attached scan and drop me a
line if you have questions.  When I write b:{0,t}, I mean that b can
vary between 0 and t nanoseconds, depending on where the particle hits
the bar.


In other news...

That is a surprisingly large double-peak in the upper-left histos.  Not
sure what that is...  Curious.

It would be interesting to regenerate the L plots with the constraint
that that paddle was generating the trigger.  Ie. Add a cut on the
self-timing spike on the R-side PMT.  This will be simpler once those
self-timing spikes are all aligned to a common channel -- then you can
cut on L.s2.rt_c[0].

> Of further concern is that there is a fairly large spike at ~4000
> channels (for both L and R).

Assuming that no-hit events show up in bin zero (suppressed on these
plots), then I would say the blue spike at ~4100 is due to the left-side
PMT carrying the trigger timing.  That should be impossible for a good
particle unless there is a hardware cabling error.  If the spike is just
random coincidences (ie. junk), then height of the spike suggests a L+R
coincidence windows of 100 bins == 50ns.  That seems a little wide to
me, but it's hard to judge factors of two off the plot.  (Basically, you
assume a flat randoms distribution and ask yourself how many bins' worth
of background have piled up in the spike on the right.  Then convert the
#bins to ns and that is your coincidence overlap.)

If, however, there is a spike at 4100 in the red histo too (hidden
behind the blue), then I would guess that those are really the no-hit
events (ie.  no hits in that channel within the TDC window setting).
The fact that they show up in a single bin is a consequence of the
analyzer assigning some fixed value to such no-hit events (ie. 0?) and
the formula used to convert and correct the common-stop TDC data to
nanoseconds in your histogram.  This assumption would mean there should
not be any hits in bin zero though -- that needs to be checked.

-- Brad

-- 
Brad Sawatzky, PhD <brads at jlab.org>  -<>-  Jefferson Lab / Hall C / C111
Ph: 757-269-5947  -<>-  Fax: 757-269-5235  -<>- Pager: brads-page at jlab.org
The most exciting phrase to hear in science, the one that heralds new
  discoveries, is not "Eureka!" but "That's funny..."   -- Isaac Asimov
-------------- next part --------------
A non-text attachment was scrubbed...
Name: hrs_tdc_timing_worksheet.pdf
Type: application/pdf
Size: 21157 bytes
Desc: not available
Url : https://mailman.jlab.org/pipermail/d2n-analysis-talk/attachments/20101123/8fe8f5e3/attachment-0001.pdf