Slewing:

Isabel collected slewing data in 6/97 and 8/97. Results of the first calibration run are
what is currently in the offline v21 software archive. In both cases, the data were fit to 5th
order polynomials. In the first data set, when the hi TOT is above a prepulsing threshold,
the high start is used and a constant offset is added to the high start to align it with the
predicted low start time. In the second data set, the offset between the hi start and the
predicted lo start is parameterized by a 3rd order polynomial.  This change should reult
in more accurate timing for the highest pulse height hits.

I have implemented the new slewing calibrations. For those PMTs that died between
6/97 and 8/97, The old calibrations were used. The only PMTs without slewing calibrations
are 5,38,81,96,101,105,150,194,195,207,213. All but 2 tubes in this list were dead prior to
the first calibration run, and of those all but PMT 207 were dead at turn on in February. PMTs
5 and 38 had pathological behavior and could not be calibrated. Tube 239 has no
hi TOT slewing curve, because it is located too far from a functional laser ball.

Instead of calculating a 5th order polynomial for each PMT with each event, the
slewing corrections have been made into a lookup table. The Slewing correction for
each PMT and each lo and hi TOT are calculated the first time the calibration subroutine
is called. This change increases the speed of the calibration code.

Months ago, Morgan, Jordan and others noticed that for events that blast the pond, in
many cases many or even most of the tubes failed calibration. The edge finder was
wrongly blamed for the loss of these tubes. The real reason for the lost hits is that the
light level measured by the PMT was higher than the laser calibration system was able
to reach. On average, only about 0.35 hits per event were are lost due to this limitation,
but the lost hits are our best hits. On David's advice, when a tube is hit with a hi TOT
above the calibration limit, the slewing curve is linearly extrapolated  to the higher value.
This correction is typically less than 2ns.

Another source of lost hits is events with a lo TOT lower than the minimum lo TOT
used in the calibration fit. 10 hits per event are lost in the way! I restored these hits by
applying the slewing correction for the lowest value of lo TOT available in the
calibration. Restoration of these tubes has NO effect on the angle ChiSq angle fitter,
as they are all fall below the PE cut.

An additional 2 hits per event were lost because lo TOT was too high and there was
no hi TOT. This is the same problem observed in the TOT-PE calibration. Long lo
TOTs can be caused by a late hit that keeps the pulse height above the lo threshold.
Potentially, 2 pe hit (one prompt, and one late) could mimic a 10 pe lo TOT. There is
really no good solution for these hits. I chose to recover them by slewing them as if
they were hits of the maximum calibrated value of lo TOT.

All calibrate PMTs with available TPeds are slewing corrected and have valid start
times. In total 213 tubes of the 228 total are calibrated.

TPeds:

The NYU team has generated 2 sets of TPeds based on the two different slewing
calibrations. In the first set, the pedestals were calculated at 500 hi TOT. As not all
the PMTs were slewing calibrated as high as 500, many PMTs were without pedestals.
For the second set of calibration data, the timing pedestals were calculated at 250
hi TOT. At this lower light level, all the PMTs with slewing curves except 234,231
and 44 were calibratable. These were not calibrated in the first data set, because
they were not calibrated up to 500 hi TOT, and died before the second calibration
run.

Among the living PMTs, only pathological tubes 5 and 38 are not calibrated. A
handful of tubes were replaced in 9/97.  Of these tubes, a few died soon after
installation, and several of the others ar baffled, making calibration difficult.
None of the replaced PMTs have been slewing calibrated.

TOT-PE:

In the new calibration, I have implemented my ADC calibration. My calibration method
was presented at the last collaboration meeting, but I will repeat it here.

I first found the pedestal and 1pe peaks for each channel in the data collected by Kelin
in 9/97. The Pedestal was always easy to identify and fit, but in some tubes, the 1pe
peak was not easily identified. For these tubes I looked at the laser calibration data
collected at low light levels. I defined 1 pe as the average ADC value when the tube
was hit and pedestal as the average ADC value when the tube was not hit. From this
I was able to measure the gain of the PMTs and therefore derive the position of the
1pe peak  for the calibration air shower data.

For each PMT and each value for  TOT between 1 and 1000,  I collected all the hits
between TOT-2  and TOT+2, and recorded the median ADC value.  With the
knowledge of the pedestal and 1pe peak positions, the median ADC value is
readily converted to PEs. A look up table for TOT-PE conversion is saved and
and read in by the calibration code when called for the first time.  For lo TOT,
only data for which no hi TOT exists is used.

Of the PMTs for which slewing and TPed calibrations exist, only tubes 135 and 219
were not TOT-PE calibrated. These tubes died prior to the collection of the
calibration data.

As I pointed out at the collaboration meeting, late light can corrupt the lo TOT PE
calibration. Shown below are plots of lo and hi TOT plotted vs PEs as calculated by
the ADC. In the first plot, all lo TOT values are plotted. In the second plot, lo TOT
is only plotted when a hi TOT hit is not recorded. It is clear from these plots that when
hi TOT is on and above threshold, it is a better measure of the number of PEs in the
hit. Also visible in the plots is a late tail to the lo TOT distribution for a fixed PE
level. This is caused by late hits extending the lo TOT. This effect is not visible
in the laser calibration data, as the light from the laser is all prompt. Assuming that
the light is prompt in lo TOT calibration will systematically over estimate the
number of PEs in the hit. Tony has demonstrated that the PMTs do not saturate until
50-100 PEs, so the ADC calibration is more accurate than the laser-occupancy method
for lo TOT - PE conversion.
totpe.ps

A final note on the laser-occupancy TOT-PE calibration:

A problem was found with the filter wheel. The density of the filter wheel is not linear
with the angular position. The calibration curve for the filter wheel is still not well
known. The filter wheel will be shipped back to the manufacturer for calibration, It
is not clear that the occupancy calibration data is recoverable. It is however clear
that an occupancy calibration will not be available soon.

Further modifications to the calibration code:

In addition to implementation of the calibration code, a few modifications have been
made. The PMT_INFO structure has been modified to include lo and hi TOT. These
variables are filled during calibration. Programs that use the new code must be
modified so that a pointer to a structure of type PMT_INFO is sent to CalibrateRaw.
Therefore, any declarations of CalibrateRaw in the users code must be changed from

      CAL_DATA *CalibrateRaw(RAW_EVENT *);
to
      CAL_DATA *CalibrateRaw(RAW_EVENT *, PMT_INFO *);

Failure to make this change will result in a bus error.
 

DataRead has been modified to support the reading of a compressed processed data
format. Details of the compressed format will be outlined in another memo.
The new code can be found in /disk02/Milagro/analysis/offline_v22.
 

The effect on Reconstruction:

With no changes to the sampling correction, the following results are found:

                                                          old calib                     new calib
Ave number of hit PMTs                   142.7                           142.7
Ave number of Calibrated PMTs      109.2                           128.2
Ave number of PMTs in fit                 39.5                             47.5
Fraction reconstructed                         86.4%                          84.9%
Mean deleo (deleo<10)                          3.38                              3.20

It is not clear to me why 1.5% more events fail the fit. This is however not the reason for
the improved deleo. The deleo peak is higher for the same number of events processed with
the new calibrations. Of the 14 or so hits that still fail the calibration,  about 10 are lost
because they fail the edge finder. David has proposed a new edge finder that preserves
more data.

                                                                 new calib with /DAW edge finder
Ave number of PMTs                                              142.7
Ave number of Calibrated PMTs                            132.5
Ave number of PMTs in fit                                       48.3
Fraction reconstructed                                               84.3%
Mean deleo (deleo<10)                                                3.19

I see no substantial difference between the DAW edge finder and the current version
of Scott H. as modified by Gus.

Plans:

1) continue to search for bugs.
2) redo weighting/sampling with new calibrations.
3) reconstruct Mrk501 and Moon.
4) wait arround for no good reason.
5) reconstruct milagrito data.
 
 

-Andy