The latest news from the ``calibration front'' has to do with the
production of the .TRK files that are now being produced at the
external labs (ALPHA cluster), in parallel with the production on
VXMACB.
As a quick reminder, these files, named xxxxx.TRK (where xxxxx
is the run number) have been traditionally produced at the end of
run by the USER3 procedure, submitted as batch jobs on VXMACB.
These .TRK files contain events for which only single fast tracks
are reconstructed, together with an ERP-mu trigger for at least 1
scintillator counter. They are the input to the ERP-mu calibration
software, run on a weekly basis and providing the corresponding
CALMOD constants.
There are a number of reasons that favor the production of these
files at the external labs instead of VXMACB:
(1) Full MACRO (raw) data are copied automatically to the ALPHA
cluster shortly after the end of run for the rare-dst production -
thus the file is available with no additional overhead.
(2) ALPHA cluster is much more powerful and less congested, compared
to VXMACB.
(3) Disk space on VXMACB::$1$DIA2 (where these .TRK files are saved)
has occasionally created problems - I remind you that this same
disk is used for other automated jobs, like LGB production.
On the contrary, disk space on the external labs is at this point
abundant.
(4) There used to be a periodic job that copies the .TRK files from
VXMACB -> ALPHA cluster, since the calibration software runs on
the ALPHAs. It can now be disabled.
(5) DREAM version, calibration databases and streamer tube geometries
are updated on the ALPHA cluster, while they aren't on VXMACB.
(6) If decided to re-process our data and come up with new calibration
constants, it is necessary to have this code running on the ALPHAs.
Comparing the .TRK files produced on the ALPHA cluster to the
ones produced on VXMACB, we found a number of differences that
can be classified in the following categories:
(a) Hits involving the 8th layer attico vertical (E08/W08) counters.
For these counters, the scintillator geometry DBs are different
(the 10cm correction on their x-position was never put on the
VXMACB geometry database). Introduces difference of 10cm on
the track position in these tanks.
(b) Hits involving the North/South face. Due to the different DREAM
version (1.20 on VXMACB, 1.32 on ALPHAs), there is a different
tracking algorithm (frontal tracking) for these hits. Introduces
difference of 1-6cm in track positions for tanks involved and
corresponding pathlengths.
(c) Some hits involving the North/South face are rejected by the
ALPHA version of the software (frontal tracking finds 2 tracks,
or 1 "curved" track while older tracking finds "bad" single track)
and other events are rejected by the VXMACB version (older tracking
sees 3 streamer tube hits, while frontal tracking sees 4 hits and
creates a fast track).
(d) Practically all events have differences at the level of mm (or less)
due to the different streamer-tube geometries used on VXMACB
(Dec. '94) and the ALPHAs (Jul. '95).
It goes without saying that the ALPHA version should be used for
calibration purposes, since this is the same version used to produce
the .DST files that are used for (practically) all neutrino analyses.
As such, we expect minimal (if any) improvements on up-mu and related
analyses, starting with CALMOD reference run # 17888 (Jul. 7, 1999).
Please give us feedback on the magnitude of these differences.
A presto,
Ioannis Katsavounidis
p.s.: Many thanks to Lynn Miller/Stu Muffson/Paolo Bernardini for
their help in putting together this new version of .TRK
production code and understanding the differences between the
new and old version.