Agenda:
1. status of geometry calibration -- Reinhard Schumacher
2. region 2 wire position -- Rustam Niyazov
3. status of t0 and x(t) calibration -- Liming Qin
4. status of pdu (hit quality control) -- Richard Thompson
5. code in the dc package -- Liming Qin
6. code in the trk package -- Franz Klein
7. DC/tracking works for the next weeks/month -- Franz Klein
I am sorry that Jim Mueller couldn't contribute to the meeting
because I failed to set up the speakerphone connection
in the Conference Room at Trailer City.
Franz
item 1:
Reinhard reported on the ongoing DC alignment project at CMU
using zero field data from October '97.
The straight tracks are fitted within the standard recsis tracking
after correcting the DC geometry wrt. the survey data
(cf. CLASnote 98-001).
First track segments in Region 2 are fitted, then extrapolated into
Regions 1 & 3. The track residua (projected tracks vs. hit clusters
in Regions 1, 3 resp.) are minimized by varying all translation and
rotation parameters. The procedure will be automized to find the
optimal variation (determined by chi^2).
Alexander Skabelin reported that the previous alignment studies of
Nicolai Pivnyuk were based on the midplane projection (axial Super-
layers in Regions 2 & 3) in Sector 1. The recent analysis takes the
stereo superlayers into account, too.
Discussion: A rotation of 0.5 mrad beyond the survey data adjustments
is sufficient to adjust the W-peak for elastics.
Geometry constants should be stored in separate files (not data state-
ments) to be callable by various programs.
item 2:
Rustam showed that the wire positions in Region 2 as used in geometry
definitions reflect the tamper lines - not the ideal (and actual) wire
position that is shifted in X by -1.01 mm (X in SCS=normal to wire
direction showing outside).
item 3:
Liming reported on t0 calibration that for more than 20% of the DC
channels the pulser info is missing. Moreover, one finds hidden
offsets between groups of channels that are not seen in the pulser
data (e.g. part of Region 2 south channels have an additional 8 nsec
offset; Region 3 has an add. offset of 10 nsec wrt. Regions 1 & 2).
Via SSLT Liming obtained a single layer resolution of 340 mu when
taking track segments which cross the cells within +/-75% of the cell
size. Liming proposed to parameterize the time to distance relation as a
function X(t,alpha,t0(beta),t_max,alpha_c(B),..)
where: alpha=local angle;
alpha_c=local angle correction dep. on magn. field;
t0=min.drift time corrected for time walk dep. on part. veloc.;
constraints on t_max=max. drift time.
(T_max differs per layer for ca. 20 nsec.)
Using these functions (as determined via SSLT) he couldn't get any
significant improvements in TBT
(e.g. the chi^2 for TBT is a few percent lower but has longer tails,
the residuals don't look better, etc.).
Discussion: Bogdan didn't adjust the t0 for Region 3 by 10 nsec
but got quite good chi^2 and residuals in TBT.
It's necessary to study the t0 of the pulser data for individual wires
and compare it with t0 distributions for real data.
item 4:
Richard reported on the status of 'pdu'. It has a 98% efficiency in
finding dead wires but less eff. for hot wires. The checking procedure
uses the DC_CHANNEL.map. Up to now the wire status is only used in
GSIMKO (for dead wires only!).
Junho Yun is developping PDU2 which allows to plot drift time distributions
vs. functional units (e.g. ADB, STB channels).
Discussion:
'pdu' should use a better normalisation in order to use the program for
real data. The handling with hot wires has to be improved.
item 5:
There are several routines in the DC package which are obsolete because
(i) parameters are now read from map files - not BOS files,
(ii) there are several choices for x(t) relations that are obsolete.
Liming will make a proposal how to clean up the code.
item 6:
As soon as the 'data cooking' hast started and Jim has committed his
changes to HBT & TBT parts of the tracking code will be cleaned up.
Jim's proposal for TBT (part of a mail from 4/5/98):
` Since I got the HBT efficiency up to >99%, I have been looking at
the problems in TBT. Most of them revolve around the hits chosen in
trk_dcam to be used in the TB fit. I have HBT producing resonable tracks
(with some extra "ghost" tracks), so one should use the HBT track for this
instead of a loop though all the segments in the cluster. I can explain
this more clearly later, but the upshot is that HBT only produces segments
now because other pieces of code have been looking at trk_segm, etc. I am
trying to excise those bits curently. Once that is finished, how I see
TBT proceding is
1) Use HBER or HBTR to get initial track parameters.
2) Swim the track once to get the track positions at every layer.
3) Find which wire the track passed closest to in each layer.
4) Find all possible segments within a certain road_width of those
wires.
5) Decide which of these segments best fits in each superlayer a
straight line using the timing info as well as the HB tracking
info from 3)
6) Do the TB fit.
7) Reject any remaining "ghost" tracks. '
item 7:
Further studies have to be done in all fields of DC/tracking:
- hit quality monitoring (statistics banks, event_monitor,
recsis_online)
- study: t0 for single channels
(leading,trailing edge of time distributions)
(comparison pulser/real data)
- doca vs. dtime functions (dep. on functional units, kinematics)
- automatic procedures for filling TDLY (pdu -> map), XVST
- studies on tracking (segment finding/linking,
left-right ambiguity fit, HB&TB swimming)
- utility functions for DC/tracking
(dc_ana* for DC / trk studies; library routines)
- tcl /input parameter control
items 5&6&7 weren't enough discussed in the meeting
'it was time for lunch':-)
They will be main items for one of the next meetings.
===============================================================
Franz J. Klein, Thomas Jefferson National Accelerator Facility,
12000 Jefferson Avenue MS 12H, Newport News, VA 23606
Email: fklein@jlab.org, Tel: (757)269-5879, Fax: (757)269-5800
---------------------------------------------------------------