TPC Drift velocity studies
Introduction
The quality of the DELPHI tracking relies heavily
on the reconstruction of tracks in the TPC, since the TPC tracks are the
main seed for the track search. They contribute also substancially in the
precise determination of the track parameters, and hence they are crucial
for energy flow as well as for precise vertex and impact parameter determination
which enter in the b-tagging.
it has been noticed that during the year,
the residuals of TPC tracks extrapolated to the vertex detector (VD) have
a symetric behaviour for tracks in opposite hemisphere (backward and forward).
Two main parameters can induce such a symetric variation : the origin of
times (t0) and the drift velocity Vd.
We have hence proposed to use the track
residuals in the VD to recompute the drift velocity throughout the years.
For this an ntuple has been produced which for each track contains its
parameters and the residuals in the relevant layers of the VD, together
with the drift velocity used for its processing.
If you want to jump to the results look
at the following pages :
Method and checks
The method consists in evaluating for eack track
the correction which should be applied to the drift velocity to make its
residual 0. The drift velocity to be used for the reprocessing is then
the mean value of these corrected drift velocities.
See Nathalie's
pages for the mathematics of it. To summarize, one tries to disentangle
what comes from drift velocity (Vd) from what comes from timing adjustment
(t0) using the fact that the latter is a constant term when looking at
the residuals as a function of Z while the former varies lineraly with
Z. t0 mismatches are sector-wide while Vd effects are TPC-wide.
In order to avoid using dubious tracks in
the 40 degree region (distortions used for the former processings showed
problems in those regions), a cut is applied on theta between 45 and 135
degrees. Another cut is applied to remove tracks too close to TPC sector
boundaries as well as VD overlap regions. Tracks of momentum larger than
1 GeV are selected and we require that both RPhi and Z residuals are smaller
than 4 mm to avoid randomly associated VD hits.
The Ntuple allows to check the VD structure.
Here
are figures showing the shape of the VD. First page is an RPhi view,
second a 3D view of half a VD, then 24 plots show for each VD sector the
distance between layers as a function of Z.
t0 evaluation
The effect on Vd determination of a correct
t0 determination can be seen on the following plots which show for the
time dependence of Vd for each sector (color points) and for the whole
TPC (black points). Points with error bars across the page are averaged
over time and allow a good estimate of the dispersion of the Vd determination
per sector.
The plots below used full t0 determination.
This led however to large changes in the drift velocity, mainly do to a
global offset in t0. It turns out that the main reason why the spread of
Vd determination improves with vdt0 corrections is because it cancels out
the relative t0 differences. Hence it was decided to only use the Vdt0
sector by sector after substracting the overall Vdt0 obtained with all
sectors. This we called minit0 in the following and in other pages.
Z's from 99, processing C
| No t0 correction |
 |
|
| Old t0 corrections |
 |
|
| New t0 correction, no barrel cut |
 |
 |
| New t0 correction, WITH barrel cut |
 |
 |
Vd determination
For each track, a new drift velocity is computed
which would reduce the residual in the outer layer of the VD to zero. The
time is split in slices long enough to give a good determination of Vd.
The method is different for me and for Nath :
In my method, time is split in 15mn slices.
These slices are then merged in chronological order until at least 2000
tracks enter in the histogram. A fit of the distribution (g+p0) is made
to determine the mean value and its error. This is what enters the plot
as a function of time
For Nathalie, see her
page for more explanation. The method is essentially the same except
that histograms are built after the tracks have been sorted chronologically.
Histos contain 2000 tracks unless there is a too long interfill. A pressure
correction is also applied which is not the case in my plots.
You can find the results (fully compatible
with Nathalie's, luckily enough!) in the following pages :
