|
Last updated by Thomas Naumann
Contents:
1. Hit finding
2. 2D pattern recognition
3. 3D reconstruction
4. Monitoring
5.
Alignment
The BST and FST reconstruction software
is coded in the
H1REC CMZ file in the patches CT_BSTREC
and CT_FSTREC.
It consists of four steps:
The raw
data from the SIFE banks are clustered to hits packed
in the BRSE
and BRUE banks
for the BST r and u detectors and
in the FRSE
banks for the FST detectors. The clustering reduces
the number
of hits per event from more than 100 to about 20
for the r detectors.
The signal-to-noise
ratio is about 13 for the r
detectors
and 25 for the u
detectors.
2. 2 D p a t t e r n r e c o g n i t i o n
The local mapping functions BSTMAPR,
BSTMAPU
and FSTMAP transform
the packed digital hits into 2D space
points in the local BST and FST
coordinate systems. They only use the
coordinates measured by a
given detector type (r,u,v) and allow
track finding in the
local r-z, u-z and v-z coordinate systems.
The pattern recognition starts from the triplet sagitta defined as
sag = (r1+r3) / 2 - r2
for hits with coordinates ri in 3 subsequent
detector planes i.
It is zero for straight and a constant
for circular tracks.
Based on the sagitta 2D track candidates
in r-z, u-z or v-z
are combined and fitted as straight lines.
The widths of the unfiltered combinatorical
sagitta
distributions
are 18 um for the r
detectors and 26 um for the u
detectors.
The tails of the sagitta distributions
are due to nonzero curvatures.
Before pattern recognition the distributions
of hits over the u
and r
detectors contain noisy channels or regions.
After pattern recognition
they reflect the signal distribution which
is a product of
physics and detector geometry.
The detector resolutions
are derived from the residual distributions of the
hit coordinates w.r.t. the straight line
fit. They are a convolution of the
intrinsic resolution of the Si detectors
and alignment. For the r detectors
a resolution of 18 um is reached:
This resolution is a mixture of the resolutions of single
and double hits.
An upper limit of 13 um for the intrinsic
detector resolution comes from
the resolution of the best aligned sectors.
3. 3 D r e c o n s t r u c t i o n
First, 2D track candidates are combined:
r-z and u-z for the BST and
u-z and v-z for the FST. This means combining
r-z and u-z hits for the
BST and u-z and v-z hits for the FST to
3D xyz points.
Now, the global mapping functions BSTALI
and FSTALI apply
the
global alignment transformations to the
3D space hits.
They can be written out as standard hit
banks BRXX and FRXX and e.g.
be visualized
in the H1 event display at debug level 1.
Then, circles are fitted in x-y. Using
the curvature, straight lines
are fitted in s-z with s being the arc
length of the helix.
The event vertex can be included to the
both fits. The default is to
include the CSKV vertex to the x-y circle
fit but to avoid this in s-z.
The resulting 5 parameters of the helix,
the covariance matrix and
the chi2 and ndf's of the x-y and s-z
fits are stored in the
BJKT
and FJKT track
banks for BST and FST. Both banks use the H1
coordinate system. Transformations to
vertex coordinates are left to the user.
BST2 in general only measures r-z. To get
the hit and track
parameters as precise as possible in alignment
and analysis
external phi information is taken. The
mapping function
BSTMAPF
performs both local and global transformations and uses
phi information.
The ideal status is documented in the Monte
Carlo reference plots.
During online reconstruction the following
quantities are monitored:
4.1 Hit related detector properties
Signal/noise, occupancies (raw+fitted), efficiencies (internal+external)
4.2 Track related alignment and reconstruction properties