Spiros found the Millepede
program for Linear Least Squares Fits with a Large Number of Parameters.
The Millepede's idea is to define two sets of parameters:
and fit local parameters for each event with fixed global ones, substruct result from resudials and then
fit global parameters. The procedure is iterative because it supposes to refit local parameters with the corrected
global ones and tigher cuts.
- local ones which are associated with the given event.
In our case these are the position of primary vertex
and direction and momentum of primary tracks)
- global ones which are associated with detector alignment and which are common for all events.
The test case provided with the package contains alignment of 10 planes with one measurement
calibrated with straught tracks.
To check performance of MIllepede for our case
I create a plane test example. The events contain 5 tracks coming from the same vertex with pT exponentially distributed from 0.2 GeV/c
- two measurement per plane and
- 6 alignment parameters
Notation: (u,v,w) are (x,y,z) is the plane local coordinated system, (α,β,γ) are rotation along (u,v,w).
Histogram is simulated misalignment, points are reconstructed ones.
- 10K events
- 100K events
Last modified: Wed Feb 1 14:49:09 EST 2006