(present: AA, MF, IG, TI, NK, ML, TS)
Michel - Meeting-time for fall term
Michel pointed out that during the fall term, the time of the
testbeam meeting will have to change because he will be teaching
most of the time. He proposes Tuesday morning. That is really the only
time slot that is free for him (MAYBE Friday morning might work too, but that's
still unclear).
Tamara's and Tayfun's lab schedules won't be known until the last minute before term,
and Richard's schedule will have to be considered. Those who have time constraints
should consult with Richard on the best time and then let Michel know.
Michel also mentioned that he will be away until the end of the month.
Richard should chair the meeting while he is away, and replacements for Richard
would be, in the unlikely(?) event that her departure will be delayed, Naoko,
or Margret.
Status Reports
Naoko
showed plots of the response vs global time - in contrast to the plots she had
at the last meeting, she now used the spread in response as the error, rather than the
error on the mean. She mentioned that A. Maslennikov's talk during the June LArg week
mentions a correction for this, but it was performed on a cluster rather than cell by cell.
Naoko also looked at the pulse shape. For this she used the code that
Michel had written to check Pavol's weights - it plots 5 time samples per event
vs time. Naoko plotted the ratio of (digital filtering peak)/(cubic peak) vs phase
and finds variations of about 10% in this ratio over phase.
She plotted this distribution for several cells, and finds that it varies from
cells to cell. She hopes that it might be possible to perform a correction
on a cell by cell basis. If that is not possible, then a correction at the cluster
level might be the only option.
Michel explained that it would probably not be a good idea to use the cubic peak to
parametrize the correction. A better approach would probably be to do a correction
independent of the cubic fit. Then he pointed out the question as to where to
set the reference point in time where the correction has the factor 1.
He mentions that it need to be determined whether this is some kind of universal
problem or whether it is cell to cell and at this point in time he has
no clear idea as to how to perform the correction. Faced with these problems
he wonders how uniformity studies can be done, and whether the MPI group
has attempted them so far (Naoko doesn't think so). He explained why with these
problems uniformity studies can't be done and that a cell by cell correction is needed
and would be much better than a cluster correction, and he explains that Naoko thinks
that for a given cell there is one function that can be applied to get a correction.
Alan wondered whether the shapes change with energy, and Michel said that
there is no evidence that they do. Then Alan's conclusion was that if you go to high
energy, where cubic fit does a good job, then a fix for cells should be possible.
Tayfun wondered whether the noise would be affected by such a correction.
Michel said that it would most likely be a second order effect and not
make much of a difference.
Margret
pointed out that there is an update to her z-coordinate note.
Michel has put it on his web-page under the geometry information.
She then showed some plots on occupancy and energy/volume for the EMEC
of e runs at 60 GeV and 148 GeV based on a clustering algorithm as well as on a radial algorithm.
Michel pointed out that she would have gotten better results
had she used his the NoiseAlg systemAlg rather than ped_rms for the
threshold selection. The tails of the occupancy distributions are affected and
will give better results if the NoiseAlg algorithm is used. Furthermore, the
OccupancyAlg systemAlg should rather be used as it is written for this purpose.
And finally, the energy/volume plots
should rather be energy/(volume*sampl.fraction) to give a real idea of the
energy density deposited. The sampling fraction for em showers changes from
about 1.4 to 2.2 over the EMEC area covered in our testbeam
and really needs to be taken into account.
Tamara
showed plots of the occupancy of pion runs and demonstrated that a cut
based on the noise algorithm is far superior to a cut based on ped_rms.
You can see this by comparing the
channels occupancy distribution optained using the pedestal rms as noise characterization
(ps)
with the
channels occupancy distribution optained using the digital filtering noise obtained with NoiseAlg
as noise characterization
(ps).
The later peaks at around 5%, as expected since the occupancy of each channel was calculated by counting
the fraction of events where the absolute value of the signal is above 2 noise sigmas.
Michel noted that using ped_rms produces wrong absolute values of the occupancy of cells with no signal in them,
as well as broadening the occupancy distribution for these channels (which should populate a Dirac delta at
5% occupancy!). While high occupancy channel would give similar results, most
other channels would have inacurate occupancy. There are still a few channels
that have inacurate noise estimate (because of the non-gaussian nature of their
noise profile) and hence have inacurate occupancy estimate.
Naoko and Michel have
studied this. He also pointed out that Naoko found that the pedestal is Gaussian,
but she thinks that the digital filtering weights distort things and
actually turn a Gaussian into a non-Gaussian. This affects about half a dozen cells.
Tamara continued her presentation with plots of an `x-weight' vs resolution.
The x-weight will be used to combine the HEC and EMEC to obtain a total
energy, and it will be chosen such that the resolution is minimized.
She minimized the resolution for various energies at impact point I and
plotted the weights as a function of energy. Then she showed a plots of
the ratio of (reconstructed energy)/(beam energy) vs beam energy.
She is improving her analysis by fitting the energy with a Gaussian over
the peak region only (rather than a larger region).
Tayfun
showed some slides of the theory of the lateral EMEC shower profile.
He has written an algorithm that takes as an input a point in 3D in local
coordinate system and produces as an output the depth in Xo at that point as
well as the size of a Moliere radius at that point. He showed us an example at
a radius of 156.5cm: He finds that the total depth of the calorimeter is 24.6
Xo, and the size of the Moliere radius is 2.02cm. By assuming that the entire
EMEc is made out of Lead, he finds that in the centre of S1 (the first EMEC
compartment) the shower is laterally 90% contained within 0.15 Rm, which
corresponds to an actual size of 0.3 cm. In the centre of S2 the shower size
is 0.8 Rm, corresponding to 1.62 cm and in the center of S3 it is 1.7 Rm,
corresponding to 3.43 cm.
Tayfun is planning to expand the code using a line through the calorimeter as
the starting point, rather than a 3D point.
Michel pointed out that what needs to be looked at is how the Moliere radius
is affected by the varying material density around a given point. He notes
that calculations with a line as an input parameter will be valuable if one
want to predict the shower behaviour in ATLAS.
Ian - TBRootAna
Ian mentioned that there isn't an awful lot new...,
however Michel pointed out that there have been touch-ups to the Geometry
package. Also Ian has succeeded in prapering a stand-alone geometry release
which allows people to use our geometry code without using TBRootAna.
Michel also noted that Ian is now preparing his term report and that it will be
a maintainer's manual for TBRootAna. Michel will probably be the one who
will take on maintaining the package, even though he will be very busy
once fall arrives.
Michel - Final note
Michel told us that there is a testbeam workshop at CERN during the
week before the September LArg week. Presentations will include
- Status report of Athena beam test code (Naoko to write)
- HEC-EMEC analysis results. This will include Naoko's work on electron analysis including
details of the timing problem and related corrections. (Naoko to write).
If anybody else has results that are ready to be presented, they should let Naoko and Rob know.
He suggested that Tamara could prepare some occupancy based plots to show the quality of the
digital filtering noise characterization. (He will discuss with Tamara what to include.)
Margret, Tayfun and Richard could prepare slides on what they have been doing.
- TBRootAna and geometry considerations (Michel to write, with help from Ian)