Discussion
January 17 2003 [RJ] [ST] [DK] [MK]
Two possible approaches emerged from our discussion:
- obtain energy resolution using selected meson(s) (mass dependent
approach).RJ explained the benefits and weakness of favorite
mesons decays and suggested the order of steps to be performed along the
search path for energy resolution:
- select η(2γ) as a first benchmark sample.
The signal is very good, the mass resolution is better
then the one from π°(2γ), and showers tend to be
more separated then those from π°
- the π°(2γ) mass resolution is dominated
by position resolution due to the small shower separation. The
position resolution can be pulled out from MC, so the rest of it
should agree with energy resolution obtained from
eta(2γ).
This can be a good consistency check.
- the η(6γ) signal is very good but
it is not clear whether the MC signal will be sensitive to
energy resolution tweaking. In any case, the η(6γ)
can be another consistency check.
- the ω(3γ) is another good signal.
It can be required to have large shower separation, and this can
be complementary study to η(2γ) or
yet another check.
- obtain energy resolution studying timing coincidence (in general
not mass-dependent approach, but can be combined with meson selection).
The idea is to use single tagger channel and timing coincidence peak:
- the first step would be to select a sample of events (say
π°(2γ) within the timing coincidence peak of
the chosen tagger. This sample should contain trues as well as
accidentals. The sample of accidentals can be found
within timing window outside of coincidence peak. We can statistically
eliminate accidentals from the first sample by subtracting properly
weighted second sample.
- after checking whether we have meaningful mass distribution,
the second step will be to subtract total energy distributions
from these two samples. If we get symmetric distribution with
a small tail we can claim that we have extracted shower energy
resolution (for the sum of two showers).
The width of accidentals window should be large enough to obtain
good statistics of accidentals and small statistical errors
after subtraction. The procedure can be repeated for the range of
tagger channels to cover energies from 4 - 5.5 GeV.
The tools and cuts
It is important not to have too much restrictive cuts, especially
in total LGD energy if we want to obtain good statistics of accidental
distributions. On the other hand, it might be hard to select clear
ω without energy cut.
Regarding the tools, the suggestion is to make executables
that select particular sample (say ROOD_good_eta_2gamma) that
can be called to pipe results for further study, rather then storing
large amount of data onto hard drives.
Development
After our discussion during the workshop two problems were left:
- - create executable with desired cuts built in:
- The psfReduce utility is modified to include
additional data selection criteria, other then selection of isolated
showers at 5 different spots in the LGD plane.
- the pixel finder is updated but not yet tested,
while cpv times are in waiting line for testing. In addition,
the mesons tool has to be amended. Other cuts, such as
photon multiplicity, total energy, and cluster separation are functional.
- the pdfReduce.c can be compiled
with various options that do not exclude each other.
- - store the output:
- the space for storing the output from different
event selections exists on IU cluster. However, Richard found
a way to compress the data without writing a new data to the disk.
The method uses the SIEVE mechanism to record what part of data
is useful to read and what part to skip. The output file
contains the pointer to the real data in its header, and the sequence of
bytes to read and skip.
- The package is already built into the standard libraries
and ready to use (update libdata). In order to filter data
one has to provide the data_setSieve function with the
output file name, input file name and the input file descriptor.
The reading is the same as for regular itape files, the SIEVE
libs will find the real data file pointed by the SIEVE file
and do a read or seek according to the instructions in the SIEVE file.
- For the sparse data the read/seek method outperforms
streaming and multi-buffering. For dense data it is better to use
fstream or multiple buffers (controlled by STREAM_THROUGH_SIEVES switch
in dataIO.c). The BUFFER size sets the transition in the efficiency
between two methods.
- all (un)desired features of SIEVE data
are still to be discovered, everyone is welcomed to try out.
The first sample:
- the first run through standard data was performed with
the set of cuts:
- Etot_lgd > 3.5 GeV
- Cluster separation > 30.0 cm
- above cuts are used to produce a sample of well separated showers
(between 6 and 8 blocks) to test PSF fitter. As a by-product,
the subset of 2γ events contains nice η signal.
Here are some examples from the selection:
March 18 2003 Comment:
The last picture of invariant mass shows that Npix=1 cut does noting
to two 2 gamma mass except reducing statistics. At this point we decided
not to use pixel cut from the (updated) BSD pixel finder. It might
be still useful with higher multiplicities.
The cuts for benchmark samples:
- 2 γ - π°
(eps)
- The last 2 γ from r8690 run:
eps
- 2 γ - η
(eps)
- 3 γ - ω
(eps)
- The last 3 γ from r8690 run:
eps
- 3 γ - ω fit
(eps)
- 6 γ - η
(eps)
- The last 6 γ from r8690 run:
eps
The method
Extracting energy resolution from:
May 31, 2003
And now, something completely different...
