GEn Analysis meeting on Jan 24, 2007

Main meeting upload page

Follow the link below to access the plots provided for the meeting:

http://hallaweb.jlab.org/experiment/E02-013/mtg/2007jan24

Meeting minutes or direct contributions

Seamus Riordan

Seamus presentation notes are included below (so they can be searchable),
and his presentation pages are: http://hallaweb.jlab.org/experiment/E02-013/mtg/2007jan24/riordan/riordan-20070124pres.ps

The last couple weeks I have spent my time working on getting a very
preliminary measurement for GEn out as well as improving the tracking code.

I will start with the tracking code first:

I have made several improvements to regarding the tracking code starting with the intention of cleaning up the code so that it is much more readable to someone who is starting to be come familiar with it. Primarily, I have moved much of the actual calculation of a track into the THaMWDCGroup class, which has always worked as a container for a set of hits. This class now uses the hits to do coarse tracking, fine tracking, and the LR differentation instead of the THaMWDC class asking for each hit and doing the calculation separately.

Also, since the THaMWDCGroup class now stores the track, there is no need for parallel arrays holding the track and the group separately. They are now permanently linked together which is far more natural and much easier to do memory management.

The code that has been rewritten was decidedly complete and will probably not be a focus of improvements in the future. By moving it out of THaMWDC, it does not get in the way of the algorithm that is of more interest, the construction of the groups. This algorithm has not been changed nor rewritten.

Also, a heap sort has been included when sorting tracks by chi^2. Before we had a worse possible case sort that went O(n)~n^2. Now it is O(n)~nlog(n).

This project has had a number of side benefits. Namely, we have had a number of improvements in the quality of tracks.

page 1


These are the differences between the tracks found with the old and new algorithm for identical events. As you can see, these tracks are different, but not terribly different. The largest change is seen in the calculation of y'.

page 2


The structure of the residuals has improved. Where before we would see two gaussian distributions, one much wider than the other, we now see primarily one distribution where the wider one has reduced in amplitude significantly. The width of the primary distribution has reduced from about 280um to 250um.

The events to the right I believe are due to t0 offsets that need to be recalculated. By changing the offsets the size of this can be changed.

page 3


The momentum resolution has also improved from about .86% to .8%. A new set of momentum coefficients appears to be necessary.

page 4


This is the old crosstalk results. What is plotted is the time difference between the wire used in tracking and an adjacent wire. What we see here is a bias towards using the wire later in time, which has been shown to be incorrect.

page 5


This is the crosstalk results using the new algorithm. As you can see, the bias is now in the proper direction and the number of wrong choices is much smaller than the number of correct choices above background. The code is now choosing the correct wire as it should be. Masking out the crosstalk hit should be beneficial in reducing extraneous noise.

Also, with the rewrite a significant speed increase has been achieved. Before, a typical event would take roughly 0.041s to perform coarse processing. Now, that time is about 0.023s. This translates into a 50% speed increase when the rest of the algorithm is considered, roughly going from 10Hz to 15Hz.

I believe there was now a bug in the old tracking code, most likely in the L/R diffentiation that would give a poor chi^2 result for the proper combination. The new code clearly does a better job at choosing the proper wire. While I have not performed the calculation, I expect that the beamline resolution should increase dramatically as well.

The code has been checked for memory leaks and performs as well as the old code.

Furthermore, the tracking yield is roughly the same. However, there is a discrepancy in WHICH events yield tracks. I have changed the requirements for accepting a tracking slightly, so there may be some room for improvement and we may be able increase our track yield by roughly 10%.

page 6


Results for GEn. I have done a preliminary calculation for GEn. There are several files relating to this run in the directory including the parameters for each run that were assumed in the calculation of A_phys.

Ameya had provided the density of N2 in the reference cell and Edna so I was able to calculate the N2 contamination in our quasielastic neutrons. It was determined to be about 7%.

False asymmetery calculations were done for DAQ and electronic livetime, false beam charge asymmetry, and false tracking asymmetry. These were done using a script provided by Rob on a RUN BY RUN basis. This was a significant contribution to our error. This can be improved by doing the false asymmetry calculations across all ranges.

Proton contamination was done by applying the same cuts to H2 runs and looking at the rate of neutrons in the sample. A contamination rate of 11% was used. Corrections were done for each helicity independently.

There have been no corrections for background.

Cuts of p_miss_par < 0.25GeV, W=0.94 +/- 0.25GeV, p_e > 1.0GeV, and preshower signal greater than 500 were used to select out quasielastic events.


A value of GEn(Q^2=1.67GeV^2) = XXXXX +/- XXXX was obtained.

There are a number of issues that I need to work on for the next few weeks:

  • Documentation for the tracking code needs to be updated with respect to the new revisions.
  • Understanding why certain events that had tracks from the old code and do not now may prove to be useful.
  • I will spend a couple days seeing if some minor improvements to speed can be applied in the grouping algorithm. One such suggestion from Nilanga can be studied.
  • Updated databases for this run period and the 3rd run period need to be generated.
  • Fringe field corrections need to be applied to the optics to see if there needs to be any further corrections to be made. I will also take this opportunity to revise the optics code.

Tim Ngo


http://www.jlab.org/~ngo/analysis/time_offset/time_offset/time_offset.html

Followed up on last meeting and showed the effective timing-offset quality for run 4490, which had the same mis-timed channels as the earlier elastic run he reported on the previous week.

  • He will correct for the time-of-flight to the ND assuming the particle is a proton, and add error bars to the plots.

Ameya Kolarkar


http://www.jlab.org/~kolarkar/gen_analysis/
Working on the pulsed-beam analysis and the dilution factor.

Aidan Kelleher


Further filling out the run-list and working on target documentation.

Sergey Abrahamyan


Working on the runlist (detector changes) and understanding the analysis.

Jon Miller


Working on a position-dependent veto efficiency measurement. See http:tiki-read_article.php?articleId=14

Robert Feuerbach


Still working on extracting GEn, primarily concerned with the non-proton background.

From H2 data, the protons have the distribution below. The plots on the next two figures are:
  • Top-left: Missing Pparallel momentum (Ppar - q) vs. W
  • Top-right: Missing Pparallel momentum (Ppar - q) vs. Pperp for W<1.16 GeV
  • Bottom-left: Missing Pparallel momentum (Ppar - q)
  • Bottom-right: Q2 for "quasi-elastic" cuts



While the "neutrons" have this momentum distribution:


Notice how the "neutrons" have a much wider distribution in Missing Pparallel. There appear to be at least TWO processes producing "neutrons" in the hydrogen data. The first is proton mis-ID, while I'm still trying to understand the other.



  • Overall asymmetry (corrected for livetime and Beam charge asymmetry) vs. Pperp, WITHOUT correcting for backgrounds (a dilution factor). helium production runs.

Other issues:

helicityEvents with tracks (Nt)Tracking skipped (Ns)fraction skipped (Ns/(Nt+Ns))
Plus2.284371e60.452721e616.540 +/- 0.022
Minus2.317644e60.460711e516.582 +/- 0.022


Tracking efficiency asymmetry = -0.001266 +/- 0.00094
Correcting for consistent asymmetry sign --> 0.0013 +/- 0.00094

to be subtracted from the observed asymmetry.