Minues from the analysis meeting, July 19 2000 -Richard Jones ----------------------------------------------------------------- I. Progress on physics analysis * Counting omegas, D. Armstrong Before he departed Scott Teige left us detailed instructions on how to carry on his procedure for counting photoproduced omegas. Dave reported that he managed to build and run the program without difficulty, and it produced spectra with the correct masses and expected widths. However when he tried to generate the same spectra with ostensibly the same gain constants and cluster finder settings, he found spectra that were quite different, with masses that were off by 5%. He will work on comparing the two codes and report at the next meeting what we are missing from Scott's magic box. * Counting neutrals, R. Jones Following up the study he showed at the last meeting on how to produce spectra reflecting a sample with any given multiplicity in the CPV, R.J. reported how several measured parameters depend on CPV multiplicity. Results were shown for the LGD total energy, the pixel multiplicity, the polar angle of the barrel track for 1-pixel events, and the neutral energy in the BGV. The strongest correlation was seen in the total LGD energy, with observable effects in the BGV neutral energy and barrel track theta. Someone suggested that he also look at the tagger time spectrum to see if there is a correlation between the coincidence probability and the CPV multiplicity. He will look at this and report on it at the next meeting. * Other observables as measures of sample purity This question was discussed and the following list was made, in order of decreasing interest (and difficulty, sorry to say). 1) total number of phi's in eta,gamma compared with background at phi 2) total number of omega's in pi0,gamma compared with background 3) total number of eta's and pi0's compared with background (also a0,f2,...) 4) CPV multiplicity, corrected for randoms 5) neutral energy in the barrel (until we can use it) Left off the list are the obvious things that one would cut on to improve the sample purity: total forward energy, some kind of recoil proton PID using the BSD,BGV pulse heights, and tagging coincidence timing. Some preliminary work has been done on all 5 observables, with write-ups on items 2 and 5. A write-up on item 4 is in progress. More work remains to be done on all of them. * Events with uncorrelated LGD/BSD - E. Walker Erin looked at the question of how often the event that produced the energy that satisfied the Esum trigger in the LGD was different from the one that produced the start (BSD.Tagger). To study this she looked at the time spectrum of all hits in the CPV. Since the CPV is not in the trigger, any excess over the flat randoms level in the inclusive CPV time spectrum must come from something associated with the trigger. From other studies we know that the majority of the events that pass the Esum condition contain at least one hit in the CPV, so the CPV time can measure the LGD time for those events. In the CPV t spectrum, there is a tail of excess events that extends beyond the BSD coincidence peak for about 125ns. Erin counted the events in the tail and found about 200K for an event sample with about 1.6M events in the coincidence peak. She concluded that 10-15% of our events are of the type where what fired the LGD is unrelated to what produced the trigger in the BSD. She then went on to estimate what the total yield of triggers would be if the LGD were self-triggering, that is, if the BSD were always on and the rates were completely controlled by the Esum. From the total rate in level-1 of about 500KHz and the fact that most of the excess events were found in the early 70ns of the ADC gate, she concluded that the total LGD self-trigger rate would be about 3 times the present rate of level-3 triggers. This surprising result generated considerable discussion. Question: Can these LGD/BSD accidentals be reduced by a recoil proton PID? * Work in progress: 1) D. Armstrong - standardizing of omega-counting code, will write up? 2) R. Jones - looking at tagger/CPV multiplicity, write-up in progress. 3) C. Steffen - looking at a recoil proton PID using BSD/BGV info. II. Progress on calibrations * R. Mammei is finishing calibration of tagger timing, has worked out a scheme with Craig to get constants into the map, working on write-up. Problems with L-R timing shown at last meeting have been solved, L-R timing now has sigma=500ps, at limit of TDC resolution. Individual tagger-CPV28 timing has sigma around 1.5ns, very good considering size of CPV28 counter. * C. Campbell is finishing calibration of BGV timing, working with Craig to get constants into the map. * D. Steiner spoke about his ongoing work to calibrate the timing of the UPV, will work with Craig to get constants into the map. * Work in progress: 1) R. Mammei - looking at anomalies in tagger timing, report at next mtg. 2) D. Steiner - UPV timing, report at next mtg. III. Software tools development * LGD bad block recovery, E. Koskinen Eric reported on his work aimed at enabling the reconstruction of showers that abutt dead blocks in the LGD. Using the knowledge of the shape of showers in the LGD, he presented an iterative algorithm that should be capable of estimating the missing energy in the cluster. He tested the algorithm by eliminating blocks known to be working, and comparing the measured energy with that predicted by his algorithm. At this stage his algorithm is working too well because it seems to predict the missing energy to several significant digits. It was pointed out that the error should be on the order of the LGD energy resolution, 10-20% depending on the energy. He is still working on the algorithm. * cluster cleanup routine, C. Steffen Craig has created a new tool that should be called by the user code after the clusters have been created. It checks the clusters for fiducial conditions and returns three health flags for the event as a whole: a) all-clusters-are-without-outer-boundary b) all-clusters-are-outside-inner-boundary c) no-clusters-are-too-close-together All of these cuts require further refinement, but this routine implements the cuts as they are presently being applied by Scott's code, Craig thinks. During discussion several refinements were suggested, including the following: i) in addition to the per-event flags a-c there should be a place where you can look up which clusters violated the rules ii) the cluster-separation cut should be energy dependent, eg. they should be allowed closer together if they are about equal in energy. iii) instead of just vetoing the whole event, try to patch up an event when two clusters are too close together, i.e. merge them. Craig is not opposed to these developments, but implementing them would require further study. He is checking in the code to cvs as it is. * Work in progress 1) Eric will continue working on his bad-block algorithm and will report his progress at the next meeting. 2) Craig will check his map-update program into cvs and make sure that any problems are fixed.