January 14, 2015 Calibration
GlueX Calibration Meeting
Wednesday, December 17, 2014
11:00 am, EDT
JLab: CEBAF Center, F326
Connection Using Bluejeans
- To join via Polycom room system go to the IP Address: 188.8.131.52 (bjn.vc) and enter the meeting ID: 630804895.
- To join via a Web Browser, go to the page  https://bluejeans.com/630804895.
- To join via phone, use one of the following numbers and the Conference ID: 630804895
- US or Canada: +1 408 740 7256 or
- US or Canada: +1 888 240 2560
- Upon connection all microphones are automatically muted. To unmute your mike on a Polycom or equivalent unit, enter *4
- More information on connecting to bluejeans is available.
- Calibrations and REST data
- Calibrations and MC
- Software Review, February 10-11, 2015
- Collaboration Meeting February 19-21 at Jefferson Lab
- Calibration status/updates
- Start Counter
- Data Monitoring
Attending: Sean, Amiran (NU); Curtis, Naomi (CMU); Kei, Simon, Paul, Nathan, Will M., Eugene, Mark I., Elton, Lubomir, Mark D., Adesh, Mike S. (Jlab); Matt S. (IU); Justin, Tegan, Noemi, Andrei (Regina); Volker (FSU)
- In the Offline Software meeting last week, it was decided to create data files in REST format because of the interest in analyzing reconstructed tracks and showers to search for bumps and other higher-level analyses. The first set is planned to be generated started on Jan. 23, and will fold in all improvements in reconstruction and calibration at that time.
- From personal experience, Sean reminded everyone that calibration tables used in reconstruction higher than hit-level will also be used when analyzing simulated data, and appropriate values should be filled into the tables under the "mc" variation.
- We will be subjected to a Software Review on February 10-11. Curtis, Mark I., and David L. have been meeting to prepare for this.
- We talked briefly about planning for next month's Collaboration Meeting. We expect that most of the calibration-related activity will be reported by the individual detector groups. If all goes well, Sean can give just a summary talk in the offline section and will coordinate with Mark I.
- Status of gain calibrations is essentially unchanged from last year. The pi0 calibration code is tested, so the challenge is to select a clean sample of pi0's of sufficient size to use. The first sample of pi0's that Matt tried to use had an S/B of ~1/1, which did not provide good results. Using various quality cuts (e.g. timing and ratio between peak integral and peak height), he can get a very clean sample but that has very limited statistics. He's currently trying to find an acceptable middle ground.
- Eugene asked about the overall scale factor calibration, and noted that there are a lot of electrons in the central part of the detector which could be used to calibration that portion at least.
- The major challenge is that there is a lot of noise in the data which didn't show up in earlier testing (Sergey P. estimated that ~80% of clusters are due to noise), and this problem needs to be carefully studied so that it can be fixed.
- Elton posted a note summarizing the low-level BCAL constants. Sean is helping to create the tables and integrate them into the reconstruction code.
- Mark D. has added the attenutation lengths derived from the data to the CCDB. The lengths are longer than expected (by ~10-20%) due to reflections, which lower the derived energies. Adding these parameters did not improve the resolution, however.
- Will M. has just run over all the data, and is looking at the total number of pi0's and using the pi0's for gain calibration.
- There was a detailed discussion with Andrei about the particulars of the attenuation lengths, since some of the effects may depend on issues related to the light guides or some other details of the detector, which may change over time. These systematics should be studied in detail, but the the particulars will be discussed more in Friday's Calorimetry meeting.
- Curtis reported that the gains seen in the data were ~3 times larger than expected. Some of this can be explained by a difference in the gas mixture. An analysis of the drift times showed that the gas mixture was closer to 54/46 Ar/CO2 than the 50/50 mix that was expected.
- A discussion between Curtis, Beni, Simon, and Naomi yielded:
- The overall drift time spectrum of all hits showed some problems.
- Some straws have a drift time spectrum with a clear end point, whereas some gradually fall to zero.
- Naomi showed some spectra she analyzed using her timing algorithm. The
- The constants need to be updated to take these effects into account.
- The gas mixture can be obtained fairly quickly, with a few thousand hits, so once we understand better what is going on, this procedure can be automated.
- The FDC also has a different gas mixture than expected. It looks like there might also be a small percentage of alcohol in the mix as well. Lubomir has read some papers in which a 1% alcohol contribution yields a 10% faster drift time.
- Lubomir is working on the FDC alignment. He was first trying to do a global fit, which did not converge well. He is now working on aligning it package-by-package.
- Vlad is almost done with the first run of gain calibrations, which he will pass on to Simon.
- Simon has been working on calibrating the t0's for the FDC and CDC and almost has a first version.
- Mark I. reported that Sasha O. is working on timewalk corrections. He is using some global calibration scheme for the whole detector, but the details of it are lacking, other than that it has 6 parameters.
- Simon has implemented a crude timewalk correction, which could also serve as a starting place for channel-by-channel corrections.
- Simon has done a very basic timewalk correction, similar to what he did for the TOF.
- No other progress to report here. Sean had talked to Eric about moving some of the calibrations done at FIU into the CCDB, but a new student is being given the task. Sean is following up on this.
- Simon has put in an overall time offset.
- Nathan has been doing a more comprehensive analyses of the TAGH data, and is in the process of extracting per-channel time offsets, efficiencies, etc.
- Simon has put in an overall time offset.
- The only other news is that Alex B. has said that the data is showing the painted fibers to be problematic.
- Extracting useful information from the TDCs was a major topic of the last run. The eventual solution was:
- Peg the F1TDCs to Beni's clock signal.
- The CAEN TDCs were pegged to a trigger signal up to a 6-fold ambiguity. Kei developed a procedure to unfold this ambiguity, and it was automized, requiring one extra constant which varied per run.
- The needed constants and values have been added to the CCDB and the data stream.
- Simon has looked at the TDC data from all of the detectors, except the BCAL, and they look free of obvious problems.
- The planned upgrade to the TS firmware should yield a better solution for the upcoming run (for the CAEN TDCs, at least).
Several topics of general interest were covered in the previous discussion.
- Having a list officially approved runs for analysis would be very useful. Justin is currently using a python script to determine a list of runs from the current run info DB. Justin agreed to post his script and the list of runs. It would be good to have a human make the final approval.
- It would be useful to compare the results of an analysis between EVIO and REST formats to help identify areas that need improvement.
- Elton asked if we should load default values into CCDB tables after they are created (e.g. 0) if their values haven't been extracted from data yet, or just leave them empty. Generally, while it depends on the exact situation, Sean leads towards loading some default value as soon as the table is created, with the caveat that one might want to wait until closer to when you might want to actually use the table since the format of CCDB tables (by design) cannot be modified once they are created.
- Simon is working on updating the geometry, primarily changing beamline element (e.g. beam profiler moved to downstream of main spectrometer), and introducing the (non-trivial) tilt of the start counter as installed.
- Sean is testing a fix to the long-standing problem of truncated processing in mcsmear. Details will be discussed in next week's Offline Software meeting.
- Simon reported that he did some photon beam simulations and found some missing hits in the CDC. He is looking into this.
- Sean needs to revisit his calculations for detector rates, and determine what else is required.
- Accurate rate calculations will also require an update of the trigger requirements.
- Kei reported on the latest round of data monitoring. He ran over all available files. Processing went relatively slow, also testing the new batch farm nodes. About half of the files are error free, the rest have various EVIO parsing errors. He and David L. are working to resolve these.
- 2 track EVIO skims and REST files were also created.
- Will move to biweekly running, no more EVIO skims will be created.
- Inputs to EventStore (event database) are also being generated. General physics skims are being generated, but more focused skims for calibration can be added if the selection requirements are sent to Paul. EventStore access is planned to be rolled out over the next month or so, depending on how testing progresses.
Curtis tossed out the idea of reprocessing the EVIO data, to reduce the data footprint and to speed up analysis. This could involve several improvements, including
- Converting from fADC mode 8 to mode 7 format
- Throwing out unneeded headers
- Using different algorithms on the saved waveforms (e.g. Naomi's upsampling fADC125 algorithm.
- Post python script and list of good runs in a publically accessible location (Justin)
- Compare results of an analysis between EVIO and the REST test run
- Follow up on start counter constants. (Sean)
- MC items:
- Fix bugs (Sean, Simon)
- Update geometry (Simon)
- Generate new simulations (Mark I.)
- Revisit hit rate calculations (Sean)
- Revisit trigger simulation