WAC Notes January 26 2021
|RCDB||HALOG||HAPLOG||DocDB||Online Prompt||BlueJeans Link||Runlist spreadsheet, respin 1 spreadsheet|
- 1 Organization
- 2 January 26th 2021
- 3 January 19th 2021
- 4 January 12th notes, validating respin 1
- 5 January 7th notes, updates from the respin 1
- 6 Suggested Changes after respin1
- 7 General
- 8 December 10th notes, updates from Dec 3
- 9 How to use Aggregator
- 10 Attendence
January 26th 2021
- Look at ifarm-presence of CREX respin1 (all good as of yesterday): https://docs.google.com/spreadsheets/d/1I9yOvt33Oms6oP8c30g66RReha_CA7hHChCbioz5QGc/edit#gid=173227492
- Only 5951, 6030 and 6031 are production runs with NO good events (per looking at aggregator and ratecheck PDF outputs) - Every other run has the outputs on disk as expected
- Robert has another check for good-quality of ROOT files - should be merged into Cameron's
- Robert's check of change in events (HAPLOG 4444) for important signals from prompt->CREX respin1
- This list of runs will get another update to make sure no script issues appeared
- We should look at all runs with too-many event changed to understand if a problem popped up, also should make sure we understand the source of changes (do they match our intentions?)
- And change in mean/RMSs check to understand burp cut problems will come later
- Run 6784 is the only Good Production run that failed the respin1 (for no clear reason)
- Paul is saving all the respin1 outputs to tape
- Our overall volatile usage is 39 TB, but we only have 22.5 guaranteed and 30 total high quota - we need to be careful
- Sakib, Victoria, and Cameron will find a time Paul to verify run-list change suggestions and new cuts (after triple checking the changed-number of events between respin and prompt and big shifts in central values or RMSs/burp cut issues).
January 19th 2021
- More looking at runs by WACs
- Cameron looked at Devi's BPM 4aeXY and 11 and 12 XY deviation counter outlier runs (generally I looked at any run with 500+ events of outlier datapoints)
- BMOD Slopes are coming along
- 92 segmentations of CREX data across the 224 slugs
- Some more issues have popped up while looking at the data
- Likely pedestal shift induced large BPM changes between prompt and respin - causes cut failures for hard BPM X limit cuts (this is not expected and is a problem for at least 1 run, run 5472)
- Burp cuts are too tight for some of the ~8000 runs (especially 7866 and it's 200 neighbors)
- BUT according to the error counter script it appears to be all due to BCM trip cut (even though BPM12X is a clear culprit...): https://prex.jlab.org/analysis/crex/summary_respin1/Slug195/run7865_000/run7865_000_ErrorCounters.pdf
- Weibin's slopes from http://ace.phys.virginia.edu/HAPPEX/4430 had some issue with the device list, but have been updated since. The slopes listed in the grand plots here are accurate: https://prex.jlab.org/analysis/crex/agg-respin1/ErrorFlag_slug/
- Robert has an initial estimate of the number of events and main det asym comparison between prompt and respin: http://ace.phys.virginia.edu/HAPPEX/4430
- BCM double peak and an_diff drifts showed up clearly in runs 6593-6595 and warrant closer inspection
January 12th notes, validating respin 1
- Comparing cuts/good event count for respin1 vs prompt
- Robert has a script that can do comparisons (and other variables and such)
- Running now
- Devi has made BPM cut checking outputs: http://ace.phys.virginia.edu/HAPPEX/4428
- No update on BMOD or Pedestal analysis yet
- Null asym checks have been investigated by Weibin: http://ace.phys.virginia.edu/HAPPEX/4430
- He has many plots of useful outputs besides null asym as well
- Slopes - BPM 1X slopes before AT running period (when the beam tune had a major change) are LARGE and bad probably (also BPM 1YM wire has clear electronics noise).
January 7th notes, updates from the respin 1
Analyses that need to be done with respin 1 data:
- Determine changes to the data as a result of changes in map files and cuts between prompt and respin 1
- Determine efficacy and quality of BCM, target and energy BPM cuts across all runs (and develop changes to cuts if necessary)
- Cyclewise, runwise, and run-range-wise BMOD data quality determinations need to be solidified (AT data from respin 1 can be a ifarm script testing ground)
- BMOD slopes calculation and map files, and Lagrange+eigenvector BPM analysis set up
- Second pass pedestal studies, including looking at RF injected runs and beam off pedestals across all of CREX for all monitors and detectors
- Look at charge-normalized and regression corrected detectors vs. another BCM (not bcm_target) to check for non-zero pedestals or correlations.
- Should be consistent at < 1 or 2 % level.
- Beam off pedestals should also not be deviated from.
- Reclassify and update all NeedCut, Suspicious, and non-production (but "good") runs based on respin 1 data quality and known needed changes
- Develop tools with respin 1 to evaluate beam correction, Aq, and pedestal systematics across the run
- Generate more useful and specific slug plots, etc.
- Pitt and Wien level plots, systematics, etc.
- Carefully check on various time scale pull plots (multiplet wise, miniruns, slugs, pitts, wiens, etc.)
- Null asymmetry check
- Sub-block analysis consistency check
- SAMs analysis (especially during AT running)
- Comparison of number of events/patterns per slug between Respin 1 and Prompt
- Also check on the size of regressed error bar
- Will be done by Robert
- Devi can look at BPM and BCM fluctuations away from comfort ranges
- Each WAC should scan their weeks' data to see if anything looks bad
- Runlist verification etc. will be done by WACs
- BMOD analysis is being done by Cameron + bmod subgroup
- Cameron and Weibin will look over pedestal checking, etc.
- Weibin and Sakib can help will Null asymmetry analysis
- Looking at cuts (especially burp cuts) should happen after Devi looks at BPM ranges, etc.
Suggested Changes after respin1
- Cuts size, extent, holdoffs?
- Regression BPM selection?
- Manual cuts for target and 12X BPMs need another look?
- Pedestals need a more detailed look
- Special runs with known tune change that get special cuts to segment miniruns properly
Tracking events and calibrations:
- Global run list - https://docs.google.com/spreadsheets/d/1l5qA0u9KhLIFA-MbcEpI3ezUD9saq6g1kzB_eTJD_po/edit#gid=192113375
- Special runs and events list - https://docs.google.com/spreadsheets/d/1mt39fCuIW3Pr8TO7fzhtFjmMFt4Op30jMm5wMQke6Xo/edit#gid=0
- Calibration runs analysis list - https://docs.google.com/spreadsheets/d/1mt39fCuIW3Pr8TO7fzhtFjmMFt4Op30jMm5wMQke6Xo/edit#gid=598758666
- CREX PQB sheet - https://docs.google.com/spreadsheets/d/1IXYl7AnaxCbTfQwMgTbbUlGdmMqnLQiTY1WADiEt-yE/edit#gid=179750759
- Prompt - Full Experiment Aggregator chained file: ~/PREX/prompt/aggRootfiles/slugRootfiles/grandRootfile/CREX-full-miniruns.root
- Prompt - PREX version: ~/PREX/prompt/aggRootfiles/PREXII_slugRootfiles/grandRootfile/PREX-full-miniruns.root (maybe 20% of the data got messed up somehow :/ )
- Kent's collab meeting list of tasks
Respin 1 outputs
- Respin1 slug plots: https://prex.jlab.org/analysis/crex/agg-respin1/ErrorFlag_slug/
- Aggregator rootfiles: /lustre/expphy/volatile/halla/parity/crex-respin1/aggRootfiles/
- Full info on using the aggregator: http://ace.phys.virginia.edu/HAPPEX/4429
- Full info on respin1 output folders, etc.: http://ace.phys.virginia.edu/HAPPEX/4427
- Any run which had special cuts applied to segment BPM 12X regions will need to be checked on after respin 1 to retune those cuts
- BCM analog difference value doesn't compute correctly in the aggregator ... is .hw_sum an invalid leaf? Is there an assumed /ppm in the grand plots script for yield_.._mean?
- All BCM Aqs have a much larger difference between wien states and away from 0.0 for respin1 vs. prompt. Is the pedestal wildly off during prompt or during respin1? More pedestal studies should be done to decide what is going on.
- The main detector's regressed RMS has a drop from prompt->respin, especially around the slug 140 region.
- The Chi2 values are all pretty small, and got smaller with respin1 vs prompt.
- Detector asymmetry "answer" did not change value or error bar very much between prompt and respin1.
Runs for checking later
slug to check =
- slug 105: run 5576
- slug 123: single arm (right) run--6106
- slug 124: 4 right arm run: 6107-6110
Missing runs in respin1:
- Ifarm failed
- Have been marked as junk more recently
- 8428,221,ALL_PREX,NoFlag,IN,FLIP-RIGHT,0. Beam OFF. Accel DOWN. JUNK
- 8429,221,ALL_PREX,NoFlag,IN,FLIP-RIGHT,0. Beam OFF. Accel DOWN. JUNK
- 8431,221,ALL_PREX,NoFlag,IN,FLIP-RIGHT,0. Beam OFF. Accel DOWN. JUNK
- 8483,222,ALL_PREX,NoFlag,OUT,FLIP-RIGHT,0. Beam tripped OFF. Accel DOWN. JUNK
- 8559,223,ALL_PREX,NoFlag,OUT,FLIP-RIGHT,0. BeamCurrent = 2uA. JUNK
Prompt labeled as NeedCut:
- 5939 - Suggest relabel as Calibration run_type - It is a 60 uA run, at this point worrying about calibrations for this run is more expensive than losing the 4 miniruns (1.5 miniruns of 150 uA statistical equivalent) data that this run contains
- 5940 - Suggest relabel as Calibration run_type - It is a 60 uA run
- 5941 - Suggest relabel as Calibration run_type - It is a 60 uA run, 1200 patterns at 120 uA...
- 6030 - Suggest relabel as Calibration run_type - It is a 50 uA run
- 6031 - Suggest relabel as Calibration run_type - It is a 50 uA run
- 6161 - ~0.8 mm beam excursion in X direction, should we keep it?
- 6201 - Cut, suggest relabel as Good
- 6206 - Cut, suggest relabel as Good
- 6210 - bcm dd width gets larger with laser noise
- 6727 - Good Production. Add to slug146 RunList
-  - Good Production. Missing from Respin
- 6796 - Needs Cut. bpm12X excursion in minirun 6
- 7402 - Marked as needcut production, looks good. Has 1 single event with 1mm 12X excursion
- 7836 - Only keep bpm12X>-1.5
- 7842 - Given there is only a small number of events after the jump in minirun 2, may be it's ok to keep it as it is and reclassify as good.
- 7865 - Prompt output shows intermittent periods of larger natural width in bpm12X. The burp cut cleans out these regions. Is this too stringent? Should we just cut the tail with an absolute cut on 12X yield or cut the whole block like the burp cut is doing?
- 7866 - Same issue as 7865.
- 7867 - Same issue.
- 7869 - Same issue.
- 7957 - bpm excursions mostly cleaned up. Less than half a minirun worth of events ommitted from prompt. Can be reclassified as good.
- 7960 - Same as 7957.
- 7961 - More than 1 minirun worth of events ommitted. Burp cut too stringent?
- 7962 - Same as 7957.
- 7966 - Like 7842, can this be classified as good even though there is a small jump in minirun 3?
- 7967 - Burp cut didn't help at all still showing large excursions in position and energy bpms. Absolute cut on the position or energy bpm?
- 7994 - still large shift in bpm12X (~0.9 mm), split it into 2 sessions?
- 8014 - Looks good now, suggest relabel as Good
- 8019 - still some large bpm12X excursion, need further cut
- 8046 - Looks good now, suggest relabel as Good
- 8086 - Looks good now, suggest relabel as Good
- 8091 - A few events of large excursion in bpm12X, need further cut
- 8096 - same as run 8091
- 8276 8277 8278 - Good Production. Add to slug 215 RunList
- 8321_08321_1 - split file. Merge
- 8390 - Looks good, cut made before respin. Suggest relabel as Good
- 8403 - Looks good. Suggest relabel as Good
Prompt labeled as Suspicious:
- 5437 - Suggest relabel as Calibration run_type - Too short run, 2000 patterns
- 5452 - Suggest relabel as Good run_flag - Had some BCM Aq noise (off of the injector laser) - totally normal and OK.
- 5951 - Suggest relabel as Junk run_type - No good data, 1 arm running anyway
- 5972 - Suggest relabel as Good run_flag - It is 1 minirun, all BMOD anyway, looks great
- 6035 - Suggest relabel as Junk run_type - It is a polluted run - Trigger Supervisor broke and the asymmetry calculation has the wrong synchronization. NOTE this run is currently NOT included in the slug plots, but is included in the Aggregator outputs (can be removed after this decision)
- 6330 - Suggest relabel as Good run_flag - Aq feedback had an issue - 4ppm at 3 sigma level, but honestly it should be considered as GOOD
- 6331 - Suggest relabel as Good run_flag - Aq feedback had an issue - 8ppm at 8 sigma level, but honestly it should be considered as GOOD
- 6332 - Suggest relabel as Good run_flag - Aq feedback had an issue - 8ppm at 8 sigma level, but honestly it should be considered as GOOD
- 6593 - Suggest relabel as Good run_flag - offline analysis for bcm analog vs dg drifts
- 6595 - Suggest relabel as Good run_flag
- 6624 - Suggest relabel as Good run_flag
- 7023 - Suggest relabel as Good run_flag - It is most of 1 minirun and has a full bmod supercycle in it
- 7232 - Suggest relabel as Good run_flag - It has clear jumps, but not too big to worry about
- 7291 - Suggest relabel as Good run_flag - It has really trippy beam and significant noise in the BPMs (as FFB turns on and off, it seems, or some other related issue). Low current data gets cut (just makes plots look weird).
- 7294 - Suggest relabel as Good run_flag - 0.6mm 12X jump during BMOD... not too bad I think
- 7296 - Suggest relabel as Good run_flag - All good here
- 7297 - Suggest relabel as Good run_flag - 0.6mm 12X jump during BMOD... not too bad I think
- 7304 - Suggest relabel as Bad run_flag - BPM 12X is the only one to see a huge shift (caused by 12YM drooping alone). This is the onset of bad BPM data for a few hours. Likely an electronics issue. Also this run is very short, only 5000 good multiplets
- 7305 - Suggest relabel as Bad run_flag - BPM 12X is the only one to see a huge shift (caused by 12YM drooping alone). This is bad BPM data for a few hours.
- 7306 has already been labelled as Good anyway - the signal on 12YM looks fine, and the noise in 12X is matched by 11X.
- 7307 - Suggest relabel as Good run_flag - BPM 12X is back under control,
- 7347 - Suggest add a cut then relabel as Good run_flag - Cut out BPM12X < -1.6mm to remove a big 12X spike during BMOD
- 7626 - Suggest relabel as Calibration run_type - It is a large charge asymmetry run, no Aq feedback - 18ppm (5 sigma != 0) - but, otherwise this run is fine... 2 miniruns
- 8082 - Aq feedback problem at the beginning, suggest cut them out to keep about half minirun events
Special run cases
- 6852 - Marked as good production, looks good. Still has large 12X noise (brief 1mm excursion)
- Has been cut to keep only evnts with Ca-48 target in position - cut out event=20000:
- 7402 - Marked as needcut production, looks good. Has 1 single event with 1mm 12X excursion
- Has a cut for IHWP change after starting the run - cut out 0:2000
- 7842 - 460 Patterns in second range of the BPM 12X pass cuts with BurstCounter==1 (should be 0). The starting CodaEventNumber is 272202, and final is 274038 (so if we respin this with exactly the same BPM and BCM cuts and pedestals, etc. then the final CodaEventNumber in the cut to accomplish the minirun-segmentation at 12X yield should be 274038)
- Split two 12X domains, needs revisiting
- 7860 - Event cut fails to kill all bad beam motion events (Entry$ != CodaEventNumber???)... probably should have the cut extended to event 375k to accomodate?
- 8321 - Respin 1 appears to have analyzed this run as 2 segments (although I thought we had set the chainfiles=1 setting appropriately to avoid this, or did that only avoid split ROOT outputs?)
- Note: The second split has 1 full minirun and it's outputs look fine
- Note: As far as I can tell, the LRBoutputs slopes are treated separately with .000 and .001 segments (so hopefully the slopes are adequate)
other while the pass 0, 1, and 2 were being executed (BUT, if the two segments were analyzed in sequence rather than parallel everything should be fine)
- Note: The aggregator hasn't been set up to analyze segment != 0 (Although it can be done by hand just fine). Meaning the minirun from segment 001 is not included in aggregator outputs at this time
- 8507 - Looks good now
- Has a cut to keep only events with Ca-40 target in position - cut out event=20000:, blinder.force-target-out=1
Too short runs (nentries <2000), probably should be cut out retroactively:
- 5941 - 1194 entries
- 6105 - 1696 entries
- 6607 - 978 entries
- 8295 - 905 entries
ifarm failed job runs:
- 6784 - ifarm failed
- 8035 - has output plots
- 8033 - has output plots
- 7962 - has output plots
- 8260 - has output plots
December 10th notes, updates from Dec 3
Prepared for respin successfully See also notes: WAC Notes December 10 2020
How to use Aggregator
See also last week's notes: WAC Notes January 12 2021
Cameron, Robert, Devi, Adam, Paul, Ryan, Weibin, Victoria, Sakib, Juliette