20200213-Analyzer-Mtg

Back to Main Page >> Analyzer_Meeting

previous meeting << >> following meeting

Logistic information

Meeting room at JLab: Counting House Conference Room

 BlueJeans calling instructions:
 Toll-Free Number (U.S.&  Canada):  888-240-2560
 International toll number:         408-740-7256
 Bluejeans CODE:                  385 600 867
 Bluejeans link: https://bluejeans.com/385600867

Agenda

New issues

1. LRB regression failures due to BPM correlation matrix close to being singular.
   • The "prex_corrolator_all" has BPM correlations calculated using 1H4A, 1H4E, 1H1, 1C11, 1C12, 1C16 for all CREX prompted runs. The matrices are stored in files like "LRBoutput/blueR6396.000all.slope.root" in a TMatrix. We could get an aggregator-type script to look at how the correlations changed over the course of the CREX running.
   • Adding 3P03 to the massive correlator? Or to a new correlator?
2. Cleanup of the web directory for plots
   • We will get rid of the "summary/old/*" subdirectories from the PREX prompt folder, but keep all of the most recent PREX prompt results.
   • We should also be able to get rid of the files like "run####_###_all.pdf" that are in the individual run subdirectories under the prex2/summary folder, as those PDF files are the concatenation of the other files in the directory.

Analyzer code updates

1. Issue 274: Burst summary mis-counts patterns in mulc_dit
   • This is actually due to not all regression/dithering IV channels being included in the global error definition. We should review error definitions to ensure that we include the regression/dithering channels (since the choice of channels may change through the run)
   • Weibin has a PR #277. This was not merged into the respin branch yet, and seems to be all for runs before run 4000, which is where the respin has reached. Propose finishing this respin, then doing the merge to pick this up for respin2.
2. Issue 269: diff* variables are missing in burst and muls trees
   • This is a labeling issue; Paul has a fix in testing
   • Paul will try to get back to this issue to see where he was in the testing
3. Pushing the last minirun into the previous if it is "too short"
   • Current plan: do one run-through without making trees, to find number of events in last burst. If "too small" then set a "max burst counter" and do the regular "pass1" analysis and continue the postpan and pass2 analysis.
     • This was merged into operations with PR #273, and the prompt analyzer has been modifed to use it.
   • Future plan: do a pass1 without making the trees, and only create the trees on second pass. Combine the slopes from the short minirun with the previous, and we need another label to tell the system when to do the burst advancement on the second pass. Postpan scripts would just run on pass2 outputs.
     • Cameron is working on a feature branch to do this combination diractly, but it is still in progress

Prompt analysis toolchain

1. Respin1
   • Status of respin?
     • All of the regular respin jobs have gone through, but Robert is running a few more runs that had been marked as bad, on Tao's request.
     • Robert was having trouble automating the ROOT file staging to tape; Paul will point him at some example scripts from Qweak.
     • Once Robert has finished his checks, he'll put a message out with the directory on volatile, and his example submission script for batch jobs.

New modification ideas

1. New SAM combinations are being considered: certain combinations that either cancel position noise to show charge sensitivity or cancel charge noise to show position sensitivity. These would probably get put in for new runs, and then also be picked up by the CREX respin1 whenever that happens
   • Cameron reports an issue in automating the aggregation of values from corrected trees: elements that are corrected based on a "mul" element as the source do not keep "asym" in the name of the corrected values, but elements corrected from a "mulc" element still do have the "asym" in their name. This makes it slighlty harder to find the element in the trees in an automatic way.

Other issues

1. Future reorganization of the web directories: See 20200116-Analyzer-Mtg January 16 agenda for discussion
2. If we end up deciding on a different set of regression/dithering monitors, we may want to start the CREX respin1 in the near future to pick up those changes up until now

Present

Phone: Paul, Cameron, Dustin, R. Radloff, Victoria, Weibin, Devi