Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • master default protected
1 result
Created with Raphaël 2.2.022May151312930Apr11Mar21Feb2011831Jan11816Dec330Nov282119131297530Oct2518174230Sep282524211315Aug144230Jul292726191816103229Jun27201918121165429May2814873125Apr2419141312109765319Mar1614Fixed remaining issues with the interstitial frame DAG conceptmastermasterAdded a file that should be able to submit the interstitial frames in convenient DAG formatFixed a typo that made for an inconvenient filename referring to an observatoryMade the plotting feature more useful by labelling observatoryAdded a nice python plotter to display the ratio of before and after SNR calculated by Ian Harry from iHope CBC dataThe vetoes file for L1Should have enabled the interstitial frames jobs to take on frames of odd lengthAs far as I can tell, the loopMICH (type 0) and initialMICH (type 2) errors from the LHO and LLO production runs are now fixed. They have been correctly resolved, meaning that the output of AMPS in the rare instances where LDAS frames have an odd length is exactly aligned with LDAS. Where LDAS has a short frame, AMPS has a short frame, and vice versa: all the remaining frames around them are normal-length in AMPS, just as they are in LDAS. I tested on the one case at LHO of initialMICH and on a loopMICH LHO job, as well as on the entirely-normal first science segment at LHO. They are fifth-decimal place differences between that first science segment and the 2013/01/22-1 run, because the latter was run on Matlab2010b whereas today I used Matlab2012a, but otherwise everything matches. I think our problems are over... or perhaps they are just beginning.Made changes that should fix the remaining bugs. I check to see if there are any anomalous frames and, in the rare cases where there are, I think an adjustment to the frame head or tail tau variables should suffice.Fixed a special case where the obliteration of 32 s windows into windows of up to 512+32 s long caused the final window to extend into another frame, which constructWhitespace had not prepared for. To minimize intrusion into other parts of the code, an edit was made directly to the failure point in the HoftEditor windower function, where the frameTail is attached; the Hoft.data and Hoft.baseline structures are extended by the required amount, which is exactly the length of the frameTail. A variety of checks are made the ensure this only occurs where needed and that it survives in cases where the window needs to be renormalized due to vetoes. All checks out with one test case, and since the entire category of so-called type 1 errors in the 2013-01-22-1 LHO and 2013-02-21 LLO production runs were similar across all the variables that are invoked, I think it is safe to say this should solve the problem for all errors involving loopPRC. Note that loopMICH and initialMICH, also known as type 0 and type 2 errors, are due to issues with 64 and 96 s LDAS frames sneaking into ligo_data_find and the cluster nodes, and I cannot resolve those without extensive recoding, so I think I will ignore them and just write over those jobs with interstitialFrame.mModified postProcessing to have consistent y limits in the output graphsMade the segmentDivider less conversative, will give frames with two science segments to the earlier segment.Now should work for LLO. Fixed a bug where the middle function subdivider, was obliterating some science segments, as it is wont to do for very short ones, while still writing out a line for the followed and preceding bits, which was breaking the cell-to-matrix conversion.Fixed segmentDivider so that it counts jobs correctly, which has the effect of making it more conservative about overlapping segments. Previously it gave a 128 s frame that was in two science segments to the former of the segments. Now, it gives it to neither.Fixed an issue with the segmentDivider so it correctly filters out any segments shorter than 60 s from the start.Making changes to run on Matlab 2012aAdded the corresponding L1 segment-getterForgot xlabel on the last plot in postProcessingFurther refinements to postProcessing including a smoothing methodMade modifications to the postProcessing script to scale to a full science runOne more nomenclature change.Modified memory requirements nomenclature in DAG generator script.Adjusted because the memory image size on Condor seems inexplicably to have gone up. Production run attempt again.Fixed bug where the divided segment list included entries with non-unique end times, because those subsegments were less than 1024 s long, which was a problem for the Science Finder class. Fixed by eliminating those, which were redundant anyway. Should be ready for production run.Fixed a bug in HoftEditor that prevented windows failing the absolute comb test from ever being written if they were the first window of a segment. Now it should be ready for a trial production run.Time for a production run.Added a ratio of differences plotFixed a typoMade modifications to enable absolute sum vetos in the comb and fixed a bug in the middle of divided segmentsMore characterizationPlaying at printing values to learn more about the dataNow informatively marginalizes frequenciesMarginalizing over the histogramSuccessfully made python histogramWriting a python script to derive statistics in post-processingNow in excellent shape for showing statisticsAdded a statistical cutUpdated and made statisticsMade program capable of plotting differences over all 200 jobs, before and afterMade improvements to spec_avg so it prints out about 5/16 Hz around 850 Hz.

AEI Privacy Policy Imprint