|
This page provides some recipes useful to ease the DIRC data quality checks. Feel free to add new stuff inside!
|
How to start well your DQ expert period |
|
Disclaimer: This short section aims at
summarizing the main things you should do to get started in your new and
wonderful position of DIRC Data Quality expert. None of the technical steps
is detailled here: you'll find more information in other sections of the
documentation.
So: follow the steps below to make sure not to forget something important
but please read the detailled guidances...
- Subscribe to the important hypernews:
- Become familiar with the
RqmRootTreeTools
package:
- Install it in a recent test release (the ROOT version should be the same
or newer than the one which is used to write the OPR ROOT-tuples).
- Make sure it is properly compiled. You should have the following
executables available via the command line:
RqmChecker,
RqmFindBad,
RqmMerge,
RqmTestRootFile,
RqmDoCharts,
RqmMakeAl
and RqmReader.
- Learn how to use it to check the QA of a list of runs, either in
standalone (see the
README file of the
package) or by using the DIRC specific framework described elsewhere.
-
A few weeks before the official start of your duty, try to check the runs
parallel to the current expert.
- Learn how to get the lists of runs available for the weekly checks -- as
of now (2006/01), they are linked in this
page.
- Make sure you understand the differences between the three main categories
of runs: new runs, reprocessed runs and reference runs.
- New runs are... new! That's the first time they reach the DQG.
- Reprocessed runs are runs which have already went through the DQG and
for which the raw data have been processed another times -- reasons for this
are manifold: failure of the previous processing for one or more subsystems,
new DB constants, new code fixing/improving some parts of the reconstruction
etc.
- References runs correspond to the last 'good' versions of the runs which
have been reprocessed.
- Get used to the OPR DIRC plots and to the DIRC stripcharts.
- Once you're on duty do not hesitate to ask questions about a particular
runs. You may either use the DIRC hypernews or contact directly some of the
previous DIRC DQ experts: Nicolas Arnaud and Georges Vasseur.
|
Where can one find the lists of runs to be checked? |
|
They are normally available in a webpage -- currently
here
-- which is updated every week. Especially if some reprocessing is ongoing it
may be quite dense.
Let's take as example the Run 5b checks for the 2006/01/26 DQG meeting. The
html code shown below between the horizontal lines come directly from the
webpage linked above. Although the webpage layout is somewhat RQM-dependent
it is a good example of the information which are made available each time a
new set of runs needs to be checked.
If the links are not broken when you read this guidance, you can click on them
to have a look at their contents. In any case, detailled information can be
found below, right after the html display.
(...)
RUN 5b DATA
Weekly DQM
LIST OF RUNS:
(...)
(...)
Complete postcript Stripcharts:
From the OPR hbook files:
(...)
(...)
Complete root Stripcharts:
From the OPR hbook files:
(...)
(...)
When a new list is published, data usually appear in three sections of the
webpage:
- list of runs
This area contains links to the text files which give the list of runs to be
checked. These lists are available for the two steps of the processing:
'Prompt Calibration' (PC) and 'Event Reconstruction' (ER). There are two
formats of run lists which are recognized by
RqmRootTreeTools:
- proclist: (files labelled 'ER/PC list')
61232 P18.6.1cV00fb
61233 P18.6.1cV00fb
61234 P18.6.1cV00fb
- filelist: (files labelled 'ER/PC hbook list')
/nfs/babar/DQG/hbook/ER7/18.6.1c/60393-P18.6.1cV00fb.merged.root.gz
/nfs/babar/DQG/hbook/ER7/18.6.1c/60395-P18.6.1cV00fb.merged.root.gz
/nfs/babar/DQG/hbook/ER7/18.6.1c/60397-P18.6.1cV00fb.merged.root.gz
Another file whose html link is usually labelled 'Detailed list with links to run-by-run ps files' provides links to the PC and ER postscript files.
run 61277 P18.6.1cV00fb [OK] has 17567 events:
PS file from PC farm, PS file from ER farm,
run 61278 P18.6.1cV00fb [FLAWED] has 103463 events:
PS file from PC farm, PS file from ER farm,
run 61286 P18.6.1cV00fb [OK] has 424383 events:
PS file from PC farm, PS file from ER farm,
The red label into brackets shows the DQM flag for the run. One can also see
the number of L3 accepted events.
In the example displayed here, there are three sets of links in this area. The
first one, without label, corresponds to new runs whose quality is checked for
the first time. The second one, 'Reprocessing' provides the list of runs which
have been reprocessed and whose quality needs to be checked again -- good raw
data badly processed are bad for physics! Finally, the last section, 'References' gives information on the previous processing of runs which have been
reprocessed. This is useful to compare the old and new processings.
- complete postscript stripcharts
RqmRootTreeTools
produces a postscript file showing the evolution of all stripcharts defined
in the package versus the run number. These files are linked to the second
area of the example html page. There is one file per step of processing and
per type of data (new runs, reprocessing and references).
- complete root stripcharts
The ROOT (not hbook despite the labels inherited from the
pre-18 series dark age...) files used to generate the stripchart postscript
files are available in the last area. There is obviously a one-to-one
correspondence between the
.ps and
.root files.
|
How to find good runs in a given range of runs? |
|
Use the database query command
BbkQAStatus. This is
the preferred (and perhaps only!) way to get updated information on a run
status: text and html files available here and there are likely to be
oldfashioned!
Example: (as of 2005/12/20)
BbkQAStatus --GoodRuns -r 50000-50010
Good runs with physics quality calibrations
Run ProcSpec
50000 P14.4.2cV00fb
50001 P14.4.2cV00fb
50002 P14.4.2cV00fb
50003 P14.4.2cV00fb
50004 P14.4.2cV00fb
50005 P14.4.2cV00fb
50006 P14.4.2cV00fb
50007 P14.4.2cV00fb
50008 P14.4.2cV00fb
50009 P14.4.2cV00fb
50010 P14.4.2cV01fb
Good runs without final physics calibrations
Run ProcSpec
In addition to the run numbers, one gets the details of the 'best'
processings currently available. The naming convention is:
P<processing release>V<Version number of the processing>fb
A large fractions of runs are reprocessed at least once, either in the same release (if the previous processing failed or if the DB conditions were
updated) or with a newer release containing improvements and/or bug fixes.
The only drawback of this command is that its ouptut is not compatible with
RqmMakeAll: one needs
to remove the initial space at the start of the lines and the extra space
between the run number and the prospec. These two operations can for instance
be done with the the following commands (which work on
Scientific Linux):
sed 's/^[ \t]*//' <dbOutput>.txt > <firstSpaceRemoved>.txt
sed 's/ */ /g' <firstSpaceRemoved>.txt > <extraSpaceRemoved>.txt
|
How to locate OPR output files (hbook/root and postscript)? |
|
By using the OPR
QA Output page!
An example of such query for run 4970 is available
here.
To search information on a run, one needs to set 'Query
Type' to 'Run range' and 'Order by' to
'Run Number' -- not the default values. The
'Run Type' options are 'PC', 'ER' or 'BOTH'; the
'Processing Type' can be 'Processed', 'Reprocessed'
or 'BOTH'.
The query output for this run can be found
here.
The OPR QA postscript files of the current processing are directly available
via hypertext links and the paths of the hbook/root gzipped OPR QA files are
also provided. To compare different processings of a given run, one needs to
click on ALL (second left column). This opens
a new webpage summarizing the different
processings of the run.
|
How to check the logbook for a given run? |
|
Through a
dedicated
web page available from the BaBar home page:
home page → 'Detector'
(right column of the table at the top of the page) → 'Electronic
Logbook' (bottom of the left menu column) → 'List Runs' (right side of
the top menubar). An output example (for run 50000) can be found
here.
- Download the latest JAS3 version from the following
webpage.
Choose your favorite platform and install the software.
- Run JAS3 and install DqmJas.
- In the 'View' menu, choose 'Plugin Manager'.
- Click on 'Available'.
- Double-click on the folder 'hep' and then choose 'babar'.
- Finally, select 'DqmJas' and click on 'Install selected plugins'.
- To install DqmJas in your BaBar SLAC account, follow the instructions
provided by the 'README' file in the DqmJas package.
|
What are the quality flags for a run? |
|
A run can be given three flags by the DQG expert.
- Good
The data are good for physics analysis.
- Flawed
Some important DIRC monitored quantities are borderline for this run.
This feature is not enough to declare the run bad by itself but the DIRC
flag will be compared to the other subsystem flags. If too many (there is
no precise rule for this!) experts find the run flawed, its DQG final flag
may be bad.
- Bad
The DIRC data are not good for physics analysis. This usually happens when
the DIRC has an hardware problem impacting the online data quality. A run
can also be temporarily flagged bad if the processing has failed -- for
instance because of a bad global calibration. Such problems are normally
fixed when the run gets reprocessed. Even if it is not permanent, the bad
flag is useful in this case: it prevents Babarians from using a run whose
data are not OK.
The same classification is used in IR2 by the DQM although the meanings of
the flags are slightly different.
- Good
No problem has been found in the JAS plots. This diagnostic needs to be
confirmed by the DQG -- DQMs are unfortunately not 100% reliable.
- Flawed
The DQM found something weird in the JAS plots which may or may not be
important for the data quality. A quick look at the logbook entries for
the run should allow one to understand why the DQM uses that flag.
Runs too short to be checked by the DQM (the Fast Monitoring needs 1-2
minutes to configure and to start filling the monitoring plots; then the
DQM needs a few more minutes to browse all of them) are also flagged
'flawed'.
- Bad
A problem has been discovered during the run. Setting a run bad in IR2
means that it will not get processed in PC (to avoid screwing up the
rolling calibrations) and so this flag should be used with care. That's
why DQMs use normally the 'flawed' flag unless the problem is really
obvious.
If later subsystem experts think that the data could nevertheless be
useful, the run may be processed and sent to the DQG which would have
the last word on it.
|
Which are the stripcharted quantities? |
|
Limited information on the DIRC stripcharts is given below. For more details,
one should look at the files
DircTreeMaker.cc/.hh
and
DircStripchartsCollection.cc/.hh.
For simplicity, DIRC stripcharts are sorted by alphabetical order.
- assocHitsPerSectorAsym
RMS of the 12 mean numbers of associated hits (1 per sector).
See bottom left plot on the page 'DIRC: Occupancy, associated hits' in the QA
postscript file.
- bunchT0Mean
Mean value of the bunch T0 distribution.
See top left plot on the page 'DIRC: Timing' in the QA
postscript file.
- collectionTimeRMS
RMS of the distribution of the 'collection time' (hit timing - reference
timing) for associated hits.
- deltaNbGammaBhabhaMeanTopGau and deltaNbGammaBhabhaSigmaTopGau
Mean and standard deviation of the core Gaussian computed in a g+g fit of the
distribution (number of associated photons - expected number of photons) for
the BhaBha sample of tracks.
See middle and right top plots on the page 'DIRC: BhaBha and Dimuon control
samples II' in the QA postscript file.
- deltaNbGammaMuMuMeanTopGau and deltaNbGammaMuMuSigmaTopGau
Mean and standard deviation of the core Gaussian computed in a g+g fit of the
distribution (number of associated photons - expected number of photons) for
the Dimuon sample of tracks.
See middle and right bottom plots on the page 'DIRC: BhaBha and Dimuon
control samples II' in the QA postscript file.
- deltatBhabhaMeanCentralGau and deltatBhabhaSigmaCentralGau
Mean and standard deviation of the core Gaussian computed in a g+g fit of the
distribution (hit timing - expected timing) for the hits associated to the
BhaBha sample of tracks.
See top right plot on the page 'DIRC: BhaBha and Dimuon
control samples I' in the QA postscript file.
- deltatMuMuMeanCentralGau and deltatMuMuSigmaCentralGau
Mean and standard deviation of the core Gaussian computed in a g+g fit of the
distribution (hit timing - expected timing) for the hits associated to the
Dimuon sample of tracks.
See bottom right plot on the page 'DIRC: BhaBha and Dimuon
control samples I' in the QA postscript file.
- deltatGauMean and deltatGauSigma
Mean and standard deviation of the Gaussian computed in a g fit of the
distribution (hit timing - expected timing) for all associated hits.
See middle right plot on the page 'DIRC: timing' in the QA
postscript file.
- deltatGauMeanPerSectorAsym
For each sector one fits by a gaussian the distribution of (hit timing -
expected timing) for associated hits. The strichpart is the RMS of these 12
quantities.
See bottom plot on the page 'DIRC: timing' in the QA
postscript file.
- fracEventsWithoutAssocHit
Fraction of events with 0 associated hits.
See top right plot on the page 'DIRC: Occupancy, associated hits' in the QA
postscript file.
- HerMeanNbHitsWeaverPlaneBin4
Mean number of hits per event in region C of the HER Weaver plane.
See plots on the page 'DIRC: HER Trickle Injection Monitoring'
in the QA postscript file.
- HerMeanNbHitsWeaverPlaneBin5
Mean number of hits per event in region D of the HER Weaver plane.
See plots on the page 'DIRC: HER Trickle Injection Monitoring'
in the QA postscript file.
- HerMeanNbHitsWeaverPlaneBin8
Mean number of hits per event in region G of the HER Weaver plane.
See plots on the page 'DIRC: HER Trickle Injection Monitoring'
in the QA postscript file.
- largerAlignmentResidual
Higher (in absolute value) alignment residual (in mrad) among the 144 quartz
bars.
See middle plot of the page 'DIRC: Number of Photons vs. bar;
θC residual vs. bar and dip' in the QA postscript file.
- LerMeanNbHitsWeaverPlaneBin4
Mean number of hits per event in region C of the LER Weaver plane.
See plots on the page 'DIRC: LER Trickle Injection Monitoring'
in the QA postscript file.
- LerMeanNbHitsWeaverPlaneBin5
Mean number of hits per event in region D of the LER Weaver plane.
See plots on the page 'DIRC: LER Trickle Injection Monitoring'
in the QA postscript file.
- LerMeanNbHitsWeaverPlaneBin8
Mean number of hits per event in region G of the LER Weaver plane.
See plots on the page 'DIRC: LER Trickle Injection Monitoring'
in the QA postscript file.
- meanTimeBhabhaMeanGau and meanTimeBhabhaSigmaGau
Mean and standard deviation of the Gaussian computed in a g fit of the
mean time distribution for BhaBha events.
See top left plot on the page 'DIRC: BhaBha and Dimuon
control samples I' in the QA postscript file.
- meanTimeMuMuMeanGau and meanTimeMuMuSigmaGau
Mean and standard deviation of the Gaussian computed in a g fit of the
mean time distribution for Dimuon events.
See bottom left plot on the page 'DIRC: BhaBha and Dimuon
control samples I' in the QA postscript file.
- minAssocHitsPerSector
Minimum of the 12 mean number of associated hits per event (1 per sector).
See bottom left plot on the page 'DIRC: Occupancy, associated hits' in the QA
postscript file.
- minOccupancyPerSector
Minimum of the 12 mean occupancies (1 per sector).
See top right plot on the page 'DIRC: Occupancy, associated hits' in the QA
postscript file.
- nbBhaBhas
Number of entries in the meantime histogram for the BhaBha events.
See top left plot on the page 'DIRC: BhaBha and Dimuon control
samples I' in the QA postscript file.
- nbExpMinusDet
For each of the two control samples (BhaBha and Dimuon), one computes the
following estimator:
(mean number of expected photons per track - mean number of associated
photons) /
(quadratic sum of the RMS of the two distributions)
Then the stripchart is the sum of the two estimators.
This quantity is not particularly well-defined but it did not appear quite
useful in the meantime...
See the 4 most right plots on the page 'DIRC: BhaBha and Dimuon control
samples II' in the QA postscript file.
- nbGammaAssocHitsMean
Mean number of associated photons per event for which the number of
associated photons is non zero: the first bin of the histogram is
excluded.
See top right plot on the page 'DIRC: Occupancy, associated hits' in the QA
postscript file.
- nbGammaPerTrackBhabhaMean
Mean number of associated photons per track in the BhaBha events.
See the middle top plot on the page 'DIRC: BhaBha and Dimuon control
samples II' in the QA postscript file.
- nbGammaPerTrackMuMuMean
Mean number of associated photons per track in the Dimuon events.
See the middle bottom plot on the page 'DIRC: BhaBha and Dimuon control
samples II' in the QA postscript file.
- nbGammaHitsPerEventMean
Mean number of hits per event.
- nbGammaPerTrackMean
Mean number of associated photons per track.
See bottom right plot on the page 'DIRC: Occupancy, associated hits' in
the QA postscript file.
- nbMuMus
Number of entries in the meantime histogram for the Dimuon events.
See bottom left plot on the page 'DIRC: BhaBha and Dimuon control
samples I' in the QA postscript file.
- nbTracksMean
Mean number of tracks per event.
- thetaCResBhabhaMeanGau and thetaCResBhabhaSigmaGau
Mean and standard deviation of a Gaussian fit on the the θC
residual for BhaBha events.
See top left plot on the page 'DIRC: BhaBha and Dimuon control
samples I' in the QA postscript file.
- thetaCResMuMuMeanGau and thetaCResMuMuSigmaGau
Mean and standard deviation of a Gaussian fit on the the θC
residual for Dimuon events.
See bottom left plot on the page 'DIRC: BhaBha and Dimuon control
samples I' in the QA postscript file.
- thetaCResSigmaCentralGau and thetaCResSigmaLargeGau
Core and tail standard deviations of a g+g fit on the θC
residual for all tracks.
See top left plot on the page 'DIRC: θC
resolution' in the QA postscript file.
- t0GlobalCalibrationPC
Value of the T0 (rolling) Global calibration.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector0
Value of the T0 (rolling) additional correction in DIRC sector 0.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector1
Value of the T0 (rolling) additional correction in DIRC sector 1.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector2
Value of the T0 (rolling) additional correction in DIRC sector 2.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector3
Value of the T0 (rolling) additional correction in DIRC sector 3.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector4
Value of the T0 (rolling) additional correction in DIRC sector 4.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector5
Value of the T0 (rolling) additional correction in DIRC sector 5.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector6
Value of the T0 (rolling) additional correction in DIRC sector 6.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector7
Value of the T0 (rolling) additional correction in DIRC sector 7.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector8
Value of the T0 (rolling) additional correction in DIRC sector 8.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector9
Value of the T0 (rolling) additional correction in DIRC sector 9.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector10
Value of the T0 (rolling) additional correction in DIRC sector 10.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
- t0RollingCalibrationPC_sector11
Value of the T0 (rolling) additional correction in DIRC sector 11.
See top plot on the page 'DIRC: T0 Global and Sector-by-Sector
Calibrations' in the QA postscript file.
This page is maintained by Nicolas Arnaud
Last significant update: January 25 2006
|
