1. Software for Spectrometer T0 jumps correction
Straw WG
Software for Spectrometer T0 jumps correction
D. Madigozhin
JINR Dubna

2. Outline Jumps correction technique reminder
Minimum Correction Strategy data flow
Full Correction Strategy data flow
Implementation status
Conclusion

3. Reminder of the Spectrometer T0 jumps correction technique
For each period of complete runs a job (analyser in “histo” mode) is submitted, that:
Extracts from each burst only the Raw Digi leading time (wrt to trigger) 2D histogram. So reconstruction in fact does not matter at this stage, any reconstructed root file is good.
Makes 480 cover-related time histograms rebinned with 25-ns bins.
Accumulates these cover histograms burst by burst until the statistically very significant main peak is found (sometimes it may be reached in 1 burst, sometimes not).
Finds the peak position and its error by parabola inerpolation between 3 points around the maximum. Accumulated histogram is then reset (and also at the begin of the next run).
Writes the graph [peak time (ns) vs burst timestamp (epoch seconds)] for each cover.
Automatic jumps detection:
“Corrected sliding average” - a ring memory for the last 50 values and errors. The next value enters into the sliding average evaluation with an automatically detected jump correction.
If the discrepance between the next point of the graph and the current sliding average is larger than 70% of expected jump (D= ns), and it is larger than 5 errors of difference 5 sqrt(dt2 + dSla2), the jump is detected and written into the correction text file.
Some very rare jumps are not automatically detected at all (taken as a slow drift), and some ones are detected too late (if the first points after the jump have enlarged errors).
Visual control of automatic jumps and a special visual Jumps Editor to correct the problematic cases that happen due to a poor statistics.
Automatic correstions and Jumps Editor corrections are sorted together in order to produce the jumps data file.
T0 Jumps corrections must be implemented prior to T0 extraction and reconstruction.

4. Analyzer for NA62Analysis SpectrometerJumpsCorrection
Minimum Correction Strategy (MCS)
Na62Reco pass 1 Reconstructed file list for a complete run(s)
Analyzer for NA62Analysis
SpectrometerJumpsCorrection
Text files *.dat :
In-run jumps
List of runs with in-run jumps
In analyzer output.root, directory SpectrometerJumpsCorrection:
Bad burst graph (flag vs timestamp)
Run limits graph (run vs timestamps)
Peak time vs timestamp graph per cover
Selection of runs without in-run jumps
Na62Reco pass 2 for the runs with in-run jumps
Root files with corrected in-run jumps after Na62Reco pass 2
FCS
Further NA62Analysis and NA62Reco passes
For MCS everything is now included into na62fw

5. Analyzer for NA62Analysis SpectrometerJumpsCorrection
Na62Reco pass 1 Reconstructed file lists for 2016
Analyzer for NA62Analysis
SpectrometerJumpsCorrection
Full Correction Strategy (FCS)
Text files *.dat :
In-run jumps
List of runs with in-run jumps
In analyzer output.root, directory SpectrometerJumpsCorrection:
Bad burst graph (flag vs timestamp)
Run limits graph (run vs timestamps)
Peak time vs timestamp graph per cover
Organize a period root files by means of “runbyryn” and “uniteruns” codes
MCS
Na62Reco pass 2 for the full 2016 data set
Initial time alignment
Run<firstrun>_<lastrun>.root period files in separate directories
First pass of “solve” code in each directory
zero.dat
Full jumps data file
Visual jumps editor “correct”
Merging of automatic and visual corrections from all periods
root files with automatic jump corrections
Text files corr_<cover>.dat with a manual extra corrections
Text files with automatic jump corrections

6. For each runs period <from run>_<to run>, a second control pass of “solve” jumps correction program shows a visual result of all jumps found for this period up to now. One can visually detect all essential deviations from a constant T0 value. A resulting final control root file contains corrected graphs for each cover. Graphical pdf file is also produced.
If the current final result is still not good enough, a visual jumps editor is used for the problematic cover, and more manual jumps are inserted for this cover. Then a control check of total jumps correction effect may be done again.
When the given interval corrections looks satisfactory, we proceed to the next runs period in the another directory, where a link from the last period control root file is made to the local “Pre.root”, that is used for initialization of all the sliding average objects. So cover-related sliding average procedure is made without interruptions for the full year data.
If additional correction is done for some runs period, a complete recalculation should be done for the rest of the year data in order to have a visual representation of the complete corrections result. Sometimes it helps to detect more jumps missed earlier in statistically poor covers.

7. More difficult case with a large errors
Automatic correction
In-run jump
Automatically corrected
Initial
automatic
corrected
More difficult case with a large errors
All jumps visual final check and correction
Visual correction with the jump editor
Initial

8. Time alignment remainder
By means of special artificial D-jumps at the start of the 2016 run in the jumps database, the initial cover times are made close to each other within one SRB.
It is absorbed by the standard T0 correction (no effect for physics), but it simplifies the visual control and software tests.
Period of runs
Chamber I
Chamber II
Chamber III
Chamber IV
Red: after the time alignment
Electronics cover index

9. Implementation status
The NA62fw correction needed both for the MCS and FCS is merged (included into standard NA62 framework code) during the last week. MCS is ready to be implemented in the 2016 run-by-run reprocessing.
FCS additional stand-alone software is developed.
It is tested on the on-line processed 2016 data, the draft jumps databases is introduced into NA62fw and may be used for reconstruction tests.
But on-line processing did not include all the data, so the complete FCS correction is not available now, it will be prepared only when the complete 2016 processing will be done at least once.
A new GitLab project is created for the FCS software development and distribution (Michal suggestion).

10. NA62-Spectrometer CERN GitLab Group
Created for stand-alone Spectrometer codes
Read-open for all CERN users
If you want to start within this Group new project tell me or Michal
The Group contains currently 2 projects, one of them is a software for FCS:
NA62-Spectrometer/T0JumpsPreparation Project
corrs : example directory with the usage scripts and Makefiles
jumpcodes : Root+C++ codes for FCS (currently 11 utilities)
ProSGCorr : interface developed in order to run SpectrometerJumpsCorrection without NA62Analysis (mainly for development).
toolsroot : a simple general Root utilities for work with NA62 root files (currently only 1 utility).
FCS complete documentation still to be written

11. Conclusion
Jumps detection software needed both for Minimum Correction Strategy (MCS) and for the Full Correction Strategy (FCS) is developed and is working;
Jumps correction software is published in GitLab;
FCS detailed documentation preparation is in progress;
FCS jumps database with 2016 online-processed data is ready, but it is not a complete data set even for pass 1;
MCS has a better chances to be implemented first in the frameworks of the existing reprocessing strategy.