Data Quality issues related to Tracking and Alignment

This page is meant to collect the observed Data Quality (DQ) issues related to Tracking and Alignment, the studies done in order to understand them, and the conclusions reached.

The main tool remains the DQ ELOG: https://lblogbook.cern.ch/Data+Quality

and the DQ Web interface to check the plots: https://lbdqmweb.cern.ch/prompt_dq/

This page should just collect some helpful documentation.

Note: For any plot of the DQ you are looking at the Reference is in RED, the run you want to check is BLUE

Known issues

Precision of the Alignment

• PV Left-Right delta x

• Mean of PV delta x slightly above 10microns: https://lblogbook.cern.ch/Data+Quality/12022
• Velo alignment unstable: https://lblogbook.cern.ch/Data+Quality/11000
• Velo halves aligned better than the reference, e.g. https://lblogbook.cern.ch/Data+Quality/12821

Due to the intrinsic resolution of the method variations from zero up to 20 micron can safely be ignored. So these runs are OK for Alignment

• Variation of Fast track IP distributions

• Fast track IP X,Y vs eta: https://lblogbook.cern.ch/Data+Quality/12746
• Fast track IP X,Y vs eta: https://lblogbook.cern.ch/Data+Quality/12359

• Fast track IP X vs phi doesn't follow the reference distribution: https://lblogbook.cern.ch/Data+Quality/12326
• Fast track IP X vs phi doesn't follow the reference distribution: https://lblogbook.cern.ch/Data+Quality/12379

These distributions are expected to look compatible with the reference within the error bars, deviations in cases of low statistics are also the be expected. Few microns of deviation, e.g. IP Y vs eta in https://lblogbook.cern.ch/Data+Quality/12359 are acceptable.

IP X vs phi: the distribution might not follow the shape of the reference, due to a weak mode of the Velo alignment (refer to Silvia's talk also linked in the references section). If the IP X vs phi distribution fluctuates around zero with maximum distance about ~5microns for most of the points (considering error bars) it is OK.

• Tracker residuals

• C-Side track IT1BottomBox hit residuals mean > 0.05: https://lblogbook.cern.ch/Data+Quality/12517

These look like fluctuations close to the threshold value for being notified. This run looks OK for Tracking and Alignment.
(Further studies needed to understand if and how to update the threshold in the instructions).

VELO centering

• PV left-Left and right-Right
  

https://lblogbook.cern.ch/Data+Quality/12315

A slow variation of the mean of these distributions within the fill is normal reflecting the beam movement and if the sum of absolute values of PV mean left-Left and right-Right in x direction is about 300 microns is OK

If you are concerned about the PV reconstruction, we start loosing PV reconstruction efficiency when the radial off-centering of the Velo is > 0.07mm. This means that you have to sum the squares of the differences of PV mean left-Left and right-Right in x direction and in y direction, take the square root and see if it is smaller or larger than 0.07mm. For more details see Silvia's talk and Agnieszka's talk.

Features due to the variation of the beam conditions

• Lower Number of PVs and different track multiplicity and type distribution for the first run of each fill
  

https://lblogbook.cern.ch/Data+Quality/12907

https://lblogbook.cern.ch/Data+Quality/12908

These differences are due to the lower luminosity at the beginning of the fill (just the first run is can be affected). For example the run reported in these elog entries (180188) looks OK for Tracking and Alignment given that it is the first run of the fill. Please check the luminosity for the concerned run in the DQ monet web page clicking on Run information.

For completeness, I add a trend plot of the mean of the number of PV distribution and the mu of the run range 180128-180253 (two beginnings of fill are included). The small variation of the number of PVs correspond to the beginning of the fill, which has lower luminosity (the small delay is due to the fact that for the number of PVs information I take the end of run time). More information in the elog entry and talk.

OT t0 calibration and LHCb clock shift

• t0 distribution shifted

https://lblogbook.cern.ch/Data+Quality/12568

Please check if the run was acquired just after a hardware shift of the clock (should be noted in the elog https://lblogbook.cern.ch/Shift/) or problems with the clock of other nature (always noted here: https://lblogbook.cern.ch/Shift/)

This run was just after a hardware shift of the clock of -0.5ns. The t0 calibration detected a consistent variation (-0.38ns)and updated the t0 value, that has been picked up at the next run change. So next run you see again the better overlap with the reference. The difference of the mean of the histograms between 179440 and 179441 is 0.19ns (we update the t0 when the difference is more than 0.1ns - numbers not directly comparable, I haven't done any fit on the DQ histograms, but it is to get the idea).

Given that we allow for shifts of ~0.1ns, this would make a difference in the efficiency of reconstructing tracks below the per-mille level (see Lucia's talk in the references for this estimation). So, please include also the OT team to inform them and in case they have different opinion, but this run looks OK.

Standing issues (still under investigation)

• Few runs with different Track Type distribution (177067-177090; 179346-179347)

https://lblogbook.cern.ch/Data+Quality/11604

https://lblogbook.cern.ch/Data+Quality/12507

Current status: Comparing the Data Quality files (after Hlt2), the following distributions look different: track type distribution (which triggered the studies), hits per TT layer (for long tracks), eta distribution for velo tracks (and related distributions), No difference is observed comparing Online Brunel files (after Hlt1). Data look good, but since it is not understood the origin of the feature, runs are still in a un-flagged (UNKNOWN) status for the time being.

Useful references

• Lucia Grillo, Data Quality studies, T&A 23/08/2016: https://indico.cern.ch/event/506008/contributions/2282941/attachments/1326031/1990634/DQ_TandA2308.pdf
  
• Silvia Borghi, New VELO Alignment with 2010 Collision data, 19/08/2010 https://indico.cern.ch/event/102191/contributions/10308/attachments/7443/11063/VeloAlign20100819.pdf
• Agnieszka Dziurda, PV: maximal VELO centering shift, T&A 14/04/2016: https://indico.cern.ch/event/485533/contributions/2133156/attachments/1257971/1857878/2016-04-14-PV-Shift-Treshold.pdf
• Agnieszka Dziurda, Silvia Borghi, PV reconstruction and VELO off-centering in fill 4452, PPTS 21/03/2016: https://indico.cern.ch/event/502824/contributions/2021706/attachments/1246336/1835866/2016-03-21-PV-Threshold-PPTS.pdf
• Lucia Grillo, OT Drift Time Calibration, T&A 27/01/2015: https://indico.cern.ch/event/363153/contributions/1778406/attachments/722338/991500/OTDriftTimeCalibrationv1.pdf

-- LuciaGrillo - 2016-08-24