TPC timeseries: Propagating TPC track to TOF cluster by matthias-kleiner · Pull Request #12312 · AliceO2Group/AliceO2

matthias-kleiner · 2023-11-23T18:16:27Z

No description provided.

matthias-kleiner · 2023-11-23T18:20:54Z

+        float tpcYatTOF = 0;
+        float tpcZatTOF = 0;
+        if (hasTOFCluster) {
+          if (trackTmp.rotate(o2::math_utils::sector2Angle(tofCl.getSector())) && propagator->propagateTo(trackTmp, tofCl.getX(), false, mMaxSnp, mFineStep, mMatType)) {


Hello @shahor02
we wanted to get for each TPC track that has a TOF cluster assigned the TPC track propagated to the TOF cluster position and then compare the propagated TPC track position with the TOF cluster position.
Can you take a look if I did it properly?
I am also not quite sure if the propagation is correctly done. I assumed here that the tfCl.getX() returns the local X position of the cluster.

One should not use the TPC innermost track param for propagation to TOF, instead use (since you don't use errors)

o2::track::TrackPar trackTmpOut(tracksTPC[iTrk].getParamOut()); ...

shahor02 · 2023-11-23T18:28:04Z

+    // loop over ITS-TPC-TRD-TOF and ITS-TPC-TOF tracks an store for each ITS-TPC track the TOF track index
+    for (const auto& tofMatch : tofMatches) {
+      for (const auto& tpctofmatch : tofMatch) {
+        auto refTPC = recoData.getSingleDetectorRefs(tpctofmatch.getTrackRef())[TrkSrc::TPC];


since you need only TPC index, you can use faster method

auto refTPC = recoData.getTPCContributorGID(tpctofmatch.getTrackRef());

shahor02 · 2023-11-23T18:37:47Z

-  void fillDCA(const gsl::span<const TrackTPC> tracksTPC, const gsl::span<const o2::dataformats::TrackTPCITS> tracksITSTPC, const gsl::span<const o2::dataformats::PrimaryVertex> vertices, const int iTrk, const int iThread, const std::unordered_map<unsigned int, std::array<int, 2>>& indicesITSTPC, const gsl::span<const o2::its::TrackITS> tracksITS)
+  void fillDCA(const gsl::span<const TrackTPC> tracksTPC, const gsl::span<const o2::dataformats::TrackTPCITS> tracksITSTPC, const gsl::span<const o2::dataformats::PrimaryVertex> vertices, const int iTrk, const int iThread, const std::unordered_map<unsigned int, std::array<int, 2>>& indicesITSTPC, const gsl::span<const o2::its::TrackITS> tracksITS, const std::vector<int>& idxTPCTrackToTOFCluster, const gsl::span<const o2::tof::Cluster> tofClusters)
  {
    TrackTPC track = tracksTPC[iTrk];


I see that you are still cloning the full TPC track. Why? For the propagation it is enough to clone TrackParCov.

Thanks, I thought since we want the errors we would still need then the full track. But you are right, thanks.

shahor02 · 2023-11-23T18:45:02Z

+        float tpcYatTOF = 0;
+        float tpcZatTOF = 0;
+        if (hasTOFCluster) {
+          if (trackTmp.rotate(o2::math_utils::sector2Angle(tofCl.getSector())) && propagator->propagateTo(trackTmp, tofCl.getX(), false, mMaxSnp, mFineStep, mMatType)) {


One should not use the TPC innermost track param for propagation to TOF, instead use (since you don't use errors)

o2::track::TrackPar trackTmpOut(tracksTPC[iTrk].getParamOut()); ...

- storing delta parameter of TPC between inner and outer param - small fixes

- storing getDXatTOF and getDZatTOF

miranov25 · 2023-11-24T12:57:20Z

Hello @matthias-kleiner

Please include mDXatTOF and mDZatTOF for combined tracks as well.

If I come across anything later, we can make the necessary updates. I'm considering including TOF dY and dZ of combined tracks not only in the skimmed data but also to the time series.

In the pull request, let's explicitly provide the values for dY and dZ, and inquire whether they seem reasonable. The variation in dY suggests that the sectors may not be aligned correctly. I am checking what could be the reason (TPC or TOF?), but the code should be checked if alignment correction was applied.

miranov25 · 2023-11-24T16:13:31Z

TOF observations based on the RootInteractive dashboard:

observed TOF dY bias in respect to TPC tracks on A side and C side is similar - bias per sector bias of O(3 cm)
observed TOF dZ bias in respect to TPC tracks on A side and C side is different- up-down bias (up 0-9 down -9-18)

shahor02 · 2023-11-24T16:19:39Z

@miranov25 is this with the corrected TPC outerParam track extrapolation or previous, inner param extrapolation to TOF?

matthias-kleiner · 2023-11-24T16:21:22Z

@miranov25 is this with the corrected TPC outerParam track extrapolation or previous, inner param extrapolation to TOF?

This is with the latest version of the code like you suggested

shahor02 · 2023-11-24T17:00:05Z

Very strange, could you produce a similar plot for the opposite polarity? What is the IR of this run?
The residuals I saw before and after the alignment (ITS/TRD/TOF refits, no TPC) are
https://alice.its.cern.ch/jira/secure/attachment/60038/algrepB0_Lev1ok3_vs_def.pdf
https://alice.its.cern.ch/jira/secure/attachment/60036/algrepBMinus_Lev1ok3_vs_def.pdf
https://alice.its.cern.ch/jira/secure/attachment/60034/algrepBPlus_Lev1ok3_vs_def.pdf
for B0, B- and B+ resp.

matthias-kleiner · 2023-11-24T17:03:14Z

Very strange, could you produce a similar plot for the opposite polarity? What is the IR of this run? The residuals I saw before and after the alignment (ITS/TRD/TOF refits, no TPC) are https://alice.its.cern.ch/jira/secure/attachment/60038/algrepB0_Lev1ok3_vs_def.pdf https://alice.its.cern.ch/jira/secure/attachment/60036/algrepBMinus_Lev1ok3_vs_def.pdf https://alice.its.cern.ch/jira/secure/attachment/60034/algrepBPlus_Lev1ok3_vs_def.pdf for B0, B- and B+ resp.

We looked at different runs: 544126.50Hz, 544124.8kHz, 544122.18kHz, 544116.38kHz. It was the same for all of them.
I can try to run it for the different polarity

miranov25 · 2023-11-24T17:39:34Z

The residuals in tofmatch are ok. The problem is only when we are doing own matching. I assume it could be because we read clusters from root file, maybe some corrections were not applied.

shahor02 · 2023-11-24T18:06:20Z

The residuals in tofmatch are ok. The problem is only when we are doing own matching. I assume it could be because we read clusters from root file, maybe some corrections were not applied.

That would be strange, but better @noferini comments.

noferini · 2023-11-24T18:26:42Z

We don't have any additional correction on TOF clusters. What is read is used.
However, when the residuals are calculated in TOF matching algorithm TPC tracks get the TOF time and then they can be shifted along Z.
If tracks are provided with a t0 they are shifted by (t0 - t_tof)*vdrift since TOF time is more precise (this is the only TOF constraint to the track).
I don't know if this can explain the difference (I assume this is not done in your own code).
Can you point me to your matching algorithm?

matthias-kleiner · 2023-11-24T20:59:55Z

+        float tpcZDeltaAtTOF = -999;
+        if (hasTOFCluster) {
+          o2::track::TrackPar trackTmpOut(tracksTPC[iTrk].getParamOut());
+          if (trackTmpOut.rotate(o2::math_utils::sector2Angle(tofCl.getSector())) && propagator->propagateTo(trackTmpOut, tofCl.getX(), false, mMaxSnp, mFineStep, mMatType)) {


Hello @noferini
the propagation with the TPC track to TOF cluster is done here.
When we calculate the difference between TPC and TOF cluster in z (tpcZDeltaAtTOF) direction, we are also doing some correction, but I am not fully sure if this is correct:
tpcZDeltaAtTOF = signSide * (o2::tpc::ParameterElectronics::Instance().ZbinWidth * trackFull.getTime0() - vertex.getTimeStamp().getTimeStamp()) * mVDrift - trackTmpOut.getZ() + tofCl.getZ();

For the difference in y we are just doing tpcYDeltaAtTOF = trackTmpOut.getY() - tofCl.getY();

@matthias-kleiner, I am not completely sure about the signs in tpcZDeltaAtTOF formula but for tpcYDeltaAtTOF there are no problems. I don't see any evident problem in the propagation call.
Do you have also tpcZDeltaAtTOF vs phi/sector?

Hi @noferini ,
yes sure it looks like this.

It looks much better than Y residual.
What is the fine-step you are using in propagation? 2 cm (default)?
Does this play a role?

miranov25 · 2023-11-25T08:07:11Z

Dear @matthias-kleiner, @shahor02 and @noferini

I've uploaded the RootInteractive dashboard with all observations to EOS. This allows you to test all residual data (DCA, delta parameters at the vertex, and TOF) as a function of the rate and kinematic variables.

The data is grouped into three categories, facilitating functional decomposition.

    varKine0=["qpt_10","tgl_10","side_type"]
    varKine1=["qpt_10","tgl_10","side_type","dSec_10"]
    varKine2=["qpt_10","tgl_10","side_type","sec_10"]

For TOF, we have data for the standard delta (as in official reconstruction) and for the TPC-only track deltas. I won't have time to follow that pull request during the weekend, but I'll post some observations here that might be helpful. From our perspective, it will be important to integrate TPC-TOF into production. Even if we don't obtain absolute values, relative values will still be useful.

More details will be provided in the following posts.

Marian

miranov25 · 2023-11-25T08:16:22Z

Data aggregation analysis and interactive dashboard:

https://gitlab.cern.ch/alice-tpc-offline/alice-tpc-notes/-/blob/f3edfc150d9330b7557fdf2e255c40168b67ad41/JIRA/ATO-615/analyzeTimeSeriesSkimmed.py#L136-190
/eos/home-m/mivanov/public/NOTES/JIRA/time_series_analysis/dashboard_SkimmedG2G0.html

miranov25 · 2023-11-25T08:23:28Z

TOF residuals as function of rate:

"official delta" from TOFmatch as function of rate

TPC-TOF delta which we extracted in order to calibrate and characterize TPC distortion

miranov25 · 2023-11-25T08:37:49Z

DCA for the TPCITS track to be checked

looks like vertex position is not used

miranov25 · 2023-11-25T08:49:56Z

TPC--> DCA Bias with CPass1 Type of Distortion Calibration

The DCA bias should also be reflected in the TPC-TOF bias once the matching algorithm is fixed.

Modulation at sector boundaries as a function of rate: O 0.5 cm

* Modulation removed by the factorized fits, except for some hotspot regions that should be treated separately.

miranov25 · 2023-11-26T21:37:31Z

Hello all.

Do you have any suggestions for what could be the problem? I asked Matthias to create a full TOF cluster and TOF match dump so we can debug it. Indeed, this will be a good option to have in general (switched OFF in production)

shahor02 · 2023-11-26T22:50:31Z

@noferini actually, in the TOF matching you are not using TOF cluster positions as they are stored, using

AliceO2/Detectors/GlobalTracking/src/MatchTOF.cxx

Line 844 in 42a8244

Geo::getPadDxDyDz(posFloat, detIdTemp, deltaPosTemp, sec);

instead. Also, in the alignment I am recalculating the coordinates from the channels ideal positions with particular "starting" alignment matrices:

AliceO2/Detectors/Align/src/AlignableDetectorTOF.cxx

Lines 95 to 109 in 42a8244

    
           const auto& clus = TOFClusters[gid.getIndex()]; 
        
           int det[5] = {}, ch = clus.getMainContributingChannel(); 
        
           o2::tof::Geo::getVolumeIndices(ch, det); 
        
           int sensID = o2::tof::Geo::getStripNumberPerSM(det[1], det[2]) + clus.getSector() * o2::tof::Geo::NSTRIPXSECTOR; 
        
           auto* sensor = (AlignableSensorTOF*)getSensor(sensID); 
        
           if (sensor->isDummy()) { 
        
             LOGP(error, "Dummy sensor {} is referred by a track", sensID); 
        
             return 0; 
        
           } 
        
           double loc[3] = {(det[4] + 0.5) * o2::tof::Geo::XPAD - o2::tof::Geo::XHALFSTRIP, 0., (det[3] - 0.5) * o2::tof::Geo::ZPAD}, locCorr[3] = {}, traCorr[3] = {}; 
        
           const auto& matAlg = sensor->getMatrixClAlg(); 
        
           matAlg.LocalToMaster(loc, locCorr); 
        
           // rotate to tracking 
        
           const auto& matT2L = sensor->getMatrixT2L(); 
        
           matT2L.MasterToLocal(locCorr, traCorr);

.

Are you sure the coordinates as they are set in the clustering are equivalent to those set in the matching?

miranov25 · 2023-11-27T10:32:45Z

Hello @shahor02 and @noferini

Should we perform the "realignment" in our code, as indicated by @shahor02 in the code snippet, or do you prefer another solution?

@noferini actually, in the TOF matching you are not using TOF cluster positions as they are stored, using

AliceO2/Detectors/GlobalTracking/src/MatchTOF.cxx

Line 844 in 42a8244

Geo::getPadDxDyDz(posFloat, detIdTemp, deltaPosTemp, sec);

instead. Also, in the alignment I am recalculating the coordinates from the channels ideal positions with particular "starting" alignment matrices:

AliceO2/Detectors/Align/src/AlignableDetectorTOF.cxx

Lines 95 to 109 in 42a8244

const auto& clus = TOFClusters[gid.getIndex()];

int det[5] = {}, ch = clus.getMainContributingChannel();

o2::tof::Geo::getVolumeIndices(ch, det);

int sensID = o2::tof::Geo::getStripNumberPerSM(det[1], det[2]) + clus.getSector() * o2::tof::Geo::NSTRIPXSECTOR;

auto* sensor = (AlignableSensorTOF*)getSensor(sensID);

if (sensor->isDummy()) {

LOGP(error, "Dummy sensor {} is referred by a track", sensID);

return 0;

}

double loc[3] = {(det[4] + 0.5) * o2::tof::Geo::XPAD - o2::tof::Geo::XHALFSTRIP, 0., (det[3] - 0.5) * o2::tof::Geo::ZPAD}, locCorr[3] = {}, traCorr[3] = {};

const auto& matAlg = sensor->getMatrixClAlg();

matAlg.LocalToMaster(loc, locCorr);

// rotate to tracking

const auto& matT2L = sensor->getMatrixT2L();

matT2L.MasterToLocal(locCorr, traCorr);

.
Are you sure the coordinates as they are set in the clustering are equivalent to those set in the matching?

miranov25 · 2023-11-27T10:55:33Z

Chiara just informed me that Francesco is on a trip, so we won't get an update soon. As we would like to test other aspects of the code in the test, I suggest merging it as it is and fixing the TOF alignment issue later. Do you agree? Can we proceed with the merge?
@shahor02, @matthias-kleiner ?

matthias-kleiner · 2023-11-27T11:01:30Z

Chiara just informed me that Francesco is on a trip, so we won't get an update soon. As we would like to test other aspects of the code in the test, I suggest merging it as it is and fixing the TOF alignment issue later. Do you agree? Can we proceed with the merge? @shahor02, @matthias-kleiner ?

It is fine from my side. I removed the draft

shahor02 · 2023-11-27T11:10:12Z

will merge after the CI but not before 4pm (in case @noferini responds)

miranov25 · 2023-11-27T15:42:04Z

Hello @shahor02
I think CI was done. There are only unresolved conversations, which we will resolve later.
Can we merge?
Thanks
Marian

noferini · 2023-11-27T22:34:23Z

@noferini actually, in the TOF matching you are not using TOF cluster positions as they are stored, using

AliceO2/Detectors/GlobalTracking/src/MatchTOF.cxx

Line 844 in 42a8244

Geo::getPadDxDyDz(posFloat, detIdTemp, deltaPosTemp, sec);

instead. Also, in the alignment I am recalculating the coordinates from the channels ideal positions with particular "starting" alignment matrices:

AliceO2/Detectors/Align/src/AlignableDetectorTOF.cxx

Lines 95 to 109 in 42a8244

const auto& clus = TOFClusters[gid.getIndex()];

int det[5] = {}, ch = clus.getMainContributingChannel();

o2::tof::Geo::getVolumeIndices(ch, det);

int sensID = o2::tof::Geo::getStripNumberPerSM(det[1], det[2]) + clus.getSector() * o2::tof::Geo::NSTRIPXSECTOR;

auto* sensor = (AlignableSensorTOF*)getSensor(sensID);

if (sensor->isDummy()) {

LOGP(error, "Dummy sensor {} is referred by a track", sensID);

return 0;

}

double loc[3] = {(det[4] + 0.5) * o2::tof::Geo::XPAD - o2::tof::Geo::XHALFSTRIP, 0., (det[3] - 0.5) * o2::tof::Geo::ZPAD}, locCorr[3] = {}, traCorr[3] = {};

const auto& matAlg = sensor->getMatrixClAlg();

matAlg.LocalToMaster(loc, locCorr);

// rotate to tracking

const auto& matT2L = sensor->getMatrixT2L();

matT2L.MasterToLocal(locCorr, traCorr);

.

Are you sure the coordinates as they are set in the clustering are equivalent to those set in the matching?

Hi @shahor02 ,
sorry for the delay. Unfortunately during the day it is difficult for me to be online and check in the next 2 days.
I had a look at the code.
What is stored in tofclusters (XYZ) is defined in the TOFClusterer.cxx (line 275) and taken from Geo::GetPos(...) method.
In principle this takes the loaded geometry and I would not expect any difference wrt what done in MatchTOF but in the case the proper geometry is not loaded during clusterization.
I will try to compare the coordinates got in the two ways as soon as I can.

noferini · 2023-11-30T16:33:25Z

Hi, I did further checks inside the TOF matching algorithm comparing the position extracted with

getPadDxDyDz()
getPos
XYZ from tof clusters
they are consistent

) * TPC timeseries: Propagating TPC track to TOF cluster * Dont copy full TPC track - storing delta parameter of TPC between inner and outer param - small fixes * Fixing deltaZ of TPC time when comparing to TOF cluster z - storing getDXatTOF and getDZatTOF * Fixing tpc only mode

TPC timeseries: Propagating TPC track to TOF cluster

aba564a

matthias-kleiner commented Nov 23, 2023

View reviewed changes

shahor02 requested changes Nov 23, 2023

View reviewed changes

matthias-kleiner added 3 commits November 23, 2023 20:52

Dont copy full TPC track

0f19627

- storing delta parameter of TPC between inner and outer param - small fixes

Fixing deltaZ of TPC time when comparing to TOF cluster z

9c8b58d

- storing getDXatTOF and getDZatTOF

Fixing tpc only mode

f5e8279

matthias-kleiner commented Nov 24, 2023

View reviewed changes

matthias-kleiner marked this pull request as ready for review November 27, 2023 11:00

matthias-kleiner requested review from davidrohr and wiechula as code owners November 27, 2023 11:00

shahor02 merged commit 9aad898 into AliceO2Group:dev Nov 27, 2023

Conversation

matthias-kleiner commented Nov 23, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

miranov25 commented Nov 24, 2023

Uh oh!

miranov25 commented Nov 24, 2023

TOF observations based on the RootInteractive dashboard:

Uh oh!

shahor02 commented Nov 24, 2023

Uh oh!

matthias-kleiner commented Nov 24, 2023

Uh oh!

shahor02 commented Nov 24, 2023

Uh oh!

matthias-kleiner commented Nov 24, 2023

Uh oh!

miranov25 commented Nov 24, 2023

Uh oh!

shahor02 commented Nov 24, 2023

Uh oh!

noferini commented Nov 24, 2023

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

miranov25 commented Nov 25, 2023

Uh oh!

miranov25 commented Nov 25, 2023 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Data aggregation analysis and interactive dashboard:

Uh oh!

miranov25 commented Nov 25, 2023

TOF residuals as function of rate:

Uh oh!

miranov25 commented Nov 25, 2023

DCA for the TPCITS track to be checked

Uh oh!

miranov25 commented Nov 25, 2023

TPC--> DCA Bias with CPass1 Type of Distortion Calibration

Uh oh!

miranov25 commented Nov 26, 2023

Uh oh!

shahor02 commented Nov 26, 2023

Uh oh!

miranov25 commented Nov 27, 2023

Uh oh!

miranov25 commented Nov 27, 2023

Uh oh!

matthias-kleiner commented Nov 27, 2023

Uh oh!

shahor02 commented Nov 27, 2023

Uh oh!

miranov25 commented Nov 27, 2023

Uh oh!

noferini commented Nov 27, 2023

Uh oh!

noferini commented Nov 30, 2023

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

miranov25 commented Nov 25, 2023 •

edited

Loading