Add extra info with charge and timing to unbinned residuals #14969

shahor02 · 2026-01-17T23:35:42Z

No description provided.

github-actions · 2026-01-17T23:35:49Z

REQUEST FOR PRODUCTION RELEASES:
To request your PR to be included in production software, please add the corresponding labels called "async-" to your PR. Add the labels directly (if you have the permissions) or add a comment of the form (note that labels are separated by a ",")

+async-label <label1>, <label2>, !<label3> ...

This will add <label1> and <label2> and removes <label3>.

The following labels are available
async-2023-pbpb-apass4
async-2023-pp-apass4
async-2024-pp-apass1
async-2022-pp-apass7
async-2024-pp-cpass0
async-2024-PbPb-apass1
async-2024-ppRef-apass1
async-2024-PbPb-apass2
async-2023-PbPb-apass5

shahor02 · 2026-01-17T23:41:43Z

@miranov25 , this adds to the unbinned residuals tree a branch with 4-bytes struct https://github.com/shahor02/AliceO2/blob/6c23010c8442f96250eb35358bd3c73ca04a9723/Detectors/TPC/calibration/SpacePoints/include/SpacePoints/TrackInterpolation.h#L104-L145 encoding the TPC dedx, TRD q's and slope (truncated to +-3.5 range), TOF cluster time wrt track ITS-TPC time and the same for the PV (i.e. one should use the relevant getter depending on the row value).
I've copied residuals for 1 ctf of OO/564356 to /alice/cern.ch/user/s/shahoian/toMI/o2residuals_tpc.root.
Hope the dump below is self-explanatory:

root [1] unbinnedResid->Scan("res.row : detInfo.word : detInfo.qTotTPC() : detInfo.qMaxTPC() : detInfo.q0TRD() : detInfo.q1TRD() : detInfo.q2TRD() : detInfo.slopeTRD() : detInfo.timeTOF() : detInfo.timePV()")
***********************************************************************************************************************************************
*    Row   * Instance *  res.row  *  detInfo. *  detInfo. *  detInfo. *  detInfo. *  detInfo. *  detInfo. *  detInfo. *  detInfo. *  detInfo. *
***********************************************************************************************************************************************
# RS: TPC                                            QTOT          QMAX
*        0 *        0 *        17 *    524297 *         9 *         8 *         9 *         0 *        32 *      -3.5 *         0 *         0 *
*        0 *        1 *        18 *    917532 *        28 *        14 *        28 *         0 *        56 *      -3.5 *         0 *         0 *
*        0 *        2 *        19 *    589850 *        26 *         9 *        26 *         0 *        36 *      -3.5 *         0 *         0 *
*        0 *        3 *        20 *    458776 *        24 *         7 *        24 *         0 *        28 *      -3.5 *         0 *         0 *
*        0 *        4 *        21 *   1179684 *        36 *        18 *        36 *         0 *         8 * -3.498290 *         0 *         0 *
*        0 *        5 *        22 *   1048636 *        60 *        16 *        60 *         0 *         0 * -3.498290 *         0 *         0 *
*        0 *        6 *        23 *    589852 *        28 *         9 *        28 *         0 *        36 *      -3.5 *         0 *         0 *
*        0 *        7 *        24 *   1310792 *        72 *        20 *        72 *         0 *        16 * -3.498290 *         0 *         0 *
*        0 *        8 *        25 *    655386 *        26 *        10 *        26 *         0 *        40 *      -3.5 *         0 *         0 *
*        0 *        9 *        26 *    852024 *        56 *        13 *        56 *         0 *        52 *      -3.5 *         0 *         0 *
*        0 *       10 *        27 *    786454 *        22 *        12 *        22 *         0 *        48 *      -3.5 *         0 *         0 *
*        0 *       11 *        28 *   1179720 *        72 *        18 *        72 *         0 *         8 * -3.498290 *         0 *         0 *
...
# RS: TRD                                                                      Q0          Q1          Q2   Slope
*        0 *      132 *       162 * 2.470e+09 *      2304 *     37699 *         0 *        18 *        12 * 0.5273504 * -2.46e-27 * -2.46e-27 *
*        0 *      133 *       163 * 2.914e+09 *     18816 *     44467 *         0 *        19 *        13 * 1.2504272 * -2.03e-11 * -2.03e-11 *
*        0 *      134 *       164 * 2.761e+09 *     16400 *     42132 *        16 *         0 *        17 * 1.0008549 * -6.42e-17 * -6.42e-17 *
*        0 *      135 *       165 * 2.567e+09 *      4096 *     39172 *         0 *        32 *        16 * 0.6846156 * -6.82e-24 * -6.82e-24 *
...
# RS: TOF : difference between the TOF cluster time and the seeding ITS-TPC track time, \mus                                 XXX
*        0 *      136 *       170 * 3.166e+09 *     63764 *     48309 *        20 *       114 *        23 * 1.6606836 * -0.022213 * -0.022213 *
...
# RS: PV: difference between the PV time and the seeding ITS-TPC track time, \mus                                                      XXX
*        0 *      144 *       190 * 3.172e+09 *     32768 *     48405 *         0 *         0 *        22 * 1.6709404 * -0.036499 * -0.036499 *
...

miranov25 · 2026-01-19T09:04:48Z

Hello @shahor02

Thank you.
I assume that detInfo.word is parallel to res – we have as many detInfo entries as residuals.
I am now checking the context.

If my understanding is correct, then it will work.
I have one more question.
I need to define the occupancy.
In the current residuals, we have neither occupancy nor time. What is the best way to add them?
I am considering the occupancy vector we have in the reconstruction, but we will need to assign time to the track to be able to use it.

In meantime I will check your file

shahor02 · 2026-01-19T12:05:44Z

@miranov25 yes, detInfo elements are in sync with res, for ITS 0's are filled.
Concerning the occupancy: the current format has no PV-level info and the tracks ordering is randomized on purpose: adding for each track the 32 floats of our usual TPC occupancy will double the output size.
If you want, I can store per track the N TPC clusters in 1 drift time from its PV time.

miranov25 · 2026-01-19T15:59:15Z

Hello @shahor02

If you want, I can store per track the N TPC clusters in 1 drift time from its PV time.

That can work.

Another options
1.) What I suggested today, however, is to integrate within the time window covered by the track, i.e. between the two drift-time boundaries (entrance and end of the TPC)

time_80 = time_event. + drift80= time_event. + 100*(1-80*|tgl|/250.)
tim250 = time_event. +drift250=ime_event. + min(100*(1-250*|tgl|/250.),0)

2.) For occupancy bins per track defined in similar way:
IROC,OROC0,OROC1,OrOC2 radius - 1 Byte per region
4 bytes per track

I will write formula soon

shahor02 · 2026-01-20T13:45:28Z

@miranov25 shall I merge this part as is or wait for your formula?

miranov25 · 2026-01-20T14:05:36Z

Hello @shahor02

We need to use a variable with "semi" relative precision. We can use a simple logarithmic function or the asinh function, which is nearly linear for small x (below the scale value) and logarithmic for larger values.

If the occupancy is given as the number of clusters, we can use the scale as in my TF1 without normalisation.

I have to check again the scale. Can you access our occupancy estimator from reconstructions?

TF1 f1("ncl","asinh(x*0.05)/0.05",0,10000000)
root [25] f1.Eval(1000000)
(double) 230.25851

shahor02 · 2026-01-20T14:10:01Z

The residuals extraction has access to David's vector of occupancies per 16 TBs. But I don't understand which value you want to put in your TF1

miranov25 · 2026-01-20T15:55:17Z

I want to calculate the mean occupancy per track segment as discussed above. #14969 (comment)

Occupancy in radial segments IROC, OROC0, 1, 2, using our vectors of occupancies.
It is easy to make a mistake.
Are the formulas clear, or should I try to implement it using David's vector?

shahor02 · 2026-01-20T16:33:15Z

If you want to provide a code snippet, go on, I'll not work on this before tomorrow afternoon.
Note that the tracks may have not IROC of OROC3 contributions, I guess in that case you should count TBs corresponding to the TPC track extrapolation to min and max possible pad rows.

miranov25 · 2026-01-20T22:52:38Z

I do not know what we have available at the moment when we produce the clusters.
The easiest way will be to loop over clusters within the stacks and sum the occupancy at a particular time bin (from the occupancy lookup).
It is not very efficient, but there is a lower probability of making a mistake.

This will automatically take into account only rows with clusters.

miranov25 · 2026-01-21T14:29:53Z

Hello @noferini,

During the TPC meeting, we discussed the stored times in the "unbinned residuals" used for TPC calibration.
To provide some context: I had previously asked Ruben to implement TOF time in the calibration output to help constrain the curvature measurement. However, Ruben noted that we are currently storing the raw TOF time, whereas we need to access TOF time minus the collision time (t-t0).
Could you please review the current time estimator in this pull request and suggest how best to apply this correction (similar to what you did for the time series)?

Thank you!

noferini · 2026-01-21T14:54:15Z

Hi @miranov25 ,
I'm going to check.
More news soon...

noferini · 2026-01-21T15:03:36Z

Detectors/TPC/calibration/SpacePoints/src/TrackInterpolation.cxx

+          if (!gidTable[GTrackID::ITSTPC].isIndexSet()) {
+            LOGP(fatal, "ITS-TPC seed index is not set for TOF track");
+          }
+          float tdif = static_cast<float>(clTOF.getTime() - mRecoCont->getTPCITSTrack(gidTable[GTrackID::ITSTPC]).getTimeMUS().getTimeStamp() * 1e6);


Hi @miranov25 ,
if I understood correctly this is the line you want to adjust.
I see that this is a delta time wrt ITS-TPC track time inside the TF.
What is the precision you need?
At this stage clTOF.getTime() is not a raw time but an already calibrated time.
It is true that the t0 is affected of a "large" smearing (~200 ps) and we can improve by using FT0 reducing the effective resolution on t.o.f. at the level < 80 ps but do you really gain anything here?
One additional comment. If you want to achieve such a precision, I guess you should also refer the time to the Primary Vertex otherwise you would have shifts at the level of 10-20 ns (time travel from PV to TOF).
Actually, what do you need?

Hi @noferini
If the TOF contributes to the global track, then its cluster time - estimated time-of-flight-from-Lintegral is assigned to a track with 10ns error.
For the global tracks w/o TOF but with the TRD, the time assignment is done with 5ns error (unless the pile-up was detected from FT0, in this case the distance from the pile-up time to TRD trigger BC time is assigned as an error, but in PbPb it is very rare).
Then these times are used to fit the PV timestamp, so it is a weighted average between the timestamps from TRD(and noTOF) fixed to BC center and TOF-defined timestamps. If precision of this vertex time is fine, then I can use it as a time for TOF. Otherwhise, I would need the time which you are using for the TOF PID

Add extra info with charge and timing to unbinned residuals

6c23010

shahor02 requested review from davidrohr and wiechula as code owners January 17, 2026 23:35

noferini reviewed Jan 21, 2026

View reviewed changes

Add extra info with charge and timing to unbinned residuals #14969

Are you sure you want to change the base?

Add extra info with charge and timing to unbinned residuals #14969

Conversation

shahor02 commented Jan 17, 2026

Uh oh!

github-actions bot commented Jan 17, 2026

Uh oh!

shahor02 commented Jan 17, 2026

Uh oh!

miranov25 commented Jan 19, 2026

Uh oh!

shahor02 commented Jan 19, 2026

Uh oh!

miranov25 commented Jan 19, 2026

Uh oh!

shahor02 commented Jan 20, 2026

Uh oh!

miranov25 commented Jan 20, 2026

Uh oh!

shahor02 commented Jan 20, 2026

Uh oh!

miranov25 commented Jan 20, 2026

Uh oh!

shahor02 commented Jan 20, 2026

Uh oh!

miranov25 commented Jan 20, 2026

Uh oh!

miranov25 commented Jan 21, 2026

Uh oh!

noferini commented Jan 21, 2026

Uh oh!

noferini Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shahor02 Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Milestone

Development

Uh oh!

3 participants