[Draft] #2963 Review TRD Hit Pos Uncertainty

added 57 commits

c9bf37b0...e7027782 - 48 commits from branch computing:master
c1ba0f6e - Format comments
90a0cab3 - Delete function GetSpaceResolution
71266483 - Rework function CbmTrdModuleRecR::MakeHit
5d57886e - Add minor corrections to MakeHit
9c812a69 - Code formatting
b46e5e37 - Consolidate overloads of GetPadPosition.
6c38a33d - Add preliminary values for position error.
5c49b6c7 - Re-add an angle correction from Etienne.
74960c3c - Remove error classes.

Compare with previous version

added 1 commit

c217202f - Code formatting

Compare with previous version

added 95 commits

c217202f...68f68cfe - 85 commits from branch computing:master
c1af4f9f - Format comments
001f5a44 - Delete function GetSpaceResolution
2530f177 - Rework function CbmTrdModuleRecR::MakeHit
aa2baf71 - Add minor corrections to MakeHit
fad350be - Code formatting
6014b773 - Consolidate overloads of GetPadPosition.
47f46df2 - Add preliminary values for position error.
996f8b70 - Re-add an angle correction from Etienne.
22746829 - Remove error classes.
e637ddb3 - Code formatting

Compare with previous version

added 34 commits

e637ddb3...73221de0 - 24 commits from branch computing:master
c34aea08 - Format comments
48cf057b - Delete function GetSpaceResolution
70fd0b1d - Rework function CbmTrdModuleRecR::MakeHit
df8a636d - Add minor corrections to MakeHit
18c49c99 - Code formatting
b3416be0 - Consolidate overloads of GetPadPosition.
fe4239d9 - Add preliminary values for position error.
bec1c820 - Re-add an angle correction from Etienne.
1b96d7cf - Remove error classes.
d0af1e93 - Code formatting

Compare with previous version

added 68 commits

d0af1e93...7bc1ea7c - 58 commits from branch computing:master
f7359731 - Format comments
b6e13c79 - Delete function GetSpaceResolution
36e7121f - Rework function CbmTrdModuleRecR::MakeHit
ce56c578 - Add minor corrections to MakeHit
03e62847 - Code formatting
49f90684 - Consolidate overloads of GetPadPosition.
a267f443 - Add preliminary values for position error.
cf8f845e - Re-add an angle correction from Etienne.
37f41c74 - Remove error classes.
2bc24998 - Code formatting

Compare with previous version

@f.uhlig I need your help here. Some of the tests in the pipeline constantly fail. Until last week, the same two tests failed on all operating systems. Now, after rebasing again, the tests suddenly pass, except for the ones on macOS. I did not change anything in the code. Thanks in advance!

@d.spicker,

this is what I see on our CDash server in one of the failing tests.

Assertion failed: (binParamMap.find(bin) != binParamMap.end()), function FollowTracks, file CbmLitTrackFinderBranch.cxx, line 161.
/Users/fairroot/builds/trd/cbmsoft/cbmroot/build/bin/cbmreco_offline.sh: line 5: 18728 Abort trap: 6           /Users/fairroot/builds/trd/cbmsoft/cbmroot/build/bin/cbmreco_offline "$@"

I have no clue what went wrong here and have to check in more detail. Unfortunately I will have no time this week and have to postpone it to next week.

added 1 commit

d1381abe - Set hit class type for newly created hits

Compare with previous version

added 153 commits

d1381abe...63eb442b - 142 commits from branch computing:master
12fd0a1d - 1 earlier commit
6ac6a1c4 - Delete function GetSpaceResolution
c984028f - Rework function CbmTrdModuleRecR::MakeHit
6f988fba - Add minor corrections to MakeHit
695d4906 - Code formatting
54c16c08 - Consolidate overloads of GetPadPosition.
c117ebd5 - Add preliminary values for position error.
b9e79517 - Re-add an angle correction from Etienne.
24c3cc31 - Remove error classes.
4d088cb3 - Code formatting
47510056 - Set hit class type for newly created hits

Compare with previous version

added 1 commit

3f151fc4 - Code formatting

Compare with previous version

added 128 commits

3f151fc4...7c017b9d - 116 commits from branch computing:master
7c017b9d...bb792b4b - 2 earlier commits
ca8a62f2 - Rework function CbmTrdModuleRecR::MakeHit
0347aea7 - Add minor corrections to MakeHit
74619ba2 - Code formatting
e445b198 - Consolidate overloads of GetPadPosition.
2045e453 - Add preliminary values for position error.
fa06d5fd - Re-add an angle correction from Etienne.
374d0ec5 - Remove error classes.
2ca6c586 - Code formatting
4f0f5603 - Set hit class type for newly created hits
8717719b - Code formatting

Compare with previous version

added 1 commit

d6455be1 - Change Hit time error

Compare with previous version

added 1 commit

b08e2dad - Change Hit time error

Compare with previous version

@p.-a.loizeau or @f.uhlig could one of you please look into this? Here, one of the tests is still failing, and I can't figure out how to solve this problem. Thanks in advance.

The problem is that your changes somehow changed the pull in time for TRD, which now fails the QA plots.

[INFO] Checking the task CbmCaInputQaTrd
[INFO] -- Hit efficiency integrated over hit distance from station center
[INFO] 
Table: Efficiency table
                               Station ID           Efficiency           
                               0.000                0.638                
                               1.000                0.883                
                               2.000                0.878                

[INFO] 
Table: Residual mean values in different stations
                               Station ID           Residual(x) [cm]     Residual(y) [cm]     Residual(t) [ns]     
                               0.00000000           0.40327491           1.09204806           29.22268691          
                               1.00000000           0.04773213           4.07323893           9.66747556           
                               2.00000000           4.51792189           4.45632399           9.47015246           

[ERROR] CbmCaInputQaTrd: Mean for Pulls for Time in TRD station 0 is found to be under the lower limit (-3.35006 +- 3.5 x 0.240549 < -0.3)
[ERROR] CbmCaInputQaTrd: Mean for Pulls for Time in TRD station 1 is found to be above the upper limit (1.93323 +-3.5 x 0.0403162 > 0.3)
[ERROR] CbmCaInputQaTrd: Mean for Pulls for Time in TRD station 2 is found to be above the upper limit (1.94694 +-3.5 x 0.043452 > 0.3)

I am not 100% sure of the best way to solve the underlying data changes (maybe cross-check that errors added to hits fit and lead to pull with right mean and width)

The short term solution would be to tune the "allowed pull mean deviation" values, similar to what I did in 67920dcc to solve the last occurrence of this error.
I think this is what you should try for this MR, except if you have a good guess as to why the mean changed so much (could also be due to the low statistics we use for the CI tests)

@p.-a.loizeau thanks for the quick reply.
After raising the mentioned value from 0.3 to 2.5, the Time-Pull is accepted. But now, the Y-Pull in Station 2 is not accepted anymore, which does not make any sense to me, since it was okay before.

[ERROR] CbmCaInputQaTrd: Mean for Pulls for Y in TRD station 2 is found to be above the upper limit (6.50244 +-3.5 x 0 > 2.5)

The test result seems to be a bit unstable. Do you have any idea what else I could do?

@s.zharko Could you help the TRD team here? probably it will be faster with you directly looking at what they do than me trying to do it

@d.spicker the pull is calculated using the MC-truth as follows: pullY = (hitY - pointY) / hitDy, where hitY is a hit y-coordinate, pointY is a matched MC-point y-coordinate shifted to z-coordinate of the hit, and hitDy is an uncertainty of the hit y-coordinate.

Such a big mean of the y-pull distribution signals either about a big shift of the hit y-coordinate or about a too small y-coordinate uncertainty, so I would recommend to tune theses two quantities more accurately with respect to the MC-information

Thank you, @s.zharko. I have now tuned the uncertainties and in most apptainers the test passed, but not in all of them. Should I still try to further change the values or could this have other reasons?

Both of them (https://cdash.gsi.de/testDetails.php?test=19638738&build=447214 and https://cdash.gsi.de/testDetails.php?test=19638911&build=447215) have the following line linked to the failed check and which seems suspicious to me:

[ERROR] CbmCaInputQaTrd: Mean for Pulls for Y in TRD station 2 is found to be above the upper limit (5.00188 +-3.5 x 0 > 2)

Here it would seem that the allowed width is 0 if I am not mistaken? so maybe some missing value in the macro or some conversion/precision error?

@f.uhlig these are both CentOs runners, could it be that there is some types conversion should be more carefully handled in this OS when providing the values from the macro to the lower layer instance?