Add dual-readout tubes barrel calorimeter in IDEA_o2 #413
Conversation
|
Hi @lopezzot Thank you so much for your contribution. While running some events, I got the following error, could you please crosscheck? |
lopezzot
force-pushed
the
lp-ideao2drbarrel
branch
from
064d386 to
432dfa4
Compare
Hi @atolosadelgado, thanks for spotting this. I did not test it before with an isotropic source so I could not find this. It was an issue with hits at the border between the barrel and the endcap calo. The last commit fixes it. I also rebased. Let me know if you have other tests in mind we can run. (Tagging @s6anloes for his information.) |
|
Hi @lopezzot Thanks for the quick fix. I have run several simulations and Geant4 does not seem to complain. I have not inspected the output, I trust you. I would advise to remove the test named I noticed that the memory consumption increases over time during the simulation, but this may be related to other issues (hopefully not a memory leak). For reference, when running few events (<100) is 3.8GB (RSS) or 4GB (VSZ), but it goes up to 5.5GB (RSS) when running more events (~2000) The average time per event, for when using particle gun of 1 electron at 10 GeV, is 1 second (seems to grow linearly with the number of events). The output file looks reasonable (13MB for 1000 events). Otherwise, I do not see major problems for merging. |
Co-authored-by: Andreas Loeschcke Centeno <andreas.loeschcke.centeno@cern.ch>
Add dual-readout capillary-tubes barrel calorimeter inside the IDEA_o2 xml file. Change IDEA_o2 muon system dimensions to avoid overlaps with the barrel calorimeter (new params checked with Mahmoud Ali).
Move the detector dimensions parameters inside DectDimensions_IDEA_o2_v01.xml.
Move the vis attributes for DRBarrelTubes subdetector to DectDimensions_IDEA_o2_v01.xml.
Increase DRTubesGranularity to 72 rotations aroung the beam pipe (as the endcap).
Make explicit printout from DRTubesconstructor.
Rename DDDRCaloTubes subdetector for consistency with k4geo standards.
Displace the barrel inner radius by 0.5 cm to avoid overlaps between the calo barrel staves and the solenoid.
Materials names defined in DDDRCaloTubes_o1_v01.xml where already defined in the endcap xml file, therefore the barrel geometry was using the endcap materials. This is a problem in case one wants to activate/inactivate optical properties for such materials in the endcap or barrel separately. Fixed by changing names of DDDRCaloTubes materials. Also, set particle direction to (0,1,0) in the IDEA_o2_vo1.py file to shoot into the barrel instead of the endcap.
Set the subdetector ID in DectDimensions_IDEA_o2_v01.xml file and assign.
Add description of the barrel dual-readout-tubes caloriemter in detectors/calorimeter README.md file.
By adding inclusion of the dual-readout-tubes barrel calorimeter in Decemeber 2024.
First attempt to produce signals from the barrel dual-readout-tubes calorimeter.
The theta coordinate was taken from the track momentum direction, instead is should be taken from the G4Step position to discriminate if a G4Step is inside the barrel or the endcap geometry.
Change the name of hit collections to avoid duplication of hits while using ROName in the hit collection name. The barrel was creating also hit collection for cher_right/left and scin_right/left which are only for the endcap.
Reinclude optical properties for fiber materials. Create PMMA_Scin material to differentiate between material used in Scin and Cher fibers in order to not have optical photons in Scin fibers but only in Cher fibers.
Remove printout and apply clang-format. I checked that this version of the custom SD action is consistent with the speedup obtained originally with the endcap geometry only (about 1 s/evt for 10 GeV e- events).
Fix printout to check that volIDs of the DRBarrelTubes are correctly reproduced.
Commet out volID printout check for barrel calo in DRTubesSDAction. This was used to check that the 64-bits volIDs recreated from the barrel copynumbers are identical to the DD4hep ones. However marking barrel fibers as sensitive makes the memory footprint explode (34 Gb before it is killed on lxplus). So the test was done on a single stave.
Change logic to differentiate between barrel and endcap calo in DRTubesSDAction. Instead of using theta from global coordinates use copy numbers.
lopezzot
force-pushed
the
lp-ideao2drbarrel
branch
from
23e7df9 to
c63359b
Compare
|
Hi again,
Done and rebased.
I just did a run with 5k events and observed that the memory (RES) taken was 5.5 Gb already at the very first events, and it was stable for the whole run. However, I also observed 3.8 Gb when I opened this PR in December. I do not know what is causing this.
I got slightly better numbers averaged over a run with 5k events: 0.69 s/evt. Would be nice to understand what is causing this shift. Anyhow, both numbers are good.
Thanks! should we proceed or you want more tests? |
|
I have run some more events (10k, proton at 10GeV), and I agree with @lopezzot results. The memory slowly increases, and there are jumps, but only up to 5.6GB.
I do not see any error/warning or exception after running with different particles/energies, so I merge, thank you! |
|
I'm also seeing the same, testing with 5k electron events. |
|
Thank you for your tests and for merging. :) |
BEGINRELEASENOTES
ENDRELEASENOTES
Hello,
this PR includes a new subdetector, the IDEA dual-readout capillary-tube barrel calorimeter. Together with the endcap calorimeter, it completes the hadronic calorimeter of IDEA_o2. The muon system dimensions have been adjusted to avoid overlaps with this subdetector. The only missing part for IDEA_o2 full-sim is now the crystal em calorimeter (#406 ).
DRTubesSDActionis also expanded to produce hits from the barrel calorimeter.Testing:
Authors:
Geometry by Andreas Loeschcke Centeno, custom sensitive detector action, integration in IDEA_o2 and testing by @lopezzot
FYI @mahmoudali2