This important pull request implements important changes in papas to the tuning of the CMS detector model, and a few changes to the tuning of the CLIC-FCCee detector model. As a result the tuning of CMS has been much improved, and the CLIC-FCCee model is now ready to be tried. In addition a few bugs were fixed in heppy.
papas changes
Changes to the CMS model
If you use the CMS model in papas, I strongly suggest you to reproduce your samples with this version of heppy. Please don't hesitate to report any difference you may observe with the full sim.
HCAL tuning
One year ago a wrong tuning of the HCAL of CMS had been introduced. This tuning implemented two things:
the HCAL resolution parameters were wrong. I had noticed this bug earlier, and already fixed these parameters to their well-known values in a previous PR, thereby obtaining a great improvement of the agreement of CMS.
because of these wrong parameters, the contributor had to introduce an HCAL response function, that would kill the response of the HCAL for low energies (in addition to what the HCAL threshold parameters are doing), and would converge to a value larger than 1 for high energies. Here, I simply bring back the response to 1.
The improvement brought by this change is visible when comparing analyses carried with papas CMS and the full simulation, and in the reconstruction of ee->Z->ddbar.
If you use the CMS model, I strongly encourage you to reproduce your samples with this version
electron and muon tuning
Previously, the electron and muon efficiency parametrization was only accounting for the pT and eta acceptance. In the acceptance, the efficiency had been set to 100% so that the user could apply her own efficiency model later on. The muon resolution was more or less good, and the electron resolution had been set to a rough calorimeter-like parametrization.
In this pull request, we apply the same e and mu parametrizations as in Delphes, which can be found here. There is however one small difference; while Delphes computes the electron resolution as a quadratic sum of the tracker and ECAL resolutions, we take a simplistic 3% value, which is roughly what can be found in the CMS electron performance paper, Fig. 12a. This simple tuning gives good results for the electron channels that have been tested for FCC-ee Higgs analyses.
Changes to the CLIC-FCCee model
A bug was fixed in the CLIC-FCCee tracker model. This bug was leading to a low tracking efficiency.
We now use the efficiency measured in the full sim. Note in particular that the acceptance in theta has been reduced w/r to what can be found in the CLIC TDR
The CLIC-FCCee model is now ready to be tried, even though small changes are expected (see below)
heppy bug fixes
IMPORTANT: wrong configuration in heppy_batch
When running heppy_batch, the user configuration is correctly imported, so that settings done globally (not in the config object but in the configuration module itself), such as the following one, are accounted for properly:
from heppy.configuration import Collider
Collider.BEAMS = 'ee'
Collider.SQRTS = 91.
The main effect of this bug was that FCC ee people would have their particle-like objects sorted by pT when running with heppy_batch, and by E when running interactively. This bug was not affecting my FCC-ee analyses, but it could be possible to see the effect, depending on what you are doing. If you're an FCC hh user, it is highly probable that you were not affected at all.
meta information from heppy_hadd wrong
Recently, a feature had been added to check chunks and count the number of correct chunks in heppy_hadd. This number was written in a yaml file, only used by me so far. This number is now correct. If you are not yet a user of fcc_datasets, you were not affected.
Planned changes
The following changes are not included in this PR, but will come soon:
clic tuning:
need to use it in the context of a few analyses to make sure that everything is correct.
parametrize track resolution as measured in the full sim.
check again lepton efficiency and resolution
need to check that heppy_hadd correctly adds cut flow counters and to add a unit test for that (possible problem reported by @jndrf )
restructuration of the heppy package to make it possible to distribute it with pip
fcc-bot
commented
6 years ago
This important pull request implements important changes in papas to the tuning of the CMS detector model, and a few changes to the tuning of the CLIC-FCCee detector model. As a result the tuning of CMS has been much improved, and the CLIC-FCCee model is now ready to be tried. In addition a few bugs were fixed in heppy.
papas changes
Changes to the CMS model
If you use the CMS model in papas, I strongly suggest you to reproduce your samples with this version of heppy. Please don't hesitate to report any difference you may observe with the full sim.
HCAL tuning
One year ago a wrong tuning of the HCAL of CMS had been introduced. This tuning implemented two things:
The improvement brought by this change is visible when comparing analyses carried with papas CMS and the full simulation, and in the reconstruction of ee->Z->ddbar.
If you use the CMS model, I strongly encourage you to reproduce your samples with this version
electron and muon tuning
Previously, the electron and muon efficiency parametrization was only accounting for the pT and eta acceptance. In the acceptance, the efficiency had been set to 100% so that the user could apply her own efficiency model later on. The muon resolution was more or less good, and the electron resolution had been set to a rough calorimeter-like parametrization.
In this pull request, we apply the same e and mu parametrizations as in Delphes, which can be found here. There is however one small difference; while Delphes computes the electron resolution as a quadratic sum of the tracker and ECAL resolutions, we take a simplistic 3% value, which is roughly what can be found in the CMS electron performance paper, Fig. 12a. This simple tuning gives good results for the electron channels that have been tested for FCC-ee Higgs analyses.
Changes to the CLIC-FCCee model
A bug was fixed in the CLIC-FCCee tracker model. This bug was leading to a low tracking efficiency. We now use the efficiency measured in the full sim. Note in particular that the acceptance in theta has been reduced w/r to what can be found in the CLIC TDR
The CLIC-FCCee model is now ready to be tried, even though small changes are expected (see below)
heppy bug fixes
IMPORTANT: wrong configuration in heppy_batch
When running heppy_batch, the user configuration is correctly imported, so that settings done globally (not in the config object but in the configuration module itself), such as the following one, are accounted for properly:
The main effect of this bug was that FCC ee people would have their particle-like objects sorted by pT when running with heppy_batch, and by E when running interactively. This bug was not affecting my FCC-ee analyses, but it could be possible to see the effect, depending on what you are doing. If you're an FCC hh user, it is highly probable that you were not affected at all.
meta information from heppy_hadd wrong
Recently, a feature had been added to check chunks and count the number of correct chunks in heppy_hadd. This number was written in a yaml file, only used by me so far. This number is now correct. If you are not yet a user of fcc_datasets, you were not affected.
Planned changes
The following changes are not included in this PR, but will come soon: