EPICAL-2 Analysis meeting
OL 260
UU
# Epical-2 - 20 Jan 2020 NKW notes
# Event Selection
1) There are some ambiguities in the procedures as described in the weekly meeting presentations.
To clarify, these need to be written in more detail, as was done for the clustering algorithms.
The performance paper on Overleaf is the most useful place to include these descriptions now.
**Action (1)** Robbie: to write into the Overleaf performance paper, both Hiroki and his algorithms;
**Action (2)** Hiroki: to check that that Robbie's description is accurate
2) The vent selection code that Fabian tested (first version produced by Robbie, which includes both Hiroki's fastjet-based selection and Robbie's) is slow. This makes it impractical to run during an analysis, as it can take several hours to run the selection on a single run, far longer than carrying out analysis after events are selected.
Conclusion is that a loose 'preselection' that would always need to be run for all analyses, should be run on all data and new .root files produced. These would have fewer events and add to each event a unique event_number, given by the sequence number from the current merged .root files. We consider the sequence number in the current .root files as our event number. In the new 'preselected' .root files, event_number would no longer be sequential.
In running preselection, also extend the data in .root files to include definition of _cluster objects_ (list of pixels that define to a given cluster); could include here cluster objects as defined by both Hiroki's topological clusterinbg and Quasim's faster/readout-oriented definitions of cluster objects, for later comparison.
**Actions (3)**: Hiroki, agree with Fabian/Robbie/Quasim/others, the implementation of adding event number and cluster objects to the preselected .root files;
**Actions (4)**: Hiroki (+Robbie), carry out preselection of events to produce the new, preselected .root files. Add new .root files to eos/cernbox.
**Action (5)**: Robbie, investigate speed of selection code, see if any quick gains by a brief look at profiling (Nigel to discuss offline few ways of doing this)
If performing the event selection and storing results of the selection (decision) in separate, standalone .root file, that is (optionally) read in by the event selection in subsequent use rather than repeating the slow event selection each time, then using the unique run and event_number will avoid potential for associating decision with the wrong event.
# Simulation tuning
Tim showed very thorough studies of the paremeter dependence in AllPix2 of resolution/mean values and also time taken.
Proposal to tune parameters using 5 GeV to retain independent predictive power for other energies.
Recommendations for beam profile parameters as input to the simulation, is to take a 2d PDF of the beam profile as defined by the pixel density on layer 0, separately for each beam energy studied in the range 1...5.8 GeV.
* The list of good runs is available on: [cernbox/eos](https://cernbox.cern.ch/index.php/s/KM1vghn5FCeYgXX)
* **Action (6)**: Fabian to produce hitmaps that Tim can use in AllPix2
* These should have the loose selection applied (no point in simulating events that we will not use - but suggest extending PDF from beam data by small amount in row/column regions, to allow for some smearing/resolution/alignment
* Initially, not important that the selection used to define the PDF is not final, main aim is to show this works technically with AllPix2.
* Hitmap can be rebinned to ensure no empty bins before use in AllPix2, as Tim noted, any gaps would lead to no particles generated in a given bin
* May have to be parametrised rather than binned, but for now binned PDF is good.
* There will be correlation between position and beam momentum as selected by a dipole magnet, effects probably small for our purposes, but to remember for future use, just in case, see [arXiv:1807.09328](https://arxiv.org/abs/1807.09328) and [https://doi.org/10.1016/j.nima.2018.11.133](https://doi.org/10.1016/j.nima.2018.11.133)
TIPP conference: [Abstract](https://tipp2021.triumf.ca/abstracts.html) deadline 7 Feb.; we should aim to present results, on some of (at least)
* simulation of a Epical-2 using AllPix2
* initial test beam performance of a digital electromagnetic calorimeter using ALPIDE sensors
* experience of ...