Open HealthyPear opened 3 years ago
@GernotMaier here there should not be big differences, taking into account that the extractor that CTAMARS uses came from VERITAS.
To summarise
Any major differences wrt to what EventDisplay does?
Major difference is obviously the image cleaning methods applied. Eventdisplay uses the optimised next-neighbour cleaning (https://ui.adsabs.harvard.edu/abs/2013ICRC...33.3000S/abstract).
This is an issue which I dedicated to the calibration phase - do you mean that you use this cleaning also in the middle of the 2 image extraction passes?
trying to understand what 'in the middle of the two image extraction passes' means. You refer to the double pass cleaning from VERITAS? I don't use this for CTA but only do a single pass cleaning using Maxim's cleaning method.
Yes, I mean the cleaning that is used in the CTAMARS version between 1st pass and 2nd pass.
So you do not use a double-pass image extraction. Does the document you link above describe also image extraction or only the cleaning afterward? In the latter case, can you point me to the document that describes the image extraction procedure you use in EventDisplay? (perhaps the references from your repo?)
I thought image extraction == image cleaning...do you mean the trace integration step or what do you call image extraction?
I call image extraction the process that translates ADC counts (or anyway sample units of the waveform) to phe which results in calibrated images which will after get cleaned and parametrized to DL1
what is a bit confusing is that in the 2-pass image extraction of CTAMARS there is a preliminary image cleaning, but the whole process is image extraction
OK. So signal extraction is simply a sliding window searching for the maximum integrated ADC counts, followed by the image cleaning described in the document above.
How large is this window in sample units?
Depends on the telescope type and the typical width of the pulse shapes.
For prod5, this is LST: 2, MST-F: 4, MST-N 4, SST: 6
Note that I up-sample the FlashCam traces by a factor of 4 (using the code provided by the FlashCam team)
I never tested cameras outside LSTCam and NectarCam, so good to know!
Any chance you can pass me these same numbers for e.g. Prod3b baseline (North & South)?
I promise I'll start Prod5...
for prod3b:
This is sub-summary issue part of #85 .
Requirements
TBD
Reference documents
benchmarks/DL1/benchmarks_DL1_calibration.ipynb
)Known sources of possible divergence and current status: