Pixel positions for Astri are wrong in h5 files about dl1-data-handler HOT 4 CLOSED

I will try to look into the current version of ctapipe and see if this has been corrected and if not, maybe we should post an issue there, as this isn't something we can directly control. The pixel positions are written here:

https://github.com/cta-observatory/image-extractor/blob/efe92aedc2ea422c4e3e1985061f7666c9f529e5/image-extractor/image_extractor.py#L275-L279

They're extracted straight from the event's instrument container without any modification (besides padding). We can manually try to correct it within image-extractor but this seems like it isn't the right way to handle it.

from dl1-data-handler.

I'm currently reading the pixel positions from the fits file from https://github.com/cta-observatory/ctapipe-extra, which are identical to the ones in the h5 files. The camera is filled row by row from bottom left to the top right:

Why it seems that the ASTRI camera should be filled from top left to bottom right by 8x8 sub-module sectors? I searched for a gamma shower that triggered ASTRI, CHEC and DigiCam at the same location. So I plotted the same event at the same location with different SST's and the shower orientation is the same. Due to the larger FoV of ASTRI, the shower of ASTRI is closer to the center of the camera than CHEC and DigiCam.

from dl1-data-handler.

I think you are confusing the real (physical) Astri camera, and the simulated one. What comes out in these files is read from the Monte-Carlo simulations, which have in general different pixel ordering for all cameras. Since the ordering doesn't matter except for calibration, this should be fine for most studies.

If we load real data, of course we would need to load the true pixel positions , which would have the physical ordering and are hopefully stored in any raw data file in the proper order. I'm not sure if there is any control over the ordering in the Monte-Carlo, but perhaps it's possible in the configuration to have them come out similar to the real data? That would be part of the MC validation, I guess. Nevertheless, if you change the ordering, you would have to re-produce all monte-carlo files, so if it's critical, it could be considered for prod4+

from dl1-data-handler.

As far as I can tell this will not require any specific action for the foreseeable future, so I'll close it. We can reopen a similar issue when the question of using real data becomes more defined.

from dl1-data-handler.