search

cta-observatory / cta-lstchain

LST prototype testbench chain
https://cta-observatory.github.io/cta-lstchain/
BSD 3-Clause "New" or "Revised" License
23 stars 77 forks source link

Issue with the MC info and the reconstruction. #135

Closed vuillaut closed 5 years ago

vuillaut commented 5 years ago

Hi

These are two related issues (discovered them at the same time) that actually compensate each other.

  1. The altitude info is not correctly set in the MC production for point source gammas. I bet it is actually the pointing information that is written.

    filename = '/fefs/aswg/workspace/MC_common/corsika6.9_simtelarray_2018-11-07/LST4_monotrigger/prod3/gamma/gamma_offaxis0.4deg_20190415/South_pointing/Data/gamma_20deg_180deg_run100___cta-prod3-demo-2147m-LaPalma-baseline-mono_off0.4.simtel.gz'
import eventio
file = eventio.SimTelFile(filename)
for event in simtel.iter_mc_events():
print(np.rad2deg(event['mc_shower']['altitude'] ))

    returns 69.99999 for every event, while one should have 70.4. For a diffuse file, the writing seems right, different for every event. Note that the point source simulated images are right though, pointing to the offset of 0.4deg. Applying Hillas to a couple of them and plotting the directions, here is what I get: image

  2. Training on diffuse gammas and reconstructing point source ones, the mean reconstructed value by lstchain however is 70.0 (while it should be 70.4 as the images do point to this direction). The current results therefore have a bias of 0.4deg that was unseen until now because we compare the reconstructed value to the simulated one which is wrong.

I am currently investigating issue 2. And will contact Yoshiki concerning issue 1.

This issue might be part of the 'too good sensitivity' explanation. This with the filtering on LST id (see comment in #133)

vuillaut commented 5 years ago

Self-correction:

It's actually the pointing that has an offset, my bad.