calculate_noise_parameters

[Elisa-Visentin]

Hi, maybe this is more a discussion than an issue...
In the calculate_noise_parameters function you default extra_noise_dim to 0 if you get a value < 0

cta-lstchain/lstchain/image/modifier.py

Line 406 in 378a4c3

extra_noise_in_dim_pixels = max(0., added_noise)

In case you apply it only once to get the NSB level for a run, I agree that you cannot return a negative value (or you can return it but then you cannot remove noise from MC so you will have to put it to 0 later on). But in case you apply this function N times on the same run (i.e., using different subruns), if the average value is slightly above 0 you 'cut' one of the tails of the distribution, thus introducing a bias.
What about adding an option (e.g., a boolean variable) to the function to allow both for the default behaviour (i.e., default to 0) and for a more 'custom' behaviour (i.e., return also negative values)?

[moralejo]

Good point, although in practice if one gets values around 0 probably it is not worth to tune the NSB in the first place - I mean it will be a small tuning anyway, and the effect (compared to using dark MC) will be small.
Rather than adding an option on what to return I think you can add an argument (with 0 as default) which is the minimum allowed value. You then set it to -100 or whatever if you want to get the negative values. The important thing is that the default behavior, without that extra argument, stays the same.

[Elisa-Visentin]

Well, it depends on the 'granularity' of your tuned MCs (and, so, on the requested accuracy for the analysis). And on the 'std deviation' of the NSB values evaluated for a single run
Of course the default behavior must not change :)