Short exposures versus narrowband filters

[Ags]

Assuming that gradients can be controlled, do narrowband filter offer any benefit with short (1 to 15 second) DSO exposures? I can understand NB filters enabling long exposures by preventing the image from saturating due to sky fog - but this won't happen with short exposures and the sky fog that is captured can be zeroed out in processing.

Edited March 13, 2020 by Ags

[vlaiv]

Yes.

Primary role of NB filters is to cut sky background and associated noise. Just reducing background levels is really not important on its own - you can control that in processing by setting black point. It is associated noise that is problematic - stronger the signal - stronger the shot noise associated with it. Sky background does not contain important signal - it just contains constant signal and that is not useful - but noise associated with it is bad for image.

Short vs long exposure - if read noise of camera is dominant noise source in single exposure - you will benefit from doing longer exposures (until read noise is not strongest per single sub). Reducing noise from LP/Sky background makes read noise dominant - and that is why it is better to do longer vs shorter exposures.

If you limit yourself to having only short exposure - then NB filter vs without filter is again better - it will remove sky background associated noise, so in this case - just going with short exposures regardless - NB filters help if target emits in NB lines of the filter of course.

[Ags]

Ah, thanks to you I have learned something fundamental. The sky fog photons scatter randomly onto the sensor pixels, and the brighter the sky fog the more similar the totals on each pixel is, so I would be able to take a very good smooth picture of the sky fog... But the absolute difference between the totals on each pixel increases (at the square root of the total sky fog signal) so that when I "zero out" the sky fog in post processing, I only subtract the baseline from each pixel, leaving 100% of the noise component. If I was using an NB filter, the sky signal would be lower so the noise from random photon scatter would be reduced in absolute terms.

Edited March 13, 2020 by Ags

[vlaiv]

2 hours ago, Ags said:

Ah, thanks to you I have learned something fundamental. The sky fog photons scatter randomly onto the sensor pixels, and the brighter the sky fog the more similar the totals on each pixel is, so I would be able to take a very good smooth picture of the sky fog... But the absolute difference between the totals on each pixel increases (at the square root of the total sky fog signal) so that when I "zero out" the sky fog in post processing, I only subtract the baseline from each pixel, leaving 100% of the noise component. If I was using an NB filter, the sky signal would be lower so the noise from random photon scatter would be reduced in absolute terms.

Indeed - same thing applies to other "signal" sources - target signal (good signal) and thermal signal (like sky signal - bad signal as you remove it by calibration but random component remains).

All of these are modeled like Poisson process.