"Blotchy" Noise With Gain 0 - Advice?

[The Lazy Astronomer]

Hi all, I was doing a test the other night using gain 0 with the ZWO 294MM. I shot 30 mins each using gain 120 (unity) and gain 0 and quickly stacked today and noticed the noise in the gain 0 stack has a 'blotchiness' to it.

I've attached a side-by-side of both stacks below at 100% zoom - any suggestions as to why this may be?

Edit: gain 120 is on the left, gain 0 on the right

Edit 2: I should also mention both are light frame stacks only, no calibration, and both are displayed with a standard STF

Edited February 12, 2023 by The Lazy Astronomer

[The Lazy Astronomer]

Might be easier to see with boosted STF

[ONIKKINEN]

Judging from the graphs:

You shot at 8e- read noise compared to the usual 2e-. 4x the read noise means you need to expose 4^2=16x longer to swamp the read noise the same way as with the higher gain that has 2e- read noise. Did you expose for 16x longer? Doubt it, so it looks more noisy because it just is more noisy.

Edited February 12, 2023 by ONIKKINEN

[vlaiv]

3 hours ago, ONIKKINEN said:

You shot at 8e- read noise compared to the usual 2e-. 4x the read noise means you need to expose 4^2=16x longer to swamp the read noise the same way as with the higher gain that has 2e- read noise. Did you expose for 16x longer? Doubt it, so it looks more noisy because it just is more noisy.

Not only that - there is also matter of quantization error / noise - which does not behave like regular noise.

With e/ADU value of about 4 - there is two bits of rounding error.

Say that you have 15e pixel value - it will be recorded as 3 ADU - and when you convert it back to electrons - it will be 3ADU * 4 e/ADU = 12e instead of 15e.

Any value in electrons will be divided with 4 and rounded to lower integer (or nearest - but effect is the same - 3 out of 4 numbers will have error).

This quantization error looks like "ladder" - or steps.

12e, 13e, 14e and 15e are all recorded as 3ADU or rounded to 12e.

High read noise is attempt to mask this - it is called noise shaping:

https://en.wikipedia.org/wiki/Noise_shaping

It does help - but when you stack many images that have same noise pattern - read noise goes down, but this type of "predictable" noise is not reduced equally as truly random noise and starts surfacing.

It is most visible in areas of the image where there is no signal (shot noise also helps to mask it).

[The Lazy Astronomer]

Thanks both. I did use longer exposures, but nowhere near 16x longer, so think I was expecting a bit more noise, but the non-random pattern of the noise confused me. If I'm understanding you correctly vlaiv, then what you've said explains it.