Noise control and Unity gain with ZWO cams

[Astrokev]

My simplistic understanding is that unity gain is the value which gives a 1:1 conversion of electron count to ADU.

My newbie question is - is this the value that theoretically also gives the lowest noise in the image?

For imaging of bright planets, where there is sometimes enough light to be able to use a gain setting lower than unity (depending on f ratio being used!), would this give lower noise?

Therefore, in principle, does setting gain as low as possible give best control of noise?

A few nights ago I started an experiment with my ASI174MM on Mars to put this to the test, but haven't completed this yet, and with the other variables involved (mainly fluctuations in seeing) I'm not sure if this will give a clear answer.

Any thoughts on this gratefully received.

[vlaiv]

No.

I mean it does not give lowest noise in the image.

There is relationship between gain and noise, but it's not simple one. Gain does not directly affect shot, LP and dark noise. It does however directly affect read noise. For this you have a graph for every camera. In case of ASI174 it is this:

In principle, read noise decreases with increased gain. Explanation for this is: read noise happens in two stages - pre amp and post amp. Pre amp stage obviously does not depend on gain, but post amp does. This part of noise is in ADUs rather than electrons and is sort of "fixed" - but since gain is changing in such way that single electron is more and more ADU units - opposite is true as well. Fixed ADU value is less and less electrons as you increase gain. For this reason, post amp read noise drops in electrons as you increase gain.

Two things happen because of this - read noise drops with gain and there is a place where it "levels out" - post amp becomes insignificant in comparison to pre amp.

Read noise is not very important in long exposure imaging. It's only impact is on choice of sub duration. If you have two otherwise identical cameras - but one with higher read noise then the other - you can still produce images with exactly the same SNR with them - just choose appropriately longer subs with camera that has higher read noise.

With planetary imaging this has important consequence - camera or setting with lower read noise is better. Because we stack tens of thousands of frames and our subs are exceptionally short - like a few milliseconds, and because each of them carries one "dose" of read noise - lower read noise is better. You can't do the same trick as above - use longer exposures with higher read noise camera as seeing gets in the way and you don't have lucky imaging any more.

There is another aspect of gain. Gain is conversion factor and resulting values are always integer values / whole numbers.

Imagine you have 2 e/ADU gain, and you collected 9 electrons. Resulting ADU value should be 4.5 but you can only write 4 or 5 - not a number in between - so you round it to say 4.

Now if you say someone I captured 4 ADU with 2 e/ADU and ask them to calculate how much electrons you captured - they will tell you 8e. By above rounding you introduced 1e of error due to rounding.

Problem with this noise is that is is not good noise - it's not random, it depends on numbers involved and as such it does not "stack" properly. In practice, this is masked by read noise and you get "pseudo stackable" noise (which means for all intents and purposes it stacks but it does not behave 100% like pure noise). For this reason I believe if you want the best performance, you should stick to gain numbers that give you the least rounding errors.

For deep sky imaging / long exposure - this is unity gain, but for planetary imaging it is low read noise that gives you low rounding error - one that produces reciprocal of whole e/ADU (ADU/e is whole number). So you want gain to be 1, 1/2, 1/3, 1/4 and so on ... whichever gives you best read noise.

With ASI, gain scale is in 0.1dB units (so 1/10 of a Decibel) which means it is well defined in logarithmic scale. From above graph we can see the dip around 300 - this is the best place to be with respect to read noise, but which one exactly?

Unity gain for ASI174 is 189, and we have 20*log(ratio) formula, and we want ratio to be whole number. For example, for number 4 we have ~12.04dB (20 * log(4)), or if we convert that to 0.1dB units - that is 120.4. We need 189+120.4 = 309.4 gain to have e/ADU value of 0.25 or 1/4.

you can see from above chart that gain about 300 gives about 0.25 e/ADU value which in turns gives close to integer ADU/e value (in this case 4 - like we calculated above).

If you want best results noise wise for ASI174, I suggest that you go with gain 309/310.

 

[Astrokev]

3 hours ago, vlaiv said:

Wow, thanks Vlaiv for the detailed explanation. I understand bits of your reply, but will have to read it over (and get my calculator out!) to fully get my head around it.

The bottom line seems clear - best gain for lucky imaging is around 310. This gives me something to experiment with next time I'm out, and see if this produces a visible difference in my images.

Thanks again - great stuff!

Kev

 

 

[inFINNity Deck]

I know this response is a bit late, but for those interested in these figures: today I received an e-mail from ZWO stating that unity gain for the ASI174 is 179, not 189 as written in their graph (the black arrow is correctly placed at 179). When opening the ASCOM driver unity gain is correctly given as 179.

Nicolàs