What does Gamma setting do?

[tooth_dr]

Im going to try imaging Jupiter tonight, using a DMK21mono or a Skyris618c. I normally use FireCapture any time I’ve tried either camera 

Is there a starting point for the settings to get me close?

Firstly what does Gamma actually do and what should it be defaulting to?

I always aim for a low gain and increase exposure time to give approx 50% histogram. Otherwise it looks too noisy when gain is higher.

Im beginning to think this is not the correct approach.

Is there an ‘optimal’ exposure speed/time to aim for instead and then move the gain to give the appropriate histogram?

I plan to use a 12” F5 scope, with probably either a 2x or 3x Barlow to suit the conditions.

TIA

Adam. 

[vlaiv]

Gamma:

It is non linear transform of the signal and should be avoided in astronomical use. Actually I can only think of one particular case where it would be useful - and that is only if gamma is implemented at analog stage (In most cameras I think it is in digital stage so not much use).

I can probably best explain it like this: Take a number range: 0 - 100 and do linear transform with it - let's say divide it with 2. Now observe sub ranges 0-10, 10-20, 20-30, ... after operation they have "equal spacing" in resulting set 0-5, 5-10, ....

Gamma is non linear transform that operates as power, so look at same range: 0-100 and use square root transform (non linear) it will "compress" range into 0-10 range, but there is another thing that happened. If you again observe sub ranges, in this case 0-10 and 10-90 and note that second range is x9 as large as first - in resulting output it will be 0-3.1622, 3.1622-10 - so first range is 3.16 long while second is 6.83 long - or about x2.16. So we ended up "compressing" x9 long original range into one that is only x2.16 as long as first range.

Gamma is used to "compress" certain range to smaller - so you can compress low numbers, or compress high numbers (depending on value of gamma). This can be used in signal compression when you don't care about certain range - too dim or too bright. It is also noted that human eye has response that interprets Gamma 2.2 as linear - or if we observe linearly bright sources - one source as four times as bright as another - our eye will not perceive it as being 4 times as bright, but if we apply gamma 2.2 - then our brain will register it as being x4 times as bright.

Only place that I think this might be useful is to "compress" high full well capacity into smaller number of bits - but it needs to be applied in analog stage, and after digitizing and before processing one should "revert" gamma (with opposite setting) to recover linear data. Gamma factor should be chosen such that it compresses high numbers while leaving low numbers as "normal" as possible - because high signal values have high SNR and gamma with successive rounding (ADC) will introduce errors - larger errors in more compressed region - so if we compress high values, since they have already high SNR - it will not hurt as much.

I don't think our current cameras even have analog stage gamma so that parameter is best left at 1 (or no gamma applied).

Parameters for lucky imaging:

I think you should base your parameters on coherence time. With seeing, depending on different layers of turbulent air having different speeds that they move with and aperture of telescope, there is something called coherence time. Different temperatures in different layers will bend the light, but if that bending is "constant" - you will not get "motion" blur, only distortion. It is important to capture frames that "freezes" the seeing - or use exposure lengths shorter or equal to coherence time. This will let stacking program select good frames (the least distorted) without impact of motion blur. Usual coherence time is in range of 5 to 15 (or even 20, 30 when atmosphere is stable) milliseconds. So aim for those exposure lengths.

After you select your exposure length - choose your gain based on two things - first is read noise - examine gain/read noise graph and see which gain settings offers you the least read noise. And second criteria is avoid clipping / over exposure, so even though there might be really high gain setting with very low read noise - don't use it if it clips your target. You should aim your histogram up to 90% or there about - don't go over that on average - because in moments of excellent seeing it can jump over 100% and clip (distortion spreads light around and so on average it might seem that target is in good histogram place - but then you will run into moments of great seeing and image will be really sharp and bright in places and it might clip).

Don't worry if your selected exposure is such that histogram is far to the left - like 10% or so - in lucky imaging you are not trying to go for optimum exposure in terms of collected light - you aim for optimum exposure to freeze the seeing. After stacking you can brighten the image in processing if it is too dim in subs.

[Stub Mandrel]

@vlaiv have you ever thought of writing a book or making a website that explains all these things and more?

[tooth_dr]

A real credit to the site @vlaiv ??

[vlaiv]

9 minutes ago, Stub Mandrel said:

@vlaiv have you ever thought of writing a book or making a website that explains all these things and more?

Actually I do plan a website / blog kind of thing that I would use to explore (write about) many different topics, most related to programming, astronomy and physics. I've got server and domain setup (well, couple of those ). Only problem is that I wanted to program it rather than use "out of the box" solution like wordpress, joomla (is that thing still alive?), drupal or similar. Up to beginning of this year I had two jobs, so I'm kind of burned out at the moment (I had to quit one of them because it was putting too much mental strain on me). So yes, there is website in pipeline along with couple of software related to astronomy (that will hopefully see a light of a day sometime - or as soon as get back into saddle and dedicate some time to those).

[newbie alert]

er you select your exposure length - choose your gain based on two things - first is read noise - examine gain/read noise graph and see which gain settings offers you the least read noise. And second criteria is avoid clipping / over exposure, so even though there might be really high gain setting with very low read noise - don't use it if it clips your target. You should aim your histogram up to 90% 

Gain ,read noise graph?? Where would you find this Vlaiv.. For planets I have my histogram 50-70% . For solar I have it at around 80%

[vlaiv]

10 minutes ago, newbie alert said:

er you select your exposure length - choose your gain based on two things - first is read noise - examine gain/read noise graph and see which gain settings offers you the least read noise. And second criteria is avoid clipping / over exposure, so even though there might be really high gain setting with very low read noise - don't use it if it clips your target. You should aim your histogram up to 90% 

Gain ,read noise graph?? Where would you find this Vlaiv.. For planets I have my histogram 50-70% . For solar I have it at around 80%

I often see it published for CMOS cameras like this:

Look at last graph - it shows read noise vs different gain settings.

Alternatively you can create such graph your self (if you know e/ADU value for each gain setting) - just shoot 2 bias files at each gain setting - subtract them, convert to e from ADU (by using e/ADU value for particular gain) and measure standard deviation of pixels, which you then divide by square root of 2 - that will give you read noise value.

[newbie alert]

10 hours ago, vlaiv said:

I often see it published for CMOS cameras like this:

Look at last graph - it shows read noise vs different gain settings.

Alternatively you can create such graph your self (if you know e/ADU value for each gain setting) - just shoot 2 bias files at each gain setting - subtract them, convert to e from ADU (by using e/ADU value for particular gain) and measure standard deviation of pixels, which you then divide by square root of 2 - that will give you read noise value.

Ha ha,simple then!

[Demonperformer]

Just leaves the question of how do you get the e/ADU figure for each gain setting ...

[vlaiv]

2 hours ago, Demonperformer said:

Just leaves the question of how do you get the e/ADU figure for each gain setting ...

You can measure that as well. Measurement process is a bit different - you need a set of flats to do that. I can't remember exact procedure, but lets derive it.

Suppose that we have some conversion coefficient (e/ADU) and we take a flat and measure mean ADU value. So our mean value will be e_count / eADU.

If we subtract two flats - we should be able to measure shot noise (as standard deviation of difference) and that will be e_count^0.5 / eADU - shot noise will be square root of number of photons and it will be then converted to ADU units.

If we take square root of mean ADU value and divide it by measured standard deviation we will get (e_count / eADU)^0.5 / (e_count^0.5 / eADU) = eADU^0.5  (after cancelling terms).

So we just need to raise it to power of two to get eADU.

This above calculation is assuming that only source of noise is poisson shot noise - not true, we need to remove read noise as well (if we use very short exposures and dark current noise is not an issue).

Ok, so procedure would be like this:

Take 2 flats and 2 bias files. From bias files as described above get read noise (in ADU units). Subtract 2 flat files and measure stdev. Find noise term as noise = sqrt(stdev^2 - read_noise^2)

Take mean value of one of the flats. Divide with noise squared. This should get you ADU value.

For best results - take multiple measurements (multiple sets of flats and bias files, you can take flats at different exposure lengths to cover range on histogram) and calculate mean value of all measured ADUs, you can calculate stdev of result set. This way you will get e/ADU value and its stdev (so you will know how precise you measured it).