ASI 1600 Gain Settings

[Rodd]

OK--everything is working properly for a change.  Thanks to all that helped me.  Now, its time to tackle gain settings.  I am currently running the ASI 1600mmcool pro on the TOA 130 at F7.7 (1000mm fl) for galaxy imaging.  I have been using unity gain (139).  As far as offset goes, I have no idea what that is and have ignored it--so I am using whatever is default for the camera. For cooling I am using -20.  I am taking 120 sec LRGB subs.  But I find there is allot of black chekerboard type pixels that really do not go away with stacking.  I have been stacking 90-120 subs for RGB and 120-150 subs for lum.   I have only used the camera 3 times and conditions were terrible--my M63 image is the sole project I have completed.  

So--I guess the question is:  What gain is best for gain 2min subs.  2min subs look pretty good as far as bightness and detail (there is no glaring areas of overexposure, and each sub shows about what I would expect in a 10 min KAF 8300 sub).  I know the answer will be it depends on my sky and optics, etc.  But there should be a general guideline.  Clear sky time is almost non existant these days, so a all night trial and error session is not desirable.  Any suggestions?

Thanks----Rodd

[kens]

As you know - it depends. The "sweet spot" for the ASI1600MM appears to be around gain 75 (or 76 depending which way your OCD tendencies lie). Lets say 70 to 80.

That's not one of the presets so you need to set an appropriate offset. I now use 50 for all gain as that seems to be the way ZWO is heading with their drivers. It is based on the offset  of their highest preset gain of 300 so can never be too low at lower gains. But it simplifies things not having to think about it and it has negligible impact on dynamic range. 

Will that give you 2 minutes subs? That's probably the wrong way to think about it. What you want to do, at any gain level, is to expose long enough for the sky background to "bury" the cameras read noise. Not an exact science but somewhere between 3 and 10 times the square of the read noise are touted. 

For instance, at gain 75 (about 2.1e-/ADU) the read noise is about 2.2e- so at offset 50, 3 times RN squared is 910 ADU and 10 times is 1140 ADU. So you want the peak of your histogram to be around those numbers. At that level the camera read noise is no longer dominant.

At gain 139 the read noise is 1.7e- and the corresponding 3xRN^2 and 10xRN^2 figures are 930 and 1240 at offset 50; or 470 and  780 at the default offset of 21.

So check the histogram peak on your subs. You'd expect to have much lower values with RGB than for L since only 1/3 the photons are reaching the sensor. If the peaks are good enough with 2 min subs all good and well. If they are too low then you either need to increase the gain (to lower read noise) or increase the exposures. You can calculate roughly how much but it will depend on the sky background on any night. If you don't want too many options you could calibrate it so the darkest nights are at least at the 3x level.

The ASI1600 (and its ilk) also exhibit fixed pattern noise and your chequerboard pattern may be that. Calibration and dithering can help there. You did not say if you are calibrating your subs at all. Note that with the ASI1600, flat darks are recommended over bias frames.

[Rodd]

9 minutes ago, kens said:

As you know - it depends. The "sweet spot" for the ASI1600MM appears to be around gain 75 (or 76 depending which way your OCD tendencies lie). Lets say 70 to 80.

That's not one of the presets so you need to set an appropriate offset. I now use 50 for all gain as that seems to be the way ZWO is heading with their drivers. It is based on the offset  of their highest preset gain of 300 so can never be too low at lower gains. But it simplifies things not having to think about it and it has negligible impact on dynamic range. 

Will that give you 2 minutes subs? That's probably the wrong way to think about it. What you want to do, at any gain level, is to expose long enough for the sky background to "bury" the cameras read noise. Not an exact science but somewhere between 3 and 10 times the square of the read noise are touted. 

For instance, at gain 75 (about 2.1e-/ADU) the read noise is about 2.2e- so at offset 50, 3 times RN squared is 910 ADU and 10 times is 1140 ADU. So you want the peak of your histogram to be around those numbers. At that level the camera read noise is no longer dominant.

At gain 139 the read noise is 1.7e- and the corresponding 3xRN^2 and 10xRN^2 figures are 930 and 1240 at offset 50; or 470 and  780 at the default offset of 21.

So check the histogram peak on your subs. You'd expect to have much lower values with RGB than for L since only 1/3 the photons are reaching the sensor. If the peaks are good enough with 2 min subs all good and well. If they are too low then you either need to increase the gain (to lower read noise) or increase the exposures. You can calculate roughly how much but it will depend on the sky background on any night. If you don't want too many options you could calibrate it so the darkest nights are at least at the 3x level.

The ASI1600 (and its ilk) also exhibit fixed pattern noise and your chequerboard pattern may be that. Calibration and dithering can help there. You did not say if you are calibrating your subs at all. Note that with the ASI1600, flat darks are recommended over bias frames.

Thanks--yes I run full calibrations (darks, bias flats and flat darks).  Having some trouble with the Lum in getting all the dust bunnies off--filter is very dirty.  The other channels calibrate pretty well.

Is offset changeable?  It may sound ridiculous, but I don't knw how to see the histogram of a sub--if I use the STF it is an artificial stretch.  I don't know how to see the natural histogram in PI.

If I change the gain to 76--do I have to change the offset (I use Maxim DL and I am not sure there is a control for offset).  What would the difference be between a 2min sub with gai 139 and gain 76, all else being equal?

Rodd

[Rodd]

36 minutes ago, kens said:

The ASI1600 (and its ilk) also exhibit fixed pattern noise and your chequerboard pattern may be that.

Here are 2 red 2min sub at gain 139--1 raw and 1 calibrated....see the checkerboard pattern?  At first glance it just seems like something that will go away after stacking 100 subs--but it does not.  A stack of 100 2min subs looks about the same as far as the black pixels.  Anyway--this is 2min.  It does not look blown out anywhere.  The background is light due to my skies--a red/yellow boundary.

Calibrated

Raw

[kens]

My ASCOM driver allows offset to be adjusted for the ASI1600MM-C but I understand that for newer cameras it is fixed - this may be the case for the Pro in which case it is probably set at 50. 

There are plenty of tools for viewing single subs. You could try FITS Liberator if yours does not allow it. What you are interested in is the mean or median value.

I can see the pattern in your raw sub. I would have thought calibration would deal with that. Perhaps you've had to do some extreme stretching ? Hard to tell from jpegs.

Can you upload (to Dropbox or wherever) and link to the raw FITs file - jpeg messes with it too much.Note that I'm no expert on fixed pattern noise as I've never had to deal with it myself. Like you I image with an ASI1600MM-C from a light polluted site.

[Rodd]

41 minutes ago, kens said:

My ASCOM driver allows offset to be adjusted for the ASI1600MM-C but I understand that for newer cameras it is fixed - this may be the case for the Pro in which case it is probably set at 50. 

There are plenty of tools for viewing single subs. You could try FITS Liberator if yours does not allow it. What you are interested in is the mean or median value.

I can see the pattern in your raw sub. I would have thought calibration would deal with that. Perhaps you've had to do some extreme stretching ? Hard to tell from jpegs.

Can you upload (to Dropbox or wherever) and link to the raw FITs file - jpeg messes with it too much.Note that I'm no expert on fixed pattern noise as I've never had to deal with it myself. Like you I image with an ASI1600MM-C from a light polluted site.

Actually there is 0 compression on teh jpegs (I always choose 0 for compression--POI gives me a choice).  But I can load a raw file.  It will have to wait until tomorrow though as I just set up for the night and the computer is in the backyard where the internet connection is shaky at best.

Do you think if I change the gain the pattern will lesson?  I used 139 becuase that is unity and I figured it was a good place to start.  I am pleased with the sub all in all--except fro the pattern noise.

Rodd

[kens]

As far as I know changing the gain will not make any difference to the pattern. What I was going to look for was where the histogram peak is. But the jpeg only had a range of 0-255 on each pixel instead of 0-65535. Just to be clear, what I was interested in is a straight out of the camera FITs file

[Rodd]

5 hours ago, kens said:

As far as I know changing the gain will not make any difference to the pattern. What I was going to look for was where the histogram peak is. But the jpeg only had a range of 0-255 on each pixel instead of 0-65535. Just to be clear, what I was interested in is a straight out of the camera FITs file

Hear is the same sub--raw

Red 2mn sec minus20-0006.fit

[kens]

That looks like a pretty good sub. The exposure looks about right with a mean of 1420 against a target value of  1240. Your Lum subs may be a bit overexposed which means more saturation and reduced dynamic range.

The pattern is not very obvious until you stretch heavily. Normally I'd expect calibration to clean it up but I recall you said you use bias frames. These are not recommended with this camera as there is too much variation from one bias frame to the next. If you use darks, flats and flat darks you don't need bias frames. I'm wondering if the master bias is causing the pattern to not calibrate out. A bit of a wild guess on my part but maybe worth a try. All you need to do is restack without the master bias. 

 

[Rodd]

24 minutes ago, kens said:

That looks like a pretty good sub. The exposure looks about right with a mean of 1420 against a target value of  1240. Your Lum subs may be a bit overexposed which means more saturation and reduced dynamic range.

The pattern is not very obvious until you stretch heavily. Normally I'd expect calibration to clean it up but I recall you said you use bias frames. These are not recommended with this camera as there is too much variation from one bias frame to the next. If you use darks, flats and flat darks you don't need bias frames. I'm wondering if the master bias is causing the pattern to not calibrate out. A bit of a wild guess on my part but maybe worth a try. All you need to do is restack without the master bias. 

 

I tried calibrating the red stack every way possible--with bias, without bias, with flat darks, without flat darks, with optimization (scaling) of darks and without optimization of darks.  The only visible difference is that optimization does not work with darks of the same exposure length as the subs (it does for the KAF 8300, but not the ASI 1600 for some reason).  Otherwise, there was no difference between using bias and not using bias.  The difference between using flat darks and not flat darks was a very slight different pixel value across the sub--like instead of a pixel being 3,371, it was 3,368--pretty much consistently a reduction of 3 or 4 across the image.  

Rodd

[kens]

Don't use dark optimization/scaling. It relies on a bias and they don't work well with these cameras.

Use flats, darks, flat darks. All at the same temperature, gain and offset. Make sure your flats and flat darks don't have very short exposures or they suffer the same issues as bias. I keep mine above 2 seconds.

I just had another look at the two images you posted originally. I'm wondering about my own sanity now as I cannot see the chequerboard pattern! Maybe I've been looking at it too long, or its an artefact of my display.

[spillage]

What settings are you using for you flats? It might be worth trying a few different ones and see what gives you the best results. So far I have found that 20k flats work for me and I also use dark flats and bias.

I did find in the beginning that flats caused me a few issues and I had to retake them a few times.

Not sure if this is relevant but I only use my light flats for all my filters.