Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

ASI2600MC Pro – Gain 0 vs. 100?

astrophotography CMOS imaging
  • Please log in to reply
74 replies to this topic

#26 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 26 August 2020 - 04:52 PM

I've settled on Gain 100 and Offset 50 too.

 

I don't mind getting more subs... Dither, Flats, Flat Darks, Calibrate & Enjoy



#27 PilotAstronomy

PilotAstronomy

    Vostok 1

  • *****
  • Posts: 137
  • Joined: 25 Sep 2017

Posted 26 August 2020 - 07:46 PM

Well, after a long, painful stacking process, the ~ 6 hours of 60 second, 100 gain image is NOTICEABLY less noisy than a similar total integration time on the 180 second, 0 gain image (both offset 50). That's after calibration, correction, debayering, subframe selection weighting, registration, local normalization, integration, and drizzle integration on the stacks. I dithered every 3 minutes on each...every 3 frames on the 60 second/100 gain target..every frame on the 180 second/0 gain target. The 60 second, 100 gain stack was 3x the files and 3x the harddrive space and 3x the processing time. But before any digging into the processing I think that effort is worth the squeeze.


  • RFtinkerer and const like this

#28 buckeyestargazer

buckeyestargazer

    Vendor - Buckeyestargazer.net

  • *****
  • Vendors
  • Posts: 5,422
  • Joined: 12 Jan 2008
  • Loc: IN, USA

Posted 27 August 2020 - 08:03 AM

Well, after a long, painful stacking process, the ~ 6 hours of 60 second, 100 gain image is NOTICEABLY less noisy than a similar total integration time on the 180 second, 0 gain image (both offset 50). That's after calibration, correction, debayering, subframe selection weighting, registration, local normalization, integration, and drizzle integration on the stacks. I dithered every 3 minutes on each...every 3 frames on the 60 second/100 gain target..every frame on the 180 second/0 gain target. The 60 second, 100 gain stack was 3x the files and 3x the harddrive space and 3x the processing time. But before any digging into the processing I think that effort is worth the squeeze.

Can you quantify that statement?  How images "look" is not a good indication of the real noise profile.  If you use Pixinsight there is a script called NoiseEvaluation that will give you a quantifiable number of the noise.  I'd be interested to see the results for both stacks.  


  • rockstarbill likes this

#29 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 27 August 2020 - 10:05 AM

Can you quantify that statement?  How images "look" is not a good indication of the real noise profile.  If you use Pixinsight there is a script called NoiseEvaluation that will give you a quantifiable number of the noise.  I'd be interested to see the results for both stacks.  

Curious what we would get out of this number in any meaningful way.  Without doing the math to know when the SNR converges or at which point you want to compare along the way, it will just lead to argument ad nauseum won't it?

 

Short of doing the math, "look" is how most people would do it right?

 

Let's say you know your scopes performance, sensors performance, the emission rate of object and the emission rate of sky glow and you calculate a total integration time against diminishing returns. (the point at which either 0 gain or unity gain will stop giving any noticeable SNR improvement)

 

The ONLY thing you could know up front is that  30 second exposures at Unity gain will equal 300 second exposures at gain 0 in total integration time X if you calculate the convergence of SNR (against diminishing returns for ease of a normalized wall)

 

Knowing that, in 6 hours, the stack at gain 0 will have a lower SNR than the stack at unity gain simply because you actually need gain 0 subs to be (roughly) 300 seconds long for the total integration time to be anywhere near in equal SNR.

 

Maybe the test should be 30-60 seconds at unity vs ~300 seconds at gain 0 - my calculations show at 6 hours the 300 second deep well would surpass 30 seconds at unity in SNR but one could move to 90 seconds at Unity  and still be slightly ahead of 300 seconds at gain 0... it's an interesting balance

 

In the end - does the deep well give something visually that the unity gain doesn't? that's what i'd be curious about or is unity gain doing a good enough job? Reality is the math is just a calculation on averages at best... sky rate/conditions/seeing changes everything and changes all the time so eyeballing is a good way to keep this hobby sane. BUT.. knowing some of the fundamental exposure times based on read error rates and how they impact stack integration time at least gives you some "Rails" to live within.


Edited by sn2006gy, 27 August 2020 - 11:07 AM.


#30 buckeyestargazer

buckeyestargazer

    Vendor - Buckeyestargazer.net

  • *****
  • Vendors
  • Posts: 5,422
  • Joined: 12 Jan 2008
  • Loc: IN, USA

Posted 27 August 2020 - 02:11 PM

Curious what we would get out of this number in any meaningful way.  Without doing the math to know when the SNR converges or at which point you want to compare along the way, it will just lead to argument ad nauseum won't it?

 

Short of doing the math, "look" is how most people would do it right?

 

Let's say you know your scopes performance, sensors performance, the emission rate of object and the emission rate of sky glow and you calculate a total integration time against diminishing returns. (the point at which either 0 gain or unity gain will stop giving any noticeable SNR improvement)

 

The ONLY thing you could know up front is that  30 second exposures at Unity gain will equal 300 second exposures at gain 0 in total integration time X if you calculate the convergence of SNR (against diminishing returns for ease of a normalized wall)

 

Knowing that, in 6 hours, the stack at gain 0 will have a lower SNR than the stack at unity gain simply because you actually need gain 0 subs to be (roughly) 300 seconds long for the total integration time to be anywhere near in equal SNR.

 

Maybe the test should be 30-60 seconds at unity vs ~300 seconds at gain 0 - my calculations show at 6 hours the 300 second deep well would surpass 30 seconds at unity in SNR but one could move to 90 seconds at Unity  and still be slightly ahead of 300 seconds at gain 0... it's an interesting balance

 

In the end - does the deep well give something visually that the unity gain doesn't? that's what i'd be curious about or is unity gain doing a good enough job? Reality is the math is just a calculation on averages at best... sky rate/conditions/seeing changes everything and changes all the time so eyeballing is a good way to keep this hobby sane. BUT.. knowing some of the fundamental exposure times based on read error rates and how they impact stack integration time at least gives you some "Rails" to live within.

I certainly don't want this to turn into one of those threads filled with super math and numbers that I certainly don't understand.  You obviously know far more about this than I do. I'm simply asking for one number to compare.  I use NoiseEvaluation to simply measure the noise of a stacked image to help me determine how much data rejection to use vs increasing noise.  So I have a simple idea of what NoiseEvaluation shows.  

 

I agree that most of us (myself included) simply "look" at our images to try and see the best outcome, but there have been many times when I'm looking at an image that looks far less noisy than another, but in reality it was the reverse after more careful inspection.  

 

I'm not claiming anything here.  I'm just curious to know that number between the two stacked images with different gains.  



#31 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 27 August 2020 - 02:37 PM

I certainly don't want this to turn into one of those threads filled with super math and numbers that I certainly don't understand.  You obviously know far more about this than I do. I'm simply asking for one number to compare.  I use NoiseEvaluation to simply measure the noise of a stacked image to help me determine how much data rejection to use vs increasing noise.  So I have a simple idea of what NoiseEvaluation shows.  

 

I agree that most of us (myself included) simply "look" at our images to try and see the best outcome, but there have been many times when I'm looking at an image that looks far less noisy than another, but in reality it was the reverse after more careful inspection.  

 

I'm not claiming anything here.  I'm just curious to know that number between the two stacked images with different gains.  

Cool :) I get it.

 

I mostly look too!

 

BTW, If the weather stays nice, I may run a test on a few of these.   Unfortunately, my sky at home has now officially gone Orange - bortle 8 so i can't shoot broadband too easily but i may do the math on my Triad filter and see how gain 0 works vs unity gain when done on the assumption that unity gain subs will be shorter and gain 0 subs longer to overcome noise. Obviously biased for a super narrowband filter

 

Or maybe i'll get out to club site and shoot Andromeda... i'd just have to do 2 nights as i'd want its  relative altitude to be the same... too many variables.

 

I am curious if the deep well plays any role in image quality or if unity gain is just ok to bet on.


  • Bill Shaheen and buckeyestargazer like this

#32 Umasscrew39

Umasscrew39

    Apollo

  • *****
  • Moderators
  • Posts: 1,411
  • Joined: 07 Dec 2016
  • Loc: Central Florida

Posted 28 August 2020 - 06:00 AM

I've been taking a number of images with the 2600 at gain 100/50 offset.  Noise is very low, detail is excellent, and color is rich.  Pretty much mimics what I was getting with the 533.  M13 was only captured in 13 minutes (30s subs) and the detail is pretty amazing for such a short period of time.  The others are 120s subs for about 3 total hours.  I haven't tried gain 0 and not sure I need to as I have been pretty happy with these results. 

 

 

get.jpg?insecure

 

get.jpg?insecure

 

get.jpg?insecure


  • Jim Waters and elmiko like this

#33 buckeyestargazer

buckeyestargazer

    Vendor - Buckeyestargazer.net

  • *****
  • Vendors
  • Posts: 5,422
  • Joined: 12 Jan 2008
  • Loc: IN, USA

Posted 28 August 2020 - 08:04 AM

I've been taking a number of images with the 2600 at gain 100/50 offset.  Noise is very low, detail is excellent, and color is rich.  Pretty much mimics what I was getting with the 533.  M13 was only captured in 13 minutes (30s subs) and the detail is pretty amazing for such a short period of time.  The others are 120s subs for about 3 total hours.  I haven't tried gain 0 and not sure I need to as I have been pretty happy with these results. 

 

 

get.jpg?insecure

 

get.jpg?insecure

 

get.jpg?insecure

Three very nice images with nice details.  I'd be happy with any of them!

But notice that there is hardly any color in the stars.  One possible reason for that is because the stars at gain 100 are getting blown out/saturated quickly, even at 30s exposures, which is exactly why I am using gain 0 for LRGB for the increased full well capacity.  Of course this depends on other light gathering properties too (you have a large mirror to suck in more light quickly).  For LRGB the read noise just doesn't matter that much.  You swamp the read noise pretty quickly with LRGB.  So I just don't think there's much advantage for using gain 100 with LRGB images (for deep sky images - solar system is another story).  


  • elmiko and Umasscrew39 like this

#34 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 878
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 28 August 2020 - 11:42 AM

I have only had a single outing with my new ASI2600MC-Pro so I am by no stretch of the imagination an expert on the camera. 
 
I did select gain 100 for my first outing but I was wondering how that would affect star colors.  However, I went with gain 100 because of the read noise advantage.
 
The is the result with an FSQ-106 at f/3.6 with the 645 reducer:

get.jpg?insecure

Not all of the stars have color but certainly some of the bright yellow stars to the right of the galaxy look fine. Those bright yellow stars were as bright as 56k counts in the original subs.

So, I am going to stick with gain 100 for now, unless I specifically point the scope toward something like the double cluster where I am not also trying to capture faint nebulosity in the frame.

Based on my limited experience with the camera I am inclined to think that the star color problem is to some extent attributable to processing choices.

Umasscrew39 can certainly look at his original subs to see which, if any, stars are saturated. I would be interested in hearing whether they are.
  • elmiko and davidparks like this

#35 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 28 August 2020 - 12:24 PM

That's a fantastic m31!



#36 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 878
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 28 August 2020 - 01:06 PM

Thanks. The new color cameras like the ASI2600 and the ASI6200 with their Sony 5th gen. Exmor R back illuminated sensors have convinced me to try a color camera for the first time since I bought my first camera [an SBIG ST-7] back in 1998.

 

I definitely like the simplicity of using a color camera and after this first outing I think that I will keep the ASI2600MC.  I use a high resolution encoder mount to do unguided images and I was worried that the color camera wouldn't work well with with the point mapping and plate solving program that I use to create tracking rate correction models, but there weren't any problems. 

 

I can definitely use the camera on the bright stuff while I am waiting for an ASI2600MM-Pro mono camera.



#37 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 28 August 2020 - 02:11 PM

 

I can definitely use the camera on the bright stuff while I am waiting for an ASI2600MM-Pro mono camera.

It works on the faint stuff too...

 

This is Abbel 85... a single night of data with 2600mc & Triad ultra shot at 80% moon in a bortle 8.. Plan to double this.  You can see the blue and red and if you zoom in some of the feint wisps of the novae bubble. The problem i have is a reducer/ff that isn't working for full aps-c so i have that replacement coming in next week.

 

supernovae 4
 
The secret sauce is to dither and bayer-drizzle. Greatly diminishes any penalty of bayer matrix and with the triad i samble ha/hb, sii, oiii for entire duration

Edited by sn2006gy, 28 August 2020 - 02:14 PM.

  • Jim Waters, buckeyestargazer, elmiko and 1 other like this

#38 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 28 August 2020 - 02:29 PM

Unfortunately you can't zoom in on CN's image viewer. Once i get the second night, i'll post final and share to astrobin :)



#39 Jim Waters

Jim Waters

    Fly Me to the Moon

  • *****
  • topic starter
  • Posts: 5,476
  • Joined: 21 Oct 2007
  • Loc: Phoenix, AZ USA

Posted 28 August 2020 - 02:33 PM

 

It works on the faint stuff too...

 

This is Abbel 85... a single night of data with 2600mc & Triad ultra shot at 80% moon in a bortle 8.. Plan to double this.  You can see the blue and red and if you zoom in some of the feint wisps of the novae bubble. The problem i have is a reducer/ff that isn't working for full aps-c so i have that replacement coming in next week.

 

 
 
The secret sauce is to dither and bayer-drizzle. Greatly diminishes any penalty of bayer matrix and with the triad i samble ha/hb, sii, oiii for entire duration

 

What scope were you using?



#40 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 28 August 2020 - 03:15 PM

What scope were you using?

Just an old Orion ED80. Nothing fancy at all.  .8x reducer that seems to have trouble at aps-c. 

 

I should have had an Epsilon 130 by now but UPS lost it... Hopefully UPS doesn't lose my Apex-l that is on the way ;)



#41 Umasscrew39

Umasscrew39

    Apollo

  • *****
  • Moderators
  • Posts: 1,411
  • Joined: 07 Dec 2016
  • Loc: Central Florida

Posted 28 August 2020 - 05:37 PM

Three very nice images with nice details.  I'd be happy with any of them!

But notice that there is hardly any color in the stars.  One possible reason for that is because the stars at gain 100 are getting blown out/saturated quickly, even at 30s exposures, which is exactly why I am using gain 0 for LRGB for the increased full well capacity.  Of course this depends on other light gathering properties too (you have a large mirror to suck in more light quickly).  For LRGB the read noise just doesn't matter that much.  You swamp the read noise pretty quickly with LRGB.  So I just don't think there's much advantage for using gain 100 with LRGB images (for deep sky images - solar system is another story).  

 

 

I have only had a single outing with my new ASI2600MC-Pro so I am by no stretch of the imagination an expert on the camera. 
 
I did select gain 100 for my first outing but I was wondering how that would affect star colors.  However, I went with gain 100 because of the read noise advantage.
 
The is the result with an FSQ-106 at f/3.6 with the 645 reducer:

get.jpg?insecure

Not all of the stars have color but certainly some of the bright yellow stars to the right of the galaxy look fine. Those bright yellow stars were as bright as 56k counts in the original subs.

So, I am going to stick with gain 100 for now, unless I specifically point the scope toward something like the double cluster where I am not also trying to capture faint nebulosity in the frame.

Based on my limited experience with the camera I am inclined to think that the star color problem is to some extent attributable to processing choices.

Umasscrew39 can certainly look at his original subs to see which, if any, stars are saturated. I would be interested in hearing whether they are.

I checked and the stars are definitely not saturated.  So, maybe it is my poor processing skills.


  • buckeyestargazer likes this

#42 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 878
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 28 August 2020 - 05:59 PM

I checked and the stars are definitely not saturated.  So, maybe it is my poor processing skills.

I wouldn’t say “poor”. ;-) But maybe somewhere along the way you did something that caused some of the field stars to loose color.



#43 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,627
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 28 August 2020 - 06:02 PM

Anyone hear about a potential ASI2600MM mono camera coming to market any time soon?



#44 sn2006gy

sn2006gy

    Vendor - Rockchuck Summit Observatory

  • ****-
  • Vendors
  • Posts: 1,542
  • Joined: 04 May 2020
  • Loc: Austin, TX

Posted 28 August 2020 - 06:05 PM

Anyone hear about a potential ASI2600MM mono camera coming to market any time soon?

ZWO denies it.

 

But QHY has one in beta and hitting retail around October.

 

With that said, I have 0 interest/desire to move to mono. The color is hitting all the right notes.


  • telfish, Peregrinatum and paolop like this

#45 Umasscrew39

Umasscrew39

    Apollo

  • *****
  • Moderators
  • Posts: 1,411
  • Joined: 07 Dec 2016
  • Loc: Central Florida

Posted 28 August 2020 - 06:13 PM

I wouldn’t say “poor”. ;-) But maybe somewhere along the way you did something that caused some of the field stars to loose color.

I follow some of the common steps of the PI gurus so not sure what that would be.  I showed some of my images to pfile (Rob) from the PI team and he said this is a common problem.  Since the stars are not saturated, he suggested I try the MaskedStretch or ArcSinH stretch as initial stretches to try to preserve the color instead of the HT.  So, I'll give that a try.



#46 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 878
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 28 August 2020 - 07:51 PM

I follow some of the common steps of the PI gurus so not sure what that would be.  I showed some of my images to pfile (Rob) from the PI team and he said this is a common problem.  Since the stars are not saturated, he suggested I try the MaskedStretch or ArcSinH stretch as initial stretches to try to preserve the color instead of the HT.  So, I'll give that a try.

If you are using the white point slider in HT to brighten up the image, then that could be a potential cause.  Moving the white point slider to the left could change what was a bright colored star to a white star.


Edited by Dean J., 28 August 2020 - 08:22 PM.


#47 Peregrinatum

Peregrinatum

    Surveyor 1

  • *****
  • Posts: 1,627
  • Joined: 27 Dec 2018
  • Loc: South Central Valley, Ca

Posted 29 August 2020 - 12:19 PM

ZWO denies it.

 

But QHY has one in beta and hitting retail around October.

 

With that said, I have 0 interest/desire to move to mono. The color is hitting all the right notes.

Cool, I'm going to be all over this one with the higher QE.  I have Bortle 7 skies so mono really helps, and it's the same amount of imaging time for me.



#48 Dean J.

Dean J.

    Viking 1

  • -----
  • Posts: 878
  • Joined: 13 Aug 2011
  • Loc: Vista, CA

Posted 29 August 2020 - 12:44 PM

Anyone hear about a potential ASI2600MM mono camera coming to market any time soon?


QHY has announced that they are going to be releasing a camera with a mono version of the IMX571 detector in October:

https://www.youtube....youtu.be&t=2753

Roll back the time to about 42:30 to see the discussion on the QHY268M.

I will get a QHY if ZWO doesn't get their act together and get their mono version out in short order.

Edited by Dean J., 29 August 2020 - 12:44 PM.

  • Peregrinatum likes this

#49 Umasscrew39

Umasscrew39

    Apollo

  • *****
  • Moderators
  • Posts: 1,411
  • Joined: 07 Dec 2016
  • Loc: Central Florida

Posted 29 August 2020 - 01:32 PM

If you are using the white point slider in HT to brighten up the image, then that could be a potential cause.  Moving the white point slider to the left could change what was a bright colored star to a white star.

I never use the white point slider and never had star color issues with other cameras but I just discovered the major problem.  For my 2600 images, I forgot to apply the preview background while doing ColorCalibration.  I am a neuroscientist and this is technically called a "brain cramp".  I did a quick redo of them and saw more color stars.  However, I also rechecked for star saturation.  Based on the suggestion of some PI gurus that star color can be preserved if you do not go over 0.8 on the normalized scale, I am pushing it in some cases.  So, this could argue with my setup to try a gain of 0.  From all of this, trying options like gain of 0, ArcsinhStretch, MaskedStretch, and PCC are all viable alternatives to reduce saturation or increase star color.  On the other had, I find star color exaggerated in many images which are pleasing to the eye but I personally do not like the look but , of course, is a personal preference thing.

 

Anyway, this was a good topic to discuss for the camera and processing techniques.  Always new things to learn and try. 



#50 PilotAstronomy

PilotAstronomy

    Vostok 1

  • *****
  • Posts: 137
  • Joined: 25 Sep 2017

Posted 29 August 2020 - 02:14 PM

Can you quantify that statement?  How images "look" is not a good indication of the real noise profile.  If you use Pixinsight there is a script called NoiseEvaluation that will give you a quantifiable number of the noise.  I'd be interested to see the results for both stacks.  

 

How they "look" is a good measure as far as I'm concerned, but in the interest of "science" I'll take a look at those stacks for ya. I'll get back to you on this.




CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics





Also tagged with one or more of these keywords: astrophotography, CMOS, imaging



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics