Am I expecting too much from CMOS

[centroid]

I am struggling to get images from my 294c CMOS camera, of the quality I used to get with my SX CCD cameras, back in the day (2002-2014).

I can put up with the quirks of CMOS, e.g. green tint, gain and black level setting etc, but as yet the image quality is disappointing. It could be me of course, but I'm not new to astro imaging, a bit rusty maybe, after an 8 year hiatus, but even so.

Here is an image of the Rossette, taken the other evening. 18 x 4 min subs, with a 115mm f/7 Triplet APO + 0.8x flattener reducer, thus f/5.6, and a 294c camera with no filter.

I never had to 'push' the processing as hard as I did with this image, with my SX CCD cameras images.  A combination of Astroart 8, Photoshop CS, and Lightroom, to extract this from what was a very dull, and bland image.

Yep, I know that 18x4 min subs is probably not enough, and I have been told that it needs a good 4 hours of integration. 4 Hours 😮, I never had to get anywhere near that with my CCD cameras.

The resulting image (attached) is way below the quality that I was used to.

[iantaylor2uk]

What gain are you using? With no filter I usually use 30-60 second subs, and only  go up to 2 mins if I use an L-enhance filter, but I tend to use pretty high gain (200 on my ZWO 071 which has a max gain of 240). Most people with CMOS cameras tend to use shorter subs (as compared to CCDs) and take many of them.

[centroid]

7 minutes ago, iantaylor2uk said:

What gain are you using? With no filter I usually use 30-60 second subs, and only  go up to 2 mins if I use an L-enhance filter, but I tend to use pretty high gain (200 on my ZWO 071 which has a max gain of 240). Most people with CMOS cameras tend to use shorter subs (as compared to CCDs) and take many of them.

The gain was 9 in Astroart (ASCOM), which is equal to 900 in Sharpcap. I set this based on a sensor analysis, at the best crossover point between read noise, and gain. 

I would be interested to hear what gain settings other folk are using with the IMX 294c sensor. 

Its not easy to quote gain, when different software uses different gain setting numbers to represent the same setting. Ascom is x1, whereas Sharpcap is x100. Never had to play with the gain conundrum with CCD. I think its called progress 😅

[vlaiv]

If you want to compare results between CMOS and CCD cameras - then you have to say (or show) what sort of results you were getting with CCDs and in what sort of time / with what gear.

If you want to see if you manged to do the best with your data - or learn how to do it, if that is not the case - it is maybe best to attach raw / linear data and see what others are capable of producing.

If you worry that you did not stack properly - giving access to original data and letting other stack that for you so you can compare is best course of action.

[vlaiv]

3 minutes ago, centroid said:

Its not easy to quote gain, when different software uses different gain setting numbers to represent the same setting. Ascom is x1, whereas Sharpcap is x100. Never had to play with the gain conundrum with CCD. I think its called progress

CCDs also had gain - but you were not able to change it. If you want something like that with CMOS sensors - just pick a gain and stick with it.

 

[iantaylor2uk]

I would encourage you to play with the gain value - a lot of people stick with low gain or unity gain as they think it is better to keep a higher dynamic range, but many people have had good results with using higher gain values, and you also get lower noise at higher gains. You also get some of the dynamic range back by stacking a few hundred subs. 

[centroid]

3 minutes ago, vlaiv said:

CCDs also had gain - but you were not able to change it. If you want something like that with CMOS sensors - just pick a gain and stick with it.

 

Indeed they did Vlaiv, but it was preset by the camera manufacturer, and it worked absolutely fine, and no 'faffing' about with settings. I guess it is what happens when the astro camera manufacturers, grab readily available CMOS sensors, that were designed for use in DSLRs, and Smart Phones etc, and re-purposes them them as astro cam sensors.

[centroid]

13 minutes ago, vlaiv said:

If you want to compare results between CMOS and CCD cameras - then you have to say (or show) what sort of results you were getting with CCDs and in what sort of time / with what gear.

If you want to see if you manged to do the best with your data - or learn how to do it, if that is not the case - it is maybe best to attach raw / linear data and see what others are capable of producing.

If you worry that you did not stack properly - giving access to original data and letting other stack that for you so you can compare is best course of action.

Some of the results can be seen on my photography website. The square format ones were taken with an SC SXVR H16 (Kodak ccd sensor), with a 110mm APO at f/7, and no filter.

I no longer have any of those ccd subs, as I deleted them back in 2014 when I left the hobby, after 12 successful imaging years.

I used to pre-process (debayer, align, stack, and calibrate) in either Astroart 3,4, or 5, or Maxim DL 4. Now it is done in Astroart 8.

I never had any problem in doing this with hundreds of CCD images, or with the sensitivity of the CCD cameras.

There is no doubt the CMOS presents a whole new set of challenges to the CCD imager, now turned CMOS imager.

I guess its very much a case of, as the guy working down a hole in a New York street, said to the lady who asked "can you tell me how I get to Carnegie Hall" to which he replied "you have to practice" 😅

[vlaiv]

9 minutes ago, centroid said:

The square format ones were taken with an SC SXVR H16 (Kodak ccd sensor), with a 110mm APO at f/7, and no filter.

You are comparing mono camera with 7.4µm pixel size shooting luminance versus color camera with 4.63µm

12 minutes ago, centroid said:

I never had any problem in doing this with hundreds of CCD images, or with the sensitivity of the CCD cameras.

Well, there is a difference between CCD and CMOS image - or shell we put it like this - there is possibility for a difference (will there be any, depends on how it is used).

With CCDs, individual subs had much more signal. Larger pixel sizes and longer exposures (needed to overcome read noise) just mean more signal per sub.

There is absolutely no difference in CMOS and CCD stacks made with summing of subs (for same total duration and other parameters) - but there is difference for average stack.

Here is example:

2 + 2 + 2 + 2 + 2  versus 5 + 5

(where 2 and 5 are signal in CMOS vs CCD subs respectively)

That is just 10 versus 10

But if you average them you get 2 versus 5.

What does this mean? Well same information is just "crammed" into left part of histogram - at lower ADU values with CMOS cameras. With CCDs it was OK to use 16bit image mode when processing, but with CMOS sensors it is no longer ok to do that if one uses short exposures and since pixel sizes are small.

Another thing that happens is - you need to stretch more. If you are careful - SNR achieved will be the same (or even higher with CMOS because higher QE and lower read noise), so stretching harder will not create issues, but you need to use software capable of handling hard stretch without introducing artifacts.

Why don't you try for example Gimp to process your image - latest version (higher than 2.10 - as that is the version with full 32bit per channel support). Save your image as 32bit float point per channel format and use Gimp to process it.

(Gimp is open source / free so it will cost you nothing to try - except some time to download and install it).

[scotty38]

FWIW on my 294MC Pro I use gain 121 and leave it there but I'm new to this game really.

[centroid]

Thanks for that input Vlaiv, comprehensive, and informative as always 👍

Are you saying that I should be processing my 14bit captured images at a lower bit setting, say 8bit?

I am learning that CMOS astro imaging, and processing, is a very different 'ball game' to CCD.  I guess that many of today's imagers came into the hobby with CMOS, and no past experience with CCD astro cams.

I would love to have kept one or more of my  SX ccd cameras, so that I could make a direct comparison, using my current setup. But hindsight is a wonderful thing, and when I gave up the hobby in 2014, I had no intention of returning to it.

Out of interest, I will try GIMP, and compare the results. I have always been happy to pay for quality software, and as such, never bothered with the 'freebies'.

[centroid]

18 minutes ago, scotty38 said:

FWIW on my 294MC Pro I use gain 121 and leave it there but I'm new to this game really.

121 in what capture package?

As I said, the same gain setting is referenced differently  in different software packages, even by different camera manufacturers.

For example, a gain of 900 in Sharpcap, is the same as a gain setting of 9 under ASCOM control.

Its all done to make life uncomplicated for the end user, 🤣.

Edited January 30, 2022 by centroid
typo

[scotty38]

12 minutes ago, centroid said:

121 in what capture package?

As I said, the same gain setting is referenced differently  in different software packages, even by different camera manufacturers.

For example, a gain of 900 in Sharpcap, is the same as a gain setting of 9 under ASCOM control.

Its all done to make life uncomplicated for the end user, 🤣.

I use NINA and the ASCOM driver as below, gain goes up to 390 as the max.

 

And from the ZWO site, 390 is what they have it would seem...

Edited January 30, 2022 by scotty38

[vlaiv]

2 hours ago, centroid said:

Are you saying that I should be processing my 14bit captured images at a lower bit setting, say 8bit?

On the contrary - I'm saying you should do all your processing in 32bit per channel - float point type.

Only time when files "are allowed" to be in 16bit fixed point / integer - is at the time of the capture. As soon as you start calibrating them (as a first step of processing workflow) - I advocate use of 32bit float point precision so one does not needlessly increase noise with rounding due to use of fixed point format and low bit count.

2 hours ago, centroid said:

I am learning that CMOS astro imaging, and processing, is a very different 'ball game' to CCD.  I guess that many of today's imagers came into the hobby with CMOS, and no past experience with CCD astro cams.

Actually - it is not. It is completely the same

Well - depends on how one looks at imaging and processing.

Some people learn by example - like a skill, while others have method of learning by association and deeper understanding.

First approach is much easier / comes natural - that is how we learn to ride a bike and similar things. Problem with that approach is that if circumstances change - things that we have learned seem not to work any more. For people that learn with second method (or rather - when person learns something with second method - I don't think there is clear distinction and we all learn some things with either methods) - nothing really changed except "problem setup / constraints". Solving of the problem still follows the same rules although parameters are different.

In that sense - for me, imaging is the same - calibration files do the same thing - remove unwanted signal. Stacking does the same thing. It is same choice of exposure length depending on read noise and other noise sources. Gain is there to represent conversion factor between electrons and ADUs and so on....

Once image is stacked - can we really tell if it came from CCD or CMOS? It is just data - it can be properly handled or improperly. If it is properly handled, then again - processing can be done the same on it regardless of the source. It is "form of the data" that dictates how will it be processed and not where the data came from.

2 hours ago, centroid said:

Out of interest, I will try GIMP, and compare the results. I have always been happy to pay for quality software, and as such, never bothered with the 'freebies'.

You should. Fact that something is free - does not mean it is "cheap" or less capable. You can always spend some money on it if you find it useful - there is almost always a donate button with such open source projects.

I guess difference between free and paid software in that regard is - with paid software you pay amount that developers think software is worth to you and with free software you can pay amount you think the software is forth worth to you

 

Edited January 30, 2022 by vlaiv
typo ...

[centroid]

If I can get somewhere near this with a CMOS camera, I will be both convinced, and happy.

I this took this image back in January 2012  with a SX  SXVR H16 CCD, (Kodak sensor), with a 110mm f/7 refractor, albeit separate LRGB. All in one evening, and not one night per filter, as someone mentioned on another group, that is how he uses his CMOS camera to get decent results.

Comment:  for some obscure reason it displays excessively bright posted here, but not on my calibrated PC monitor, or on my website.

Edited January 30, 2022 by centroid
added comment

[Stuart1971]

6 hours ago, centroid said:

I am struggling to get images from my 294c CMOS camera, of the quality I used to get with my SX CCD cameras, back in the day (2002-2014).

I can put up with the quirks of CMOS, e.g. green tint, gain and black level setting etc, but as yet the image quality is disappointing. It could be me of course, but I'm not new to astro imaging, a bit rusty maybe, after an 8 year hiatus, but even so.

Here is an image of the Rossette, taken the other evening. 18 x 4 min subs, with a 115mm f/7 Triplet APO + 0.8x flattener reducer, thus f/5.6, and a 294c camera with no filter.

I never had to 'push' the processing as hard as I did with this image, with my SX CCD cameras images.  A combination of Astroart 8, Photoshop CS, and Lightroom, to extract this from what was a very dull, and bland image.

Yep, I know that 18x4 min subs is probably not enough, and I have been told that it needs a good 4 hours of integration. 4 Hours 😮, I never had to get anywhere near that with my CCD cameras.

The resulting image (attached) is way below the quality that I was used to.

I have asked this same question in a thread in here, since buying my QHY268c, and also made the same comments on not getting the quality images I got with my SX CCD, but am sticking at it, things have got a bit better, but I am not getting the results that other people are with the same camera and much less integration time, its quite disheartening when I see what some people get with say 2 hours of 2 min subs, and if I try the same settings I barley have anything to show for it…☹️
But my SX CCD is now sold, so I have to stick with it….

[centroid]

Reduced brightness for posting here.

 

Edited January 30, 2022 by centroid

[centroid]

19 minutes ago, Stuart1971 said:

I have asked this same question in a thread in here, since buying my QHY268c, and also made the same comments on not getting the quality images I got with my SX CCD, but am sticking at it, things have got a bit better, but I am not getting the results that other people are with the same camera and much less integration time, its quite disheartening when I see what some people get with say 2 hours of 2 min subs, and if I try the same settings I barley have anything to show for it…☹️
But my SX CCD is now sold, so I have to stick with it….

At least its not just me Stuart.

If I hadn't imaged successfully for 12 years, with a variety of SX CCD cameras, I wouldn't have known any different. But I have experienced what CCD can produce, and I have yet to see it from this CMOS camera, and its re-purposed sensor.

However, its all I have to work with now, and having paid £900 for it, I will persist, and maybe one day it will either impress me, or go in the bin. At which point I will either grab one of  the last remaining SX CCD cameras, or retire again from astro imaging, never to return.

I have seen some good images taken with CMOS astro cams, but I have seen many more that are very 'lack lustre'. Back in the CCD days, it was the other way around.

Edited January 30, 2022 by centroid

[SamAndrew]

If you're dealing with light polution, my experience has been that mono is superior; easier to process and with better results

I picked up a 2nd hand 294C a couple of years ago, but quickly sold it on as I much prefered the data from my 1600MM

 

 

[ONIKKINEN]

3 hours ago, centroid said:

If I can get somewhere near this with a CMOS camera, I will be both convinced, and happy.

I this took this image back in January 2012  with a SX  SXVR H16 CCD, (Kodak sensor), with a 110mm f/7 refractor, albeit separate LRGB. All in one evening, and not one night per filter, as someone mentioned on another group, that is how he uses his CMOS camera to get decent results.

Comment:  for some obscure reason it displays excessively bright posted here, but not on my calibrated PC monitor, or on my website.

Isnt LRGB like (at least) 50% faster than OSC? And the pixel sizes between that camera and your current one are also wildly different so i dont know if these 2 could be compared. Youll probably need 2 nights to reach this if the shooting conditions are the same as they were then, and more than 2 if they have gotten worse.

[Budgie1]

With the ASI294MC Pro I got from @SamAndrew I've only used the ZWO native driver and the ASCOM driver and both have the same gain settings. Software wise, I've used the ASIAir Plus but mostly I use APT, this uses the same gain & offset settings from the native & ASCOM drivers.

Similar to @scotty38, I use a gain of 120, based on the ZWO information he posted above. I have only ever use this gain setting and the only setting I've tried to change is the offset. I started with an offset of 8 (seems to be the default offset for the ZWO native driver) and then tried 30 (default for the ASCOM driver). I found the noise levels on the final image went up when I use offset 30, so I went back to 8 and stuck with that. 

If you're not using any filter with this camera, I would advise at least trying a UV/IR cut filter because the front window in ASI294MC doesn't have this coating and it will help to control the stars.

Rather then posting some of the images I've managed to get with this camera, there are a few HERE.

[MartinB]

Could the issue be that you are comparing OSC with mono Dave?  In general I would opt for mono nearly every time although the latest batch of colour CMOS cameras with a decent nebular filter and fast optics are delivering some very impressive results.

I am posting a Rosette image from 2019 which I hope demonstrates why I am very happy with CMOS.  This image was taken with a mono 12bit CMOS camera (ZWO ASI 1600MM pro) using Baader 7nm filters.  30x5mins Ha, 7x5mins OIII and 20 x 5mins SII.  The data was captured over a few weeks and there were lots of passing cloud throughout the capture, hence paucity of OIII particularly.  I don't think I would have got any data using my QSI 532 which needed 30 min subs to be getting over the read noise hump.

I was using a 200mm camera lens at F2.8.  Although I was undersampled it was getting into the right sort of range given the the cameras relatively small pixels.  I have added a crop which I hope demonstrates that the resolution isn't that bad.  There is no way I could have done anything like this with my QSI.  I have also attached the Ha mono which shows how well the camera has picked up the faint outer nebulosity, not bad for a 12 bit "re purposed chip"  The mount was a little SW EQ35M pro and wasn't behaving itself very well.  So, a 200mm camera lens, a pretty basic mount and, by CMOS standards, an old chip.  Using this set up I would feel able to use 30 second unguided exposures using my ancient Vixen Super Polaris mount and a cheap old ebay 200mm lens.  The door to serious deep sky imaging has been properly opened to would be imagers on a tight budget.

 

 

 

Terrible framing, I hadn't realised there was so much nebuloisty over to the right.

[tomato]

Hi Dave, I can’t give you a direct comparison of a OSC CCD vs CMOS as I moved to mono CCD imaging after a short time with a little Atik 314 OSC. However here is the closest I can get which is a 6 panel M 31, capturing Lum with a KAF 8300 sensor and RGB with a QHY 268c, both on an Esprit 150 dual rig so the same scope and imaging conditions. To be fair, the CCD was using the Esprit 0.77 focal reducer while the CMOS had the field flattener so the CCD was imaging at 1.33 arc sec per pixel, with the CMOS at 0.714, but the FOVs were very similar.

I cut my CMOS imaging teeth over a season imaging with ASI 178 mono sensors before moving to the OSC, but early on my processing skills lacked a lot so I wasn’t doing either sensor justice. The CMOS break through was doing shorter subs, I have never bothered playing around with gain or offset.

LUM CCD

 

RGB CMOS

[Stuart1971]

18 minutes ago, tomato said:

Hi Dave, I can’t give you a direct comparison of a OSC CCD vs CMOS as I moved to mono CCD imaging after a short time with a little Atik 314 OSC. However here is the closest I can get which is a 6 panel M 31, capturing Lum with a KAF 8300 sensor and RGB with a QHY 268c, both on an Esprit 150 dual rig so the same scope and imaging conditions. To be fair, the CCD was using the Esprit 0.77 focal reducer while the CMOS had the field flattener so the CCD was imaging at 1.33 arc sec per pixel, with the CMOS at 0.714, but the FOVs were very similar.

I cut my CMOS imaging teeth over a season imaging with ASI 178 mono sensors before moving to the OSC, but early on my processing skills lacked a lot so I wasn’t doing either sensor justice. The CMOS break through was doing shorter subs, I have never bothered playing around with gain or offset.

LUM CCD

 

RGB CMOS

Very nice, what amount and duration of frames in the CMOS image, and any filters used…?

[tomato]

Thanks, the 268c was in DSO photographic mode, 120 sec exposures using an IR/UV cut filter. Each panel was 40 minutes of integration.