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

Which do you prefer - CCD or CMOS?

  • Please log in to reply
18 replies to this topic

#1 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 25 May 2019 - 04:39 AM

I am really battling here with my CMOS camera.  After lots of reading, questions here and more experimentation I really am considering moving back to CCD. 

 

As much as i like the ZWO 1600 pro mono and on first impressions the images that i am now getting look ok in their raw format, when put together with the other filters, the results are pretty underwhelming.   (30xL@120s - 20xR/G/B@240s all at G76/50)  then run through PI BPP.

 

As a quick comparison the other night, i attached my Atik 460ex to the same scope and image train that i have been using the 1600 on and took a couple of 300 and 600s subs of M51.  A quick look in PI and i think the results are so much better. 

 

I thought perhaps its my processes in PI which are not delivering the results, so i ran a set of LRGB images i took last year through and the results came out nice - same process with PI on the 1600 subs... horrible.  Noisy, no defined details and difficult to adjust the colours to anything remotely like what M101 should look like!

 

I really am tempted to get an Atik 383L again as the results i had with that camera were very good..  Or, find a similar camera with the KAF CDD in it.

 

Its quite annoying as i see some very good images from others using mono CMOS cameras which are much the same as mine..   I am either still doing something very fundamentally wrong or just need to get back to what i know and get something to show for my time imaging.. and perhaps revisit CMOS at a later date.

 

Anyone else relate to this and have made the same conclusion??

 



#2 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23069
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 25 May 2019 - 11:27 AM

Well first off, there is nothing wrong with preferring CCD. CCD cameras have not been invalidated by CMOS, and they still have their advantages. 

 

I would be curious to see some of your data, though. Darks, flats, biases/flat darks and lights. I'd like to process it a bit myself and see what I can do. CMOS is different than CCD in key ways, so it does require some differences in how you do things. For one, dithering is often more important with CMOS, and you have not mentioned whether you dither or not or how often. Additionally, dark frames MUST be well-matched, in that they match the exposure time, temp, gain and offset of the light frames exactly, and cannot be scaled.

 

Now, there are a couple key differences between the ASI1600 an Atik460 that should be noted here. For one, you used exposures significantly longer. You used 300 an 600 second subs vs. 120 and 240 second. So you have 2.5x the signal per sub with the 460. On top of that, the 460 pixels are larger, so you actually have more than 2.5x the signal, you have nearly 3.6x the signal. So if you are comparing 460ex subs with 1600 subs, of course the 460ex subs are going to look better. They have significantly more signal...tough to beat that.

 

Now, this is not to say that there may indeed be something wrong with the ASI1600, or how the data has been handled. If there was not enough dithering, or if the darks are not well matched, etc. then the ASI1600 data could indeed end up looking worse in the end. However I would be curious to know what 300 second long L and 600 second long RGB subs from the ASI1600 look like at gain 0. The ASI1600 has a 20ke- FWC, while the 460ex has an 18ke- FWC. So you should be able to use the same longer exposures. In fact, with the smaller pixels, you might even be able to use exposures longer than these, perhaps 450s and 900s. Just as a test case, to see if you can use the ASI1600 in a manner that would give you the results you are looking for. It may indeed just be how you are using it, so changing something up may be worth the time to try and experiment before you swap out cameras.

 

I also want to stress the importance of dithering. If you feel you are not getting clean results, and you ARE using well matched master dark frames that are not scaled, then you should try to dither every frame. With longer exposures at Gain 0, you should be able to dither every frame without much overhead cost. Dithering aggressively every frame, and dithering randomly in both axes, should help smooth out any remnant FPN.

 

Finally, if you still cannot get the results you are looking for with longer subs at a lower gain and with well-matched darks and dithering, then it sounds like you might want to try another camera. The ASI1600 uses the Panasonic M sensor. While this sensor, IMO, works really well at high gain (i.e. gain 200), at lower gains the quantization error bites you and it has more FPN issues. If your goal is to do LRGB imaging, the ASI1600 may not be the best camera for the job, and if resolution and details are not your top concern then an Atik 383L may indeed be a better option. The IMX183 sensor has different characteristics than the Panasonic M. This sensor, also a CMOS, works much better at lower gains and not quite as well at higher gains. It may be that the IMX183 would work for you. It depends on the size of the field you need, as the IMX183 is more around the size of the Atik460 than the Atik383. The main difference between the 460 and 183 would be the pixel size, with the 183 having pixels that are much smaller (almost 1/4 the area). This might be a bonus for resolution, but you might suffer on SNR.


  • psandelle and Akwilliams like this

#3 jgraham

jgraham

    Hubble

  • *****
  • Posts: 19980
  • Joined: 02 Dec 2004
  • Loc: Miami Valley Astronomical Society

Posted 25 May 2019 - 12:33 PM

Hmmmm, personally, I have never noticed any differences between my CMOS and CCD cameras. I acquire, calibrate, and process my images extactly as I have always done. I know that CMOS sensors had problems with consistency way back in their early days, but those days are long gone. Even then I didn't have any real problems with image calibration.

 

Weird.


  • Akwilliams likes this

#4 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 25 May 2019 - 12:53 PM

John, you have been a world of help now and previous.   What I will do is upload the full set of subs, darks and flats and you can take a look for yourself and see if the data is good or there might be something amiss.

 

Regarding dither,  I am currenty set in SGP for High Dither every 3 frames.

 

Ill post a link up shortly


Edited by Akwilliams, 25 May 2019 - 01:03 PM.


#5 psandelle

psandelle

    Mercury-Atlas

  • *****
  • Posts: 2510
  • Joined: 18 Jun 2008
  • Loc: Los Angeles

Posted 25 May 2019 - 01:23 PM

Hang in there (and everything John said). I find CMOS cameras much more finicky than my ol' CCD cameras were, but the advantages are outweighing, at the moment. But boy can they be "techy" when they want to be, especially with calibration files. I actually went to APP for pre-processing over PI lately (just plain easier). Hang in there!

 

Paul


  • Akwilliams likes this

#6 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 25 May 2019 - 01:46 PM

Here is a link to subs and calibration frames.    LRGB Lights, Darks for L@120s - 15, Darks for RGB@240s -15 and a sample of flats for the filters

 

https://horsebridged...I4B09Q?e=NNrNfz



#7 stargzr66207

stargzr66207

    Ranger 4

  • -----
  • Posts: 336
  • Joined: 28 Aug 2014
  • Loc: Kansas, USA

Posted 25 May 2019 - 01:47 PM

This is a VERY interesting thread!! I have both CCD and CMOS cameras: An SXVR-H694C one-shot
ccd (since 2006) and an ASI 071 MC Cool (since 2017). I have been pleased with the results
from both, but am ALWAYS searching for ways to improve my results. The comments by John Rista
and Paul are very interesting, but I need "translation" of a couple of terms: John, what is
"remnant FPN", and Paul, what is APP? Thanks in advance! I look forward to following this
thread as it develops.

Ron Abbott

#8 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 25 May 2019 - 02:08 PM

Ideally, I do not want to get rid of the ASI1600 mono.  I do think it’s a great camera and proved very reliable.  I am hoping that it’s something I have done or are doing that is giving me grief.

 

When I make the comparisons between the 1600 and the 460 it was purely to see if the my process of getting the images and then processing was correct, not to actually make a direct comparison between the two... sorry if it didn’t come across that well.   

 

When I did have the 383L, I found that I could take good images and be able to process them quite nicely and I guess I am expecting similar results from the 1600, but because it’s so vastly different when I do the basic steps of processing in PI i became very disillusioned... hence considering going back to the 383 which is something I know.

 

Hopefully after a bit of expert analysis on the subs and calibration frames by John, things might be a bit different.  

 

In the meatime, I will try gain 0 and see what comes out.  Might take a while to accumulate the data as it’s not dark here till 11pm and wether is very unpredictable!



#9 liors

liors

    Mariner 2

  • *****
  • Posts: 212
  • Joined: 06 Aug 2011
  • Loc: Israel

Posted 25 May 2019 - 02:34 PM

Hi There,

 

First let me say that I never imaged with cooled CMOS sensor and honestly I never felt the need for those cameras, I imaged with DSLR`s and CCD`s, BTW- I own an SX Trius 694 which is the same sensor as you have in the Atik and I`m very happy using this camera with great results.

 

Let me share with you a bit of what I know about the CMOS sensor compering with CCD,

 

CMOS works with 14bit vs CCD 16bit, not sure if I`m right here but I heard that calibration and stacking programs software works better with 16bit because they designed for it, again I`m only guessing.
With CMOS 14bit sensor you can`t put a zero value for Bias frames vs 16bit CCD that you can.

 

With CMOS sensors you have to set the correct Gain and Offset manually vs CCD that it`s been set by manufacturer, some imagers will choose to use what it calls Unity Gain which convert one photon to one electron, this is a good way to start but not good to use it as a default set for all imaging targets which some are brighter and some are fainters, for example, if I`m using a DSLR Canon 6D and I`m choosing iso 640 which is the closest to Unity Gain, first I`ll need to increase exposure time and will get a higher read noise, if I choose iso 1600 instead than I can decrease exposure time and get a lower read noise.

 

CMOS sensors have a high QE and very low read noise, perfect for NB filters, some of the Sony sensors like 460ex, 490ex have a high QE and low read noise as well, this is why I like those sensors, the problem with the mono Sony sensors are that they come in small sizes, Sony has other sensors which they slightly bigger then the crop DSLR`s sensors, the also have a relatively high QE but they all comes in color versions like QHY8,10,12 and still you need to set Gain and Offset manually.

 

Kodak CCD sensors comes in varies sizes from very small to very big like the 16803 sensor, the reason I`m not using those cameras are due to their high read noise and force me to increase exposure time dramatically, most of those sensors are mono if not all of them.
CMOS sensors also comes in varies sizes from small to very big like the QHY367 which has the Nikon 810 sensor, again, the problem with CMOS sensors that a few of them are mono and most of them color sensors.

 

At last let me tell you that both sensors you have are very good, the Asi1600 mono sensor size is very close to Kodak KAF8300 size, if you choose to continue imaging with your Asi1600 you must learn how to use a 14bit camera properly and to find the optimization of Gain and Offset, I also recommend not to stick to Unity Gain as a default for all imaging targets.

 

Good luck!!!


  • Akwilliams likes this

#10 SXBB

SXBB

    Vostok 1

  • -----
  • Posts: 101
  • Joined: 05 Jul 2015

Posted 25 May 2019 - 02:54 PM

FPN = Fixed Pattern Noise - Becomes apparent at lower gain settings but  can be eliminated with aggressive dithering.

 

APP = Astro Pixel Processor - An alternative Image processing application.

 

I recently purchased an ASI1600 and as soon as I get a good run of weather & open skies (Very Soon I Hope!), I'll be able to render an opinion.  I have used my QHY168C and enjoyed it compared to my STT8300m but OSC CMOS vs Mono CCD isn't easy to compare.

 

OK, Actually I love the QHY168C...  I'll be testing out the ASI1600mm Pro next week under dark, Arizona skies!

 

Best,

 

Bruce


  • psandelle and moonrider like this

#11 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 25 May 2019 - 03:17 PM

It'll be good to hear your opinions.   I think i should point out, that my sky on a typical clear night is around Bortle 4



#12 stargzr66207

stargzr66207

    Ranger 4

  • -----
  • Posts: 336
  • Joined: 28 Aug 2014
  • Loc: Kansas, USA

Posted 25 May 2019 - 03:55 PM

Bruce,
Thanks for the translation on FPN and APP. It's hard to keep up with all the acronyms used
in digital imaging and processing!!

Ron Abbott

#13 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23069
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 25 May 2019 - 04:37 PM

John, you have been a world of help now and previous.   What I will do is upload the full set of subs, darks and flats and you can take a look for yourself and see if the data is good or there might be something amiss.

 

Regarding dither,  I am currenty set in SGP for High Dither every 3 frames.

 

Ill post a link up shortly

So I would say, dither every frame. I think every 3 frames may be limiting how well you can deal with any remnant FPN and quantization error stuff.

 

If you also increase your exposures, say 1.5-2x longer than you are using now, or even drop the gain and use 300-600s subs, you should be able to dither every frame without much loss to overhead. You are not stacking nearly enough subs to be able to do "sparse dithering" at every 3 frames. For sparse dithering to work, you need to be stacking a LOT of subs, hundreds. I myself dither about every 2 subs if I'm stacking around 100, and every 3 subs only if I'm stacking 300 or so. If I stack less than 100, I dither every sub. So I think that your sparse dithering with shallow stacks could be part of the problem. (May not be the whole problem, but definitely part of it.)

 

Further...lets say you stack the same number of subs with the ASI1600 as you did with your CCDs. If you were using longer exposures with the CCDs, then 30x300, or 30x600, is a LOT more signal than 30x120 or 20x240. You would have 2.5 hours with 30x300, and only 1 hour with 30x120. Lower read noise lets you use shorter exposures, and may potentially allow you to bury the read noise deeper than with a CCD...however, if you do not integrate the same total amount of exposure, then you will never get results as good as you did with CCD. So, for 120s L subs to compare to the CCD in this case, assuming you were stacking 30 CCD subs, you would need to stack 75x120s L subs with the ASI1600. Similarly, if you stacked 30x600s subs with a CCD, that would be 5 hours of integrated exposure. You would need to stack 75x240s subs to get the same integration.

 

Total integrated exposure counts, and in fact is still and always will be king. CMOS's advantage over CCD (in most, but not all, cases as there are some very low read noise CCDs out there as well) is the ability to use shorter subs, but you then end up needing to stack more of them to get the same total signal as CCD. If you are stacking the same number of shorter subs, then you have less signal, and yes, that will mean lower SNR, higher noise, etc.


  • psandelle likes this

#14 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23069
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 26 May 2019 - 03:28 PM

Here is a link to subs and calibration frames.    LRGB Lights, Darks for L@120s - 15, Darks for RGB@240s -15 and a sample of flats for the filters

 

https://horsebridged...I4B09Q?e=NNrNfz

I am not seeing anything in your data that is inherently "bad" in any way. Looks pretty normal to me. I would be curious to know the answers to the questions I asked previously. Whether you are going for equal sub count or equal integration time. If the former, then that is probably why your ASI1600 results look worse...you would simply not have the same total signal and thus a lower SNR.



#15 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 26 May 2019 - 04:18 PM

Thanks for looking Jon, its certainly encouraging that the data is relatively normal.  

 

It would certainly be integration time to make a true comparison with what i have seen with my CCD's.     I have in total about 30x 120s with Lum and 20x 240s with RGB on M101..  I would probably need another hour on each filter to get to a comparison with data i have with a CCD.

 

If the data is good and even with the amount of subs i have on all filters, i would have thought i would be able to process at least something half decent.   I am still on a steep learning curve with regards to the gain and offset.   So i'll plug away at it for the time being.. but i am certainly more encouraged to keep with it than i was a few nights ago.

 

Just to add..  Its clear tonight, so all setup and just got an image run on the go.  Based on the fact that the images you have looked at look pretty normal, I'm going for M51 with gain at 76 and an offset of 50 @ 120s Lums.   The first images are looking quite nice.  Good detail.. much more so than M101.  Ill get as much as i can tonight - probably a couple of hours in total and see how things look initally and add more data to it as and when the skies allow.


Edited by Akwilliams, 26 May 2019 - 05:17 PM.


#16 Jon Rista

Jon Rista

    ISS

  • *****
  • Posts: 23069
  • Joined: 10 Jan 2014
  • Loc: Colorado

Posted 26 May 2019 - 05:50 PM

Thanks for looking Jon, its certainly encouraging that the data is relatively normal.  

 

It would certainly be integration time to make a true comparison with what i have seen with my CCD's.     I have in total about 30x 120s with Lum and 20x 240s with RGB on M101..  I would probably need another hour on each filter to get to a comparison with data i have with a CCD.

 

If the data is good and even with the amount of subs i have on all filters, i would have thought i would be able to process at least something half decent.   I am still on a steep learning curve with regards to the gain and offset.   So i'll plug away at it for the time being.. but i am certainly more encouraged to keep with it than i was a few nights ago.

 

Just to add..  Its clear tonight, so all setup and just got an image run on the go.  Based on the fact that the images you have looked at look pretty normal, I'm going for M51 with gain at 76 and an offset of 50 @ 120s Lums.   The first images are looking quite nice.  Good detail.. much more so than M101.  Ill get as much as i can tonight - probably a couple of hours in total and see how things look initally and add more data to it as and when the skies allow.

Honestly, try longer exposures. Longer at gain 76, and also at gain 0. Give 300s and 600s subs a try at both gains.



#17 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 26 May 2019 - 05:58 PM

Ok,  ill try those figures :-)



#18 ezwheels

ezwheels

    Ranger 4

  • -----
  • Posts: 393
  • Joined: 04 Jul 2017
  • Loc: Oakland, CA.

Posted 26 May 2019 - 07:19 PM

I am not seeing anything in your data that is inherently "bad" in any way. Looks pretty normal to me. I would be curious to know the answers to the questions I asked previously. Whether you are going for equal sub count or equal integration time. If the former, then that is probably why your ASI1600 results look worse...you would simply not have the same total signal and thus a lower SNR.

Agree with Jon's comments, but if you are not able to get the clear skies to test the difference with more CMOS data you could conversely compare the shorter combined CMOS data with less subs from the existing CCD stack so that the exposure times are more equal.


  • danielguo likes this

#19 Akwilliams

Akwilliams

    Mariner 2

  • -----
  • topic starter
  • Posts: 276
  • Joined: 27 Aug 2016

Posted 27 May 2019 - 05:09 AM

So, heres a sample of what i got last night before the clouds rolled in.

 

I tried the following: 

 

300s at Gain 0

600s at Gain 0

300s at Gain 76

600s at Gain 76

 

I forgot that i had changed the offset to 21 - I had meant to test at offset 50.

 

Just dong a visual of Gain 0 vs Gain 76 on 600s - The Gain 76 looked by far the best.  Infact, i was pretty impressed.  I never thought that being able to push 600s exposures with this camera where i image from would be possible.   But it think the proof is in the pudding.   I only managed a couple of L's a couple of Rs and 1 each of G & B before i had to stop. 

 

Here's a sample of each

https://horsebridged...ZqMALQ?e=MMMMCE

 

So, I will continue with M51 doing 600s subs at Gain 76/21 and see what comes out.  


  • Jon Rista likes 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






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics