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

My test of 16 bit versus 8 bit captures. Images from 11/2/2024

  • Please log in to reply
18 replies to this topic

#1 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 04 November 2024 - 11:09 AM

I wanted to check for myself to see if capturing in 16 Bit (actually the Saturn-M SQR only goes to 14 Bit) versus 8 Bit is worth the effort in handling slower fps and files twice the size in terms of contrast, better representation of structures,  etc.  My main focus is animations so this also equates to longer processing time to handle the much larger SER files.  The disks were imaged on 11/2/2024 and were 1 minute apart so of course there could have been some variations in seeing during the two captures.  Both were processed the exact same way using Autostakkert to align and stack the best 15% of 500 frames taken at full resolution with the Lunt LS60T DS and Saturn-M SQR with a Meade 2X Shorty Barlow.  Both disks were then processed in Photoshop 2025 with an action to ensure the processing was identical.  My conclusion is that for animations, recording in 16 bit does not offer enough difference from 8 bit to warrant the effort of handling larger files, longer processing time, etc.  I can see just a slight improvement in overall representation of structures but not enough to change since changes in seeing during captures for animations will most likely reduce the benefit of the much faster fps and smaller files to work with.  Of course, ymmv!  Using 16 bit (or whatever your particular camera actually captures) for a still image makes sense, although you still have to accept reduced fps which might be quite slow for some cameras.  Luckily, the Player One 533 chip can run at nearly 20fps at full resolution in 16 bit and 42 fps in 8 bit.

You can look at the full resolution images to really pixel peek between the two.

 

LInk to Astrobin

 

get.jpg?insecure


  • BYoesle, Mark Strollo, apo130max and 11 others like this

#2 spicerack0

spicerack0

    Explorer 1

  • -----
  • Posts: 92
  • Joined: 30 Sep 2024

Posted 04 November 2024 - 02:49 PM

  For what its worth,  when i boost the gamma on your image with no other changes, the black background is noticeably darker on the left image.   Thats a huge file btw  lol 6000x3000 resolution

Untitledb.jpg


Edited by spicerack0, 04 November 2024 - 02:54 PM.

  • BYoesle, rigel123 and C0rs4ir_ like this

#3 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 04 November 2024 - 02:56 PM

  For what its worth,  when i boost the gamma on your image with no other changes, the black background is noticeably darker on the left image.   Thats a huge file btw  lol 6000x3000 resolution

attachicon.gif Untitledb.jpg

Interesting.  Yeah, I left the full resolution on AB for those that really like to pixel peek.



#4 StargazerLuigi

StargazerLuigi

    Surveyor 1

  • -----
  • Posts: 1,956
  • Joined: 19 Apr 2014
  • Loc: New London Twp, Pennsylvania, USA

Posted 04 November 2024 - 05:07 PM

Thanks for the comparison. I've not done any animations (yet), but I've been doing my captures in '16-bit' mode (I have the same camera as you do, so really 14-bits). That's 18 gigs of disk space per capture (1000 frames), which adds up quickly. Eventually I delete the ser files.

 

With my setup that limits me to 20fps at most, which I've not found to be a problem.


  • rigel123 likes this

#5 NuovaApe

NuovaApe

    Mariner 2

  • -----
  • Posts: 286
  • Joined: 07 Aug 2021
  • Loc: England

Posted 04 November 2024 - 08:52 PM

Hi Warren - nice Go Compare! Maybe that advert is not on US TV.

 

Most end user displays are 8 bit RGB (8,8,8) anyway.

 

My workstation leisure station is plugged into my 55" OLED TV.

OLED panels are 10 bit RGB and I can see the benefits of "16bit" captures vs 8bit.

 

But when reduced back down to 8bit for public consumption I question myself whether it was worth the effort.

 

It can be in tricky situations. An extra 4bits can make a difference. Whispy faint proms mostly. Instead of the last bit flipping 0/1 it can be an extra 4bits flipping 0..15. That can reduce quantization markedly. But it's rarely useful TBH.

I've not seen the sun for 2 weeks in UK. The most boring layer of blanket cloud you've ever seen.

I'd be happy with a 1bit image right now smile.gif

 

Regards, Ed.


  • BYoesle, rigel123 and C0rs4ir_ like this

#6 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 05 November 2024 - 10:57 AM

Great discussion.  From this I've decided to continue capturing animations in 8 bit and may experiment a bit with 16 (14) bit for my full disk shots.


  • LarryAlvarez and NuovaApe like this

#7 Vansh

Vansh

    Sputnik

  • *****
  • Posts: 37
  • Joined: 30 Oct 2023
  • Loc: Houston, TX

Posted 06 November 2024 - 04:17 PM

Hmm how much would 16-bit vs 8-bit matter in terms of capturing the prominences (much fainter) as well as the full disk (much brighter) at once? I'm debating if it's worth going 16 bit to brighten the prominences afterward without losing detail in their brightness.



#8 StargazerLuigi

StargazerLuigi

    Surveyor 1

  • -----
  • Posts: 1,956
  • Joined: 19 Apr 2014
  • Loc: New London Twp, Pennsylvania, USA

Posted 06 November 2024 - 04:27 PM

The downside to going with 16-bit is larger capture files and slower frame rates. I would rather have the data and only throw away the extra bits at the last possible moment. Five years from now we could all have 16-bit displays and 16-bit capture will be child's play for USB-9.

 

Teasing out the fainter prominences is all in post, at least in my case.


  • rigel123 likes this

#9 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 06 November 2024 - 05:03 PM

Hmm how much would 16-bit vs 8-bit matter in terms of capturing the prominences (much fainter) as well as the full disk (much brighter) at once? I'm debating if it's worth going 16 bit to brighten the prominences afterward without losing detail in their brightness.

The key to capturing the prominences as well as the disk is to make sure you are not clipping the black point during capture.  As long as the histogram is not buried against the left side, then the data is there to pull out the proms by raising shadows.  Reducing the amount of photosphere that shows through the chromosphere by double stacking helps to balance the brightness of the disk with the proms but there are processing steps that you can take to bring out proms with the surface even when single stacked.


  • BYoesle and groom like this

#10 ch-viladrich

ch-viladrich

    Surveyor 1

  • -----
  • Posts: 1,692
  • Joined: 14 Jul 2013
  • Loc: France

Posted 07 November 2024 - 08:08 AM

Hmm how much would 16-bit vs 8-bit matter in terms of capturing the prominences (much fainter) as well as the full disk (much brighter) at once? I'm debating if it's worth going 16 bit to brighten the prominences afterward without losing detail in their brightness.

Indeed, with the PlayerOne IMX533, I have found out that the signal/noise ratio for faint prominences in Ca K is better with 14-bit acquisition compared to 8-bit ...provided the seeing is good.

If the seeing is not good enough, then 8-bit acquisition is better (for faint prominences) because of faster fps.

I have some pictures showing this very clearly. Just need to take the time to put this on a web page.

 

Otherwise, 8-bit is better (or equal) for the disk imaging (because of larger fps).

 

The PO IMX533 is the only camera I have found this results. For all other cameras I have been using, I've found no difference between 8-bit and 12-bit acquisition (except for such extreme targets as solar total eclipses and the earthshine, which have a very large dynamic range).


  • BYoesle, rigel123 and KMH like this

#11 HPaleske

HPaleske

    Surveyor 1

  • *****
  • Posts: 1,666
  • Joined: 09 Apr 2007

Posted 08 November 2024 - 05:52 AM

Hello Warren,

I can confirm what Christian said about the CaK shots in Halpha. All my Halpha images were made with 16Bit in FC (fps 110). As soon as I switch to 8 bit I lose information in the subtle tonal values and ugly double lines appear. Interestingly, you often see this on images taken by advanced Halpha amateurs. These structures can be seen above all at the edge of the chromosphere or at the hard contrasts between the umbra and the surroundings.
However, I have not yet seen any such artifacts in Christian's halpha images, which speaks for good image processing.  

 

Theoretically, this should also be visible in your recording set. However, I suspect that the focal length is not high enough to see this.

 

 

cs Harald

www.unigraph.de


  • rigel123 and C0rs4ir_ like this

#12 C0rs4ir_

C0rs4ir_

    Viking 1

  • *****
  • Posts: 997
  • Joined: 12 Sep 2021
  • Loc: Germany

Posted 08 November 2024 - 06:18 AM

Hello Warren,

I can confirm what Christian said about the CaK shots in Halpha. All my Halpha images were made with 16Bit in FC (fps 110). As soon as I switch to 8 bit I lose information in the subtle tonal values and ugly double lines appear. Interestingly, you often see this on images taken by advanced Halpha amateurs. These structures can be seen above all at the edge of the chromosphere or at the hard contrasts between the umbra and the surroundings.
However, I have not yet seen any such artifacts in Christian's halpha images, which speaks for good image processing.  

 

Theoretically, this should also be visible in your recording set. However, I suspect that the focal length is not high enough to see this.

 

 

cs Harald

www.unigraph.de

Hello Harald,
I think the noise level of the aquired data is also a factor. The usual h-alpha filtering methods need to have the gain raised significally. In this case a jump from 8 to 12 bit might not have a big impact on the achievable dynamic range. This might change if one is recording data using only very low gain.

 

What gain levels are you using with the Apollo Max camera? I assume they are very low.

 

Regards

Dennis



#13 HPaleske

HPaleske

    Surveyor 1

  • *****
  • Posts: 1,666
  • Joined: 09 Apr 2007

Posted 08 November 2024 - 07:02 AM

Hello Dennis,

could be. I have not yet done this test with increased gain values and 8 bits. My recordings are usually always made with a gain of 70-130 in FC. These ugly gradients always occur with this setting. They can be reduced in the dark sky background by an additional noise effect. Unfortunately, I can't remember when I did these tests. I'll take a sample image the next time I observe (it may take a while!).

 

cs Harald

www.unigraph.de


  • C0rs4ir_ likes this

#14 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 08 November 2024 - 07:12 AM

Hello Warren,

I can confirm what Christian said about the CaK shots in Halpha. All my Halpha images were made with 16Bit in FC (fps 110). As soon as I switch to 8 bit I lose information in the subtle tonal values and ugly double lines appear. Interestingly, you often see this on images taken by advanced Halpha amateurs. These structures can be seen above all at the edge of the chromosphere or at the hard contrasts between the umbra and the surroundings.
However, I have not yet seen any such artifacts in Christian's halpha images, which speaks for good image processing.  

 

Theoretically, this should also be visible in your recording set. However, I suspect that the focal length is not high enough to see this.

 

 

cs Harald

www.unigraph.de

It makes sense to me that you would see a difference in your high resolution images when imaging at the higher bit level.



#15 HPaleske

HPaleske

    Surveyor 1

  • *****
  • Posts: 1,666
  • Joined: 09 Apr 2007

Posted 10 November 2024 - 05:21 AM

Hello,
I have now looked in the archive and found a comparison image from 8 bit to 16 bit. Unfortunately, I have deleted the exact settings in FC. However, I will make similar comparisons again during the next observations.

You can clearly see the gradients in the 8-bit image, which are caused by the lower information. I can't say whether these disappear when the gain is turned up (e.g. to 170 to 200).

Both images were processed immediately without tonal value corrections.

 

GIF animation 8 bit vs 16 bit (Firecapture):

 

https://unigraph.de/...lapl5_ap546.gif

 

 

cs Harald

www.unigraph.de


  • C0rs4ir_ likes this

#16 C0rs4ir_

C0rs4ir_

    Viking 1

  • *****
  • Posts: 997
  • Joined: 12 Sep 2021
  • Loc: Germany

Posted 10 November 2024 - 05:37 AM

Thanks Harald, that is interesting.



#17 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 10 November 2024 - 06:18 AM

Hello,
I have now looked in the archive and found a comparison image from 8 bit to 16 bit. Unfortunately, I have deleted the exact settings in FC. However, I will make similar comparisons again during the next observations.

You can clearly see the gradients in the 8-bit image, which are caused by the lower information. I can't say whether these disappear when the gain is turned up (e.g. to 170 to 200).

Both images were processed immediately without tonal value corrections.

 

GIF animation 8 bit vs 16 bit (Firecapture):

 

https://unigraph.de/...lapl5_ap546.gif

 

 

cs Harald

www.unigraph.de

Hmm, to me it looks like details are actually better in the 8 bit or it could be seeing was just better at the moment the 8 bit was captured.



#18 HPaleske

HPaleske

    Surveyor 1

  • *****
  • Posts: 1,666
  • Joined: 09 Apr 2007

Posted 10 November 2024 - 07:01 AM

Hello,

 

Warren,
I don't think so, as the time difference between the two recordings is only 228 seconds. One can assume that the transparency conditions were approximately the same. I don't think the exposure time was changed either.

 

For me, the 16-bit image looks much better, with more information, especially in the low-contrast areas. You can hardly tell the difference on the surface. However, this effect also occurs with Umbren.

 

It could also be that the stack program cannot cope with the large amount of 16-bit information and therefore produces blurring. This should therefore not occur with good seeing.

 

Perhaps the recording time was also shorter ( fps~ 160). I don't know anymore. I usually have 2000 pictures taken. At fps 100 that's 20 seconds, at fps 160 that's 12.5 seconds. 8 seconds could be a lot even at high resolution.

 

It almost looks as if you can look deeper into the chromosphere with 8 bits.

 

 

cs Harald

www.unigraph.de


Edited by HPaleske, 10 November 2024 - 07:43 AM.


#19 rigel123

rigel123

    ISS

  • ****-
  • topic starter
  • Posts: 26,490
  • Joined: 29 Jun 2009
  • Loc: SW Ohio

Posted 10 November 2024 - 08:45 AM

Hello,

 

Warren,
I don't think so, as the time difference between the two recordings is only 228 seconds. One can assume that the transparency conditions were approximately the same. I don't think the exposure time was changed either.

 

For me, the 16-bit image looks much better, with more information, especially in the low-contrast areas. You can hardly tell the difference on the surface. However, this effect also occurs with Umbren.

 

It could also be that the stack program cannot cope with the large amount of 16-bit information and therefore produces blurring. This should therefore not occur with good seeing.

 

Perhaps the recording time was also shorter ( fps~ 160). I don't know anymore. I usually have 2000 pictures taken. At fps 100 that's 20 seconds, at fps 160 that's 12.5 seconds. 8 seconds could be a lot even at high resolution.

 

It almost looks as if you can look deeper into the chromosphere with 8 bits.

 

 

cs Harald

www.unigraph.de

I'm thinking you are on the right track about the slower fps at 16 bit as well as the longer time to capture those 2000 frames causing some blurring.

 

I agree about the 8 bit appearing to look deeper into the chromosphere.  Maybe if you tweaked Gain up a little on the 16 bit you could reduce the exposure a bit, but it would take a lot to make up for the slower fps.  I think it all comes down to compromises between frame rate, amount of data collected, and visual appeal.


  • BYoesle 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