So I have been playing around a bit with 12 bit and 10 bit dpx's off the F35 and I have yet to be able to see any benifit of the 12 bit. So far 10 bit has been as solid as 12 bit. Have any of you encountered a situation where the 12 bits was an advantage over 10 bit? So far I have been shooting 12 bit just to have the highest quality available but one of the quirks with the Codex I am using is that you have to disconnect it from the camera to switch from 10 bit to 12 bit which is a little annoying at times.
12 bit v 10 bit Opinions
- Thread starter dhessel
- Start date
ProfessorU
Veteran
In theory, I'd use 12 in very low contrast situations, and for greenscreening and effects. But I can't say that I've done side-by-sides to know for sure it would make a difference.
When I had my FS700 with the 7Q 12-bit RAW recorder, I converted the same shot into 10 bit 422 ProRes HQ and 12 bit 4444 ProRes.
I then did the same extreme grade (nothing I would normally do, but I tried to break the codec) to both the 10 bit and 12 bit footage, and there was very, very little difference if any between the two clips.
I then did the same extreme grade (nothing I would normally do, but I tried to break the codec) to both the 10 bit and 12 bit footage, and there was very, very little difference if any between the two clips.
I am spitballing off some anecdotal experiments I've made but I believe you'll get better results with de-noise on a 12bit image. NeatVideo does some voodoo **** with a dense negative. Overall even in low contrast mildly underexposed images I find myself liking the F35 grain too much to bother denoising as it helps not only retain detail but breaks up banding, etc. Why re-grain when you keep grain. Definitely something I agree with JMtheDP on. The in-cam noise reduction is good in a pinch if you're baking your gamma and sending your footage off to never never land but I'll keep all that nonsense off if I have full control.
I think a better question is, why not use 12-bit?
With storage so cheap you really have nothing to lose.
It definitely makes a difference for grading, keying, banding prone situations and scenes with extreme contrast.
The F35 is (conservatively) marketed by Sony as a 12-stop camera. 12-bit technically helps ensure that you are getting all those 12-stops. The Genesis for example is limited to 10-bit, which is fine since it has around 10-stops of dynamic range.
For anything that I care about I shoot 12-bit.
With storage so cheap you really have nothing to lose.
It definitely makes a difference for grading, keying, banding prone situations and scenes with extreme contrast.
The F35 is (conservatively) marketed by Sony as a 12-stop camera. 12-bit technically helps ensure that you are getting all those 12-stops. The Genesis for example is limited to 10-bit, which is fine since it has around 10-stops of dynamic range.
For anything that I care about I shoot 12-bit.
We had an in house demo of one of Sony's new midrange LCD laser projectors a week or so ago. It could resolve value scales higher than 10 bit. 16 bit tiff stills looked amazing. I could see minor grading issues with my footage on that projector that would not show up on a TV or computer monitor, even a 10 bit one. Wish I could afford one for a grading theater.
Digital encoding is just a numeric scale applied to a defined analog value range. It is the analog value range that determines DR, not the digital bit depth, which only affects how well resolved deep shadow values are at the low end of the range for linear raw recording. and not even that for gamma encoded video. Bit depth has no more effect on DR than measuring centimeters vs millimeters has on the length of a 1 meter ruler.
combatentropy
Veteran
10 bit should be enough, especially if it's log.
The more bits, the less banding. But all digital pictures have banding, even 100-bit (if there were such a thing). But like resolution, the human eye can tell only so much.
For banding, you're safe if each f-stop has at least 70 levels, or bands (maybe even just 30). Note that it depends on the f-stops of the display, not the original scene before capture. I believe most TV screens top out around 7 stops at full brightness --- and who would want a much brighter TV? "Our eyes! it hurts us!" (Gollum (not really)). I believe movie screens were once measured at 9 stops (?).
Anyway, 9 x 70 = 630 (9.5 bits, if spread logarithmiically, not linearly). And that's best case. More like 6 stops for 99% of your viewers (6 x 70 = 420), and more like 30 steps, not 70 (6 x 30 = 180, less than 8 bits). And in the darker stops, less than 30 are really needed. All this is why 8-bit is usually good enough.
Now if you plan to grade the heck out of your image, you'll want more than the 7 or 8 bit minimum, because you might be moving bands around from one stop to another (like if you apply a strong S curve). Still, 10-bit log gives you 1,024 levels. For a 7-stop display that's more than twice as many bands as you need per stop.
The more bits, the less banding. But all digital pictures have banding, even 100-bit (if there were such a thing). But like resolution, the human eye can tell only so much.
For banding, you're safe if each f-stop has at least 70 levels, or bands (maybe even just 30). Note that it depends on the f-stops of the display, not the original scene before capture. I believe most TV screens top out around 7 stops at full brightness --- and who would want a much brighter TV? "Our eyes! it hurts us!" (Gollum (not really)). I believe movie screens were once measured at 9 stops (?).
Anyway, 9 x 70 = 630 (9.5 bits, if spread logarithmiically, not linearly). And that's best case. More like 6 stops for 99% of your viewers (6 x 70 = 420), and more like 30 steps, not 70 (6 x 30 = 180, less than 8 bits). And in the darker stops, less than 30 are really needed. All this is why 8-bit is usually good enough.
Now if you plan to grade the heck out of your image, you'll want more than the 7 or 8 bit minimum, because you might be moving bands around from one stop to another (like if you apply a strong S curve). Still, 10-bit log gives you 1,024 levels. For a 7-stop display that's more than twice as many bands as you need per stop.
Last edited:
Why not? Here are the main reasons.
1.) I have the Codex with mine which is nice it also means that I can shoot vari-frame or any frame rate from 1 to 50 FPS, but as far as I can tell the Codex only supports 12 bit upto 30P and one of the side effects of shooting 12 bit on the Codex is in order to switch from 10 bit to 12 bit, you have to power down the Codex which takes at least 20 seconds, power of the F35, remove the Codex, Power on the F35, switch to 12 bit, power down the F35, put the Codex back on, power up the F35, power up the Codex which again takes at about 20 seconds, switch the Codex from 10 bit to 12 bit. If I only shoot 10 bit I think I might be able to just leave the F35 and Codex in vari-frame mode and switch FPS and speed ramp at will.
2.) 12 bit dpx files are odd in that the red and blue channels are flipped, so you always have to swap red and blue channels when working with them. 10 bit doesn't have this issue.
3.) 12 bit ProRes 4444 is not supported on windows without 3rd party codec and Resolve doesn't work with 3rd party codecs, again 10 bit ProRes 4444 is fine.
10 bit is plenty to hold 12 stops of DR with a log profile and if I am not mistaken the slog on the F35 was specifically designed to store 12 stops in a 10 bit file. So far I have been shooting 12 bit to have the hightest quality negative I can but am wondering if its worth the extra qwirks.