12bit vs 10 bit test

You will find differences between 10 and 12 bit log, if the camera records any, in the sparse (not overexposed) highlights.
If you record slog (i.e. slog1) then I am not sure there is enough dynamic range in the curve to show any differences.
I would use slog2 (or slog3).

The whole point of this was to push the exposure to see where it would fail, how much we could recover, and whether there was any difference between 10 bit and 12 bit in terms of recovery.
I am not saying you are implying that, but 12 bit log has no higher dynamic range than 10 bit log, a higher bit depth simply smoothens the granularity of the curve.
In the case of log the region to look at would be the highlights, in the case of linear this would be the shadows.
 
Last edited:
You will find differences between 10 and 12 bit log, if the camera records any, in the sparse (not overexposed) highlights.
If you record slog (i.e. slog1) then I am not sure there is enough dynamic range in the curve to show any differences.
I would use slog2 (or slog3).


I am not saying you are implying that, but 12 bit log has no higher dynamic range than 10 bit log, a higher bit depth simply smoothens the granularity of the curve.
In the case of log the region to look at would be the highlights, in the case of linear this would be the shadows.

Great insight. F35 only has slog (slog1)
 
Does anyone see any difference in color on the 12 bit files?

I downloaded them and quickly graded them and I would say no I don't really see any difference. I can tell on the very over exposed ones that 12 bit allows me to bring a bit more back, but they are so far gone that they are not useable even if the 12 bit is a little better. For the ones that are over exposed but salvagable I didn't see any real benifit. Same holds true for under exposed.
 
This is one of the great things about the F35. Light weight 10bit files look amazing, because they are already in the right place. With other cameras, 10bit is generally pretty healthy, but enough grading takes place that the image begins to break up a little. I was doing some custom LUT testing with the F55. Wasn't able to perfect it, but I am thinking of shooting with a burned in custom out that I don't have to push so hard in the grade. the difficult thing about this is the work flow is so custom, it is difficult to collaborate or use in a variety of environments.

But the F35 gets it correct, right out of the gate, so all the LUT and this or that, I never have to think about, just go shoot.
 
I really like this test. it is the ultimate torture test for any camera! Was great to see that there are only nearly imperceptible differences between 10 and 12 bit. Since, I really love shooting in 10bit, just a lighter more elegant file. I do see that in certain cases, I would opt for the 12bit 444. especially in color grading where making clean color selections is important. But for a lot of what I am doing currently, I am very excited about not just the image quality, but the feel of the images I am getting.
 
Which is which? I pushed the midtones in darktable and cropped in darktable. It is evident that the waveform from 12 bit has more spikes than the one from 1O bit, this would mean more color definition, in other words more colors. But is this viewable on my 8bit monitors, or even can everybody's eyes see the difference, if there is one? Interesting test. The pictures are not exactly the same, maybe folliage at dusk, not on a windy day, would be more revealing.

dfgd.jpgfgsd.jpg
 
I really like this test. it is the ultimate torture test for any camera! Was great to see that there are only nearly imperceptible differences between 10 and 12 bit. Since, I really love shooting in 10bit, just a lighter more elegant file. I do see that in certain cases, I would opt for the 12bit 444. especially in color grading where making clean color selections is important. But for a lot of what I am doing currently, I am very excited about not just the image quality, but the feel of the images I am getting.

I agree 100%. I have been hearing about the sweetness of shooting 444 12 bit files out of the F35... but honestly, whatever the camera is doing is magic, because the difference is imperceptible. Great camera right out of the box. Colors are spot on. If I wanted to key, well maybe I would shoot 12 bit... honestly, I hardly ever do :smile:

Macgregor told me today, that you can also get 10 bit 444 out of the F35 with dual link. And yes, there is... but how to record it? Just set up my menu on the F35 to do some tests. Here is the proof. 10bit 444 may be a great compromise with the best color fidelity straight out of the sensor without chroma subsampling and with the 10bit luma ramp that satisfies most of us... Will post some more tests on a different thread. I think the idea would be to place several items evenly illuminated in tungsten and check color reproduction. Might be the internal camera processing gets 422 = 444... push it in the grade and see how many nodes before s*%t starts to fall apart.
Any ideas on how to approach this?

But how to record 10bit 444... my recorder does ProResHQ 10bit 442 or ProRes444 12 bit... am I missing something?

Sony F35 - Recording Modes.jpg
 
Which is which? I pushed the midtones in darktable and cropped in darktable. It is evident that the waveform from 12 bit has more spikes than the one from 1O bit, this would mean more color definition, in other words more colors. But is this viewable on my 8bit monitors, or even can everybody's eyes see the difference, if there is one? Interesting test. The pictures are not exactly the same, maybe folliage at dusk, not on a windy day, would be more revealing.

View attachment 131825View attachment 131826

Yeah man. I will try more foliage at dusk, but did you download the footage in the dropbox? There is foliage there shot on purpose as greens tend to always be the most problematic.
 
Back
Top