S-Log and 10bit Recording

[Steve Kalle]

I have read that 6db of sensitivity is equal to 1bit and 10bits requires a camera with at least 60db. Sony states 63db for the F3 but no test has come close to that. So, lets say its at least 60db at 0db gain. With the S-Log upgrade, the F3 loses 6db of sensitivity when outputting 444. Under the best circumstances, the F3 then can only output 9bits at 54db or as low as 48db if some tests are to be believed which is only 8bits.

Please correct me if I am wrong in my understanding.

So, is normal 9-10bit 422 better than 8-9bit 444 S-Log? Think of it this way: with S-Log, you lose 1bit from the entire image. Plus, more bits are being used in the shadows and highlights; thus, you lose even more bits for the other 75% of the image. (or whatever percentage to which it equates)

[Nate Weaver]

Steve,

I would not bother with trying to be too scientific about it.

The end result is that when recording 10-bit, the image is much more elastic in color correction before you see posterization if you bring black levels and gamma up and down. 8 bit tends to not tolerate a lot of manipulation that way, 10 bit makes a huge difference.

Even if you shot +18 gain, and the noise floor has raised all the way into your "8 bit area", so to speak, you will be able to manipulate that entire tonal range, noise and all, much better without introducing posterization in the blacks that will really make that noise much, much more distracting and ugly.

It really is a revelation the first time you start manipulating a 10 bit source if all you've ever had is 8 bit. It was a real eye opener for me. 4:4:4 on the other hand, I've only ever seen the benefits of while watching other people do keying on green screen I've shot. Nice for sure, but for my dollar, 10 bit is a much bigger deal.

[Steve Kalle]

Thanks Nate. You are just the person I was hoping to respond due to your experience with Resolve (which I am currently learning through FXPHD).

What is your thought on S-Log encoded video versus normal 10bit 422 when it comes to grading? What 3D LUT do you usually apply to S-Log footage or is it better to use curves? While I am learning Resolve on Mac, I am applying what I learn to After Effects, which also supports LUTs. Personally, I love the flat look of S-Log and Red footage; so, I usually have my black and black gamma raised in the PP settings so I can get the exact look I want in post.

Last question to you for now: what is your experience with footage from an EX1/3 + nanoFlash and how do you get around the high amount of noise in their images when grading?

Thanks

[Timur Civan]

Nate is 100% correctamundo.

the 10 bit image, ( a clean, sharp one at that...) is the real holy grail here. The sLog will just give you some extra highlight protection, and in a 10 bit recording stream, its actually useable.


Imagine a camera that records 12 stops to 8 bit XDcam hd, (cough* F3).....

Thats is 256 steps of grey to represent 12 stops of brightness. divide..... and you get 21 shades of grey per stop. Thats asking for trouble in grading, and post.

[Steve Kalle]

Timur,

That is certainly the best way to approach this issue.

I thought S-Log also helped add detail to shadows whereas normal gamma modes usually crush the blacks. At least, this is what the Abel Cine blog showed in comparing S-Log. From what I have seen, I consider S-Log to look much like Red in that they are flat.

[Alister Chapman]

The reason why 10 bit is crucial for S-Log is because the amount bits of data used to record the signal decreases logarithmically with image brightness. Most of the data range is allocated to shadow and midtones. So if you only have 8 bits your highlights will posterise when you apply the LUT due to a lack of bit depth.

10 bit brings little advantage on noisy cameras with conventional gamma curves because the noise will normally limit how far you can push the image before posterisation effects become apparent. In a noisy camera the noise will modulate every part of the image. If the 8 bit codec can resolve the amplitude of the noise then increasing to 10 bits will only bring a very marginal improvement in image quality as in essence all your are doing is recording the noise with greater accuracy. What can often be more important is the bit depth of the in camera A to D and processing as if this is done at a low bit depth it will introduce banding effects that may not be visible in an unprocessed image but quickly become visible when graded. Very often this is mistaken as the difference between 10 bit and 8 bit outputs as cameras with 10 bit outputs can often have better A-D's than cameras with 8 bit outputs.

[Aaron Newsome]

Also I'll point out that you can already record 10bit 422 without upgrading the firmware.