FS5: Does Slog3 clip at 94% in HD with the fs5?
- Thread starter lagmandokk
- Start date
FrameFarmMedia
Veteran
I was told it clips at 94 because the Slog3 curve is designed for 16 stops of DR to be used with future sensors also. Don't know it's true but it makes sense when you look at the curve.
cameramanben
Well-known member
Donny,
a Sony MLUT can output any value between 0% and 109%, but will only ever be fed S-Log3 values up to 94% (equivalent to S-Log2 values up to 109%).
FrameFarmMedia -
[*** danger! more geekery ***]
there is an assumption that that may be the case (I've suggested that myself), and when you compare to C300 C-Log that one clips at 109% IRE for 5 1/3 stops so the only way to allow for significantly more highlight range was by completely changing the curve, ie C-Log2. Equally, S-Log2 taps out at 109% for six stops over. S-Log3 would allow for an extra 2/3 stop or so headroom if taken to 100%.
However, S-Log2 is identical to S-Log underexposed by half a stop which creates an extra half stop of headroom. As sensors got quieter in the shadows, the difference it made to detail below mid gray was sufficiently small to make the shift worthwhile without completly rethinking the log curve (ie something like S-Log3).
My main suspicion is around Arri LogC. I've always pointed out that S-Log3 is very similar to LogC at Arri's default ISO of 800. I took LUTCalc's LogC calculations from an Arri white paper, which covered how - uniquely - Arri changes the parameters of LogC with ISO, so that black to mid gray is always 10-bit 95 to 420, but as the ISO is increased the highlight contrast decreases so that nothing is ever clipped. I curve fitted my own calculations for the parameter values between those given by Arri. This is in contrast to Canon C-Log which keeps all reflectance values consistent but clips shadows or highlights as the ISO shifts and Sony which consistently records the full dynamic range along to avoid clipping but with a metadata value for the CineEI ISO with the expectation of sorting it out in post.
Anyway, I've just been revisiting the Arri calculations in LUTCalc as it occurred to me that the Arri ACES IDT is the definitive source. Looking at that I discovered that Arri treats ISO 400 to be the nominal ISO for LogC. Using that, S-Log3 is near identical to ISO400 LogC - basically just overexposed 1/3 of a stop (Arri cameras have around 7 1/3 stops headroom at ISO 800, so about 6 1/3 at ISO 400 compared to Sony's 6 stops at native ISO).
As such my thought is that the S-Log3 levels have been consciously chosen to make working with it feel familiar to those used to handling LogC.
As I mentioned in a previous post, that is nothing more than an educated guess but it does make reasonable sense. ;-)
[*** end of geekery ***]
Ben
a Sony MLUT can output any value between 0% and 109%, but will only ever be fed S-Log3 values up to 94% (equivalent to S-Log2 values up to 109%).
FrameFarmMedia -
[*** danger! more geekery ***]
there is an assumption that that may be the case (I've suggested that myself), and when you compare to C300 C-Log that one clips at 109% IRE for 5 1/3 stops so the only way to allow for significantly more highlight range was by completely changing the curve, ie C-Log2. Equally, S-Log2 taps out at 109% for six stops over. S-Log3 would allow for an extra 2/3 stop or so headroom if taken to 100%.
However, S-Log2 is identical to S-Log underexposed by half a stop which creates an extra half stop of headroom. As sensors got quieter in the shadows, the difference it made to detail below mid gray was sufficiently small to make the shift worthwhile without completly rethinking the log curve (ie something like S-Log3).
My main suspicion is around Arri LogC. I've always pointed out that S-Log3 is very similar to LogC at Arri's default ISO of 800. I took LUTCalc's LogC calculations from an Arri white paper, which covered how - uniquely - Arri changes the parameters of LogC with ISO, so that black to mid gray is always 10-bit 95 to 420, but as the ISO is increased the highlight contrast decreases so that nothing is ever clipped. I curve fitted my own calculations for the parameter values between those given by Arri. This is in contrast to Canon C-Log which keeps all reflectance values consistent but clips shadows or highlights as the ISO shifts and Sony which consistently records the full dynamic range along to avoid clipping but with a metadata value for the CineEI ISO with the expectation of sorting it out in post.
Anyway, I've just been revisiting the Arri calculations in LUTCalc as it occurred to me that the Arri ACES IDT is the definitive source. Looking at that I discovered that Arri treats ISO 400 to be the nominal ISO for LogC. Using that, S-Log3 is near identical to ISO400 LogC - basically just overexposed 1/3 of a stop (Arri cameras have around 7 1/3 stops headroom at ISO 800, so about 6 1/3 at ISO 400 compared to Sony's 6 stops at native ISO).
As such my thought is that the S-Log3 levels have been consciously chosen to make working with it feel familiar to those used to handling LogC.
As I mentioned in a previous post, that is nothing more than an educated guess but it does make reasonable sense. ;-)
[*** end of geekery ***]
Ben
Last edited: