F55: S-SLOG3 vs. S-LOG2 -- can you tell the difference?

Im a bit lost here you shot raw and recorded a bigger DR than either profile.

Your journey is therefore..

raw->log>look

Basically two luts or transforms.

Now the first two transforms might simply be..

(raw value + lut1 transform value +lut2 transform value)

Slog3
Raw+7+3=Raw+10

and Slog2

Raw+5+5=Raw+10

No difference

But if one had a different source data.. like clipping in one of the profiles but not the other

You could run the maths and get..

7+3=10

and

clipped+5=error

--

So apart from the mild colour differences that are seen one is not really actually testing the DR that the two 'looks' record?
 
Sam, I don't know what tangent you're going off on but I just wanted to see if there was a difference between how graded S-LOG3 and graded S-LOG2 looked in post. Assuming that no highlights are blown with S-LOG3, I'm confident in saying someone can get identical results no matter which one they choose. If your testing shows different, I'd love to see the actual video footage (not math or charts) that shows that there is a difference.
 
The tangent is..

A) I think using raw source equalises any difference through 'oversampling' of the data, aka capturing everthing the sensor can do and not throwing some of it away.

B) Sony issued data suggests Slog2 to be 'thicker' than Slog3 and have less usable DR.

Slog2 should therefore be a more appropriate choice when shooting in controlled environments - such information should be of value to the shooting community.

Im shooting in the studio tomorrow, and will try a little test tonight if I can..

S
 
LennyLevy said:
Very pretty effect Doug. Obviously they look identical and that does tell us some things, but is it an accurate idea of what would have happened had you shot the footage not as RAW but as XAVC in the camera where the amount of highlight or shadow attitude would have differed. Obviously the color is identical in your tests, but again would that also be the same if you didn't start from the same RAW file?
Click to expand...

Good question. And I didn't want to speculate with an answer until I had a chance to compare them with some further testing. As you can see in this video, the results are almost identical between XAVC and RAW. I even tested everything at two different exposure levels just to see if that made any difference since, of course, S-LOG2 and S-LOG3 are curves and might not behave the same way at different exposures. The main difference I can see is that the XAVC files are more cyan than RAW. Another difference is that the RAW files are noticably less noisy (may not show on Vimeo) and quite a bit easier to grade. Also it confirms my preference for S-LOG2. S-LOG3 gets all the buzz but I still like S-LOG2 better.

I hope you can follow what I'm doing with the tests because I don't want to spend any more time explaining it. :-)

Password: SLOG
https://vimeo.com/226979713
 
Love how you do empirical testing, there's no better kind. To the untrained eye, I'd say there appears to be virtually no difference or very subtle differences, which proves that with the right grading (or LUT), s-log footage can be graded to look like anything, including other s-log footage. I have a 50/50 chance on this, but will say Side A is s-log 2 and Side B s-log 3. The only difference I noticed was at the 1:06 mark where you have the green grass. I notice the grass looks more saturated on the left, so I'm guessing that's the larger color s-gamut color space. Side B looks more subdued green, so I'm guessing that's the cine side. Really fantastic test.
 
Since its on the table now I felt like sharing my test I did in January:

https://vimeo.com/227074533/3e03efd054

In resolve I did a lot more than this, I tested lots of LUTs and color space conversion (both with resolves tool for this and with custom LUTs made in LUTcalc). I also tested the Alexa mini, RED helium and lots of other cameras, but have not made any real tests between them. In terms of FS7 and Slog, basicly what I came up with was:

1. Matching the two profiles to an acceptable level (maybe not always acceptable) was simple, but I could never get them to match perfectly. The colour chart kept playing with me.
2. In terms of tonality Slog2 looks slightly better overexposed while Slog3 looks better underexposed.
3. In terms of color I like Sgamut3.cine better overall, feels a bit more vivid in skin even after LUT.

In general the difference was not enough to make me want to switch between the profiles. Its not worth the more complicated post workflow having more than one profile for me. I decided to keep it simple and going with Slog3 Sgamut3.cine. For future HDR-stuff I might switch to Sgamut3, since that colour space is larger than rec 2020.

If you feel like I missed something, made errors or just curious to try this out in resolve here are some DPX stills from the vanilla log images.

DPX stills:
https://www.dropbox.com/sh/dgiumxlh57rli05/AAAWiDW_dkyJbR4q82XJKrSYa?dl=0

PS. forgot to mention that we choose to rate the camera as ISO800. So the camera at 0 is exposed for TRUE ISO800 (not Sony EI800) which we felt was more in line with reality than Sonys ISO2000 claim. Maybe we should have exposed for ISO1250 to be truly accurate, but we wanted to see how it compared to other cameras that was all lit for ISO800.

Cheers,
Ed
 
Last edited:
Doug Jensen said:
Also it confirms my preference for S-LOG2. S-LOG3 gets all the buzz but I still like S-LOG2 better.

I hope you can follow what I'm doing with the tests because I don't want to spend any more time explaining it. :-)
Click to expand...
Doug - one quick question about this - because I want to understand the logic behind your statement. Since in post #1 of this thread, it seems to me you have proven that using the same Sony LUT for each version of Slog yields nearly identical results in the image. So I assume your preference for 2 over 3 is because you feel you can push or pull the image further (or perhaps more easily) than you can using 3? Or am I misunderstanding something?
 
Sorry if I wasn't clear on a couple of things.

1) The same LUT should not be applied to both versions of S-LOG. Each version requires a different LUT to arrive at similar "looks". In the tests above, and in my real-life work, I use:
S-LOG3: SGamutCineSlog3_to_LC-709TypeA
S-LOG2: From_Slog2SGamut_to_LC-709TypeA

2) If the appropriate LUT is applied to each version of S-LOG then they will be virtually indistinguishable from each other as shown in the split-screen video: https://vimeo.com/226519193
People have a 50/50 chance of identifying one or the other, but really, there isn't much difference between them. BTW, S-LOG2 is on the left and S-LOG3 is on the right.

3) But there are some slight differences between the two S-LOGs if you look closer at the video with the flowers. I prefer the slightly brighter and more saturated colors of S-LOG2. Granted the differences are very slight and it would only take a couple of very small changes to make S-LOG3 look identical to S-LOG2, but it's one less step with S-LOG2. https://vimeo.com/226979713 (password is SLOG)

4) My preference for S-LOG2 is because it allows highlights up to 107% before clipping occurs, while S-LOG3 clips after only 93%. Why would I want to give up 14 IRE of headroom if I don't have to? I've already showed that both of them can be graded to look the same in post so I don't see any reason not to choose the one that has more headroom on highlights? Obviously I wouldn't inentionally expose anything important in the 94-107% region of the image, but it can make a difference when you shoot towards the sun at sunset, have bright backlit hair or skin outdoors, have specular hightlights on water, chrome, etc. Why would I want to clip those highlights sooner than I need to if all other things are equal? S-LOG2 is every bit as easy to grade as S-LOG3 if you used the right LUTs. So even though the benefit of 2 over 3 is very slight, I don't see any benefit of choosing 3 over 2.

BTW, neither of these videos shows the benefit of better highlight handing of S-LOG2 but it is something I've seen and tested before. So someone will either need to take my word for it or do their own testing because I'm not going to post anything to prove this point.

So the bottom line is that most of the time I could use S-LOG2 or S-LOG3 and get nearly equal results with each one. But my preference for 2 is because of the brighter colors and better highlights handling.

I hope that helps clear things up.
 
Last edited:
Doug Jensen said:
It does matter. Thanks for chiming in. I'm glad not to be the only one.
Click to expand...

Im still trying to get my head round this one (I know that annoys you)

It seems like neither 2 or 3 use the full 'levels bucket' - just because 2 uses the top bit of the bucket doesn't mean that it is any better at using the bucket?
 
Up to now I have not read sound points here, why s2 should be better then s3 or the other way round. But I can imagine that there can be a ratio, but that has neither been discussed in this thread, nor in the last one.

But before stating what I think here - to apply LUTs is one way only. Another one - and a better one - would be to use the quite simple ACES 1 workflow as available by now in tools like Resolve 14 or in other tools.

Given the different shape of s2 and s3, a clear difference is that s2 allocates a higher number of luminance signals to the highlights, but much less to the darker stops. If one shoots and grades for rec709 with the limited number of 6 stops only, that the higher number of luminance signals in the midlights and highlights can be great. But the much lower number in the lowlights can be an issue.

For HDR and rec2020, in terms of dynamic range s2 and s3 seems to be identical. Sure, with s3 we loose some signal range above 90%, but to be honest: who cares with 10 or 12bit? The main advantage that I see for rec2020 and HDR is the more even distribution of signals to each of the stops. Because if I grade for HDR I wish to utilise my 14 stops from the FS7/5, and a more even signal distribution is here a clear advantage because it gives me more latency for the postpro.

So instead of saying quite unreflected that s2 or s3 is better, we have to think about why and for what purpose it is better. Since I intend to go for HDR I prefer s3 - and for rec709 I tend to use HG anyway. And I tend to use only ACES 1.0 in the future.
 
Maybe Slog3 is actually disigned for cameras which can record a wider DR than the FS7 - aka future HDR cameras - that will continue beyond the 900 odd clip point.. and therefore using 3 would actually be dumb with the 7
 
wschmid said:
Given the different shape of s2 and s3, a clear difference is that s2 allocates a higher number of luminance signals to the highlights, but much less to the darker stops. If one shoots and grades for rec709 with the limited number of 6 stops only, that the higher number of luminance signals in the midlights and highlights can be great. But the much lower number in the lowlights can be an issue. .
Click to expand...

Yes this is my understanding. That for 709 S2 is basically better for a controlled DR scene.. but if you have a complex scene and need to dig into the shadows during the grade then S3 is probably a better choice as S2 has weak shadows.

For HDR and future wider DR cameras S3 will become the obvious choice?
 
morgan_moore said:
Maybe Slog3 is actually disigned for cameras which can record a wider DR than the FS7 - aka future HDR cameras - that will continue beyond the 900 odd clip point.. and therefore using 3 would actually be dumb with the 7
Click to expand...

That is true for sure - because cameras with better sensors then in the FS5/7 can utilize the two additional stops foreseen by s3.
 
morgan_moore said:
Yes this is my understanding. That for 709 S2 is basically better for a controlled DR scene.. but if you have a complex scene and need to dig into the shadows during the grade then S3 is probably a better choice as S2 has weak shadows.

For HDR and future wider DR cameras S3 will become the obvious choice?
Click to expand...
For HDR I think that s3 will be the better choice - as said above. For rec709 one can choose between HG and s2.

But one can use s3 also without risk for rec709 I think. In the future it may become more approriate to think about how we grade a film. Will we deliver to rec709 only as most of us do today? Or will we deliver both - a rec709 version beside a rec2020 version? Maybe it will be both.

And if it is both, a good ACES workflow may become more important. Because with ACES the postpro both to rec709 but also to HDR becomes easy.
 
wschmid said:
For HDR I think that s3 will be the better choice - as said above.
Click to expand...

You're probably correct about S-LOG3 being better for HDR and possible future cameras with more DR. All the official Sony workflows and documents that I've read about HDR use S-LOG3 as the starting point. But I'm not concerned with HDR today nor with cameras that don't exist yet. Neither of those things have been a part of this discussion until now and were not factored into my conclusion that I prefer S-LOG2 right now. It's good to have an eye down the road -- but right now I'm concerned with getting the best results possible with the cameras and deliverables I have to deal with today.
 
We are not talking about a fiction or futur cameras. With 14 stops with s2/3 both the FS5/FS7 are ready for HDR today.
 
You must log in or register to reply here.
Back
Top