S-Log w/ 1.31 Firmware PDF

[Doug Jensen]

So would I.

[Doug Jensen]

Here is another series of test shots using the new picture profile version of S-LOG. Read the Vimeo description if you want to know the details.

Sony PMW-F3 Picture Profile with S-LOG testing - PART 2 on Vimeo

[Alister Chapman]

Interesting. Your Vortex-Q profile is very nice, massive improvement over the standard no PP F3 which really is quite shocking.

I think all the Log profiles are over exposed. The log curve uses a lot of compression/data reduction above 70% and sticking white at 90 is putting a lot of the image up into the compressed range giving a somewhat strange look. The S-Log at 70 looks to have the best DR (as expected), but as we are looking at Log in a REC-709 space it's hard to tell as the highlights are severely squashed.

I think if your going to use the Log curve your still going to have to expose well to the left (White at 70) to keep the majority of the image in the more linear parts of the curve. This means a heavy grade is still necessary, in which case you may as well try to retain as much data as possible and thus stick with the standard curve. Once I get hold of 1.4 I'll have a play, but I think that possibly the only useful thing might be to adding some extra saturation.

I've been playing with the F65 and the new S-Log2 curve. What's amusing is that on the F65, S-Log2 is a LUT! If you take the 16 bit RAW output and look at that without a LUT, for a normal exposure all you get is a near black image with just a few highlights visible. So to make on-set monitoring possible the output is converted to S-Log2 which is even more severe than S-Log and S-Log2 is regarded as a viewing LUT!

[Doug Jensen]

Hi Alister, thanks for the comments. I don't disagree with you that the curtains are overexposed in every shot except the last one. That was pretty much the purpose of the test, to see how a little over-exposure in the background would look with the various gammas when I tried to keep the exposure on the face and chart as consistent as possible. In other words, how much contrast can I get away with and still have a decent shot? The final S-LOG shot is the only one where I set the white chip on the chart at 70%.

My goal is to find a way to use the higher dynamic range of S-LOG, but still not have to grade. Maybe it can't be done, but that's my goal. Some people like to grade in post, but my goal has always been to grade in camera and get a nice finished look right then and there. Clients like to see it on the monitor, I don't have to spend time grading (a skill I am not good at and a chore I find extremely boring!!), I don't have to wait for renders, and I don't have a whole other set of files to wrangle.

And if I can use S-LOG bu still avoid grading -- then there's no reason I can't record to SxS cards and not have to bother with an external recorder. On the other hand, if you're going to grade, then I think just about everyone would agree that an external recorder with at least 10bit 4:2:2 capability is mandatory.

Other people may want to grade in post, and I respect that workflow, it's just not the way I want to work. Yup, that's my broadcast roots showing through. I am a TV guy. Guilty.

In my opinion, if you underexpose anything to the point (S-LOG or not) where whites are only hitting 68%, then of course highlights are going to be protected, but then you're absolutely going to have to grade, too. To me, 32% headroom on every shot seems like a waste of DR when only a few shots here and there will really need it. That's 1/3 of my exposure with nothing in it except for the occassional specular highlight. it just makes no sense to me to be so conservative and then have to grade every single frame because I was conservative. I'd never record audio that way, so why record light that way?

My holy grail is a PP with good dynamic range, nice highlight handling, good saturation, the right amount of sharpness, the correct white balance, blacks that touch 0%, and no grading necessary. The new Picture Profile version of S-LOG allows me to come very close to my goal, but is it any better than the cine gammas? That is the question I am attempting to answer now.

One thing I have found is that it is challenging to set up a test that really tells me anything meaningful. For example, in the case of this test with the mannequin and curtains, I had to dim my lights on the chart and face down to about 20% just to reach the point where I could get the curtains to look too hot. If this was a real shoot, I could have raised my lightst to 50% and thus kept the curtains in perfect control. In other words, in order to see what S-LOG might do for me, I had to paint myself into a shooting situation that I could easily have solved in another way -- just add a little light by turning a knob.

[Leonard Levy]

Doug,
Thanks for the explanation as I could not understand what you were up to until reading it. I look forward to seeing how this compares to cine gammas or even to 709 with a well adjusted knee, gamma and black.
I'd be thrilled if you came up with something closer to an Alexa in 709.

I do have to say though that this dream of being able to post with no grading at all seems either illusory or ultimately just sloppy to me. I've been shooting 709 most of my life and always try to get the best picture I can on set, often with a good tech painting each shot. Nevertheless a well adjusted final print always could use grading to some extent (especially whenever you are outside as its hard to judge a monitor.) I just looked at a show the director proudly told me didn't need any grading. It was quite well shot, but I wouldn't let it out of my kitchen without some obvious adjustments. Moreover I think its absolutely impossible to come up with a single profile that would work for every shot.

Sorry you hate taking the time to grade. I love it and have gotten into it professionally because I just couldn't stand seeing how uneven my own shows looked after my clients clumsy editorial. They just never reached their full potential. Of course i've worked with directors who can't stand to wait for lighting either. Good work always takes some time and effort.

Ultimately you should test with a real person because flesh tones are the first to suffer if there's any compression in that range. Also I suggest opening the windows and use what's outside for your overexposure test as the colors and gradations there will also tell you a lot more about your highlights than a white curtain.


Lenny Levy

[Doug Jensen]

Originally Posted by Leonard Levy

I do have to say though that this dream of being able to post with no grading at all seems either illusory or ultimately just sloppy to me.

I'd say it's just the opposite of sloppy. Working hard to get as perfect an in-camera picture as possible (white balance, exposure, black levels, color, detail, etc.) takes more effort at the time of shooting and is anything but sloppy. I'd say shooting with S-LOG, where you basically just under-expose everything, have a couple of crude white balance options, and no other picture profile settings is the epitome of being sloppy. It's the classic "fix it in post" way of working that I dislike.

I already know it is possible to get a great looking picture straight out of the camera because I've been doing exactly that with dozens of cameras (including the F3) for over 30 years. Major broadcasters, especially live sports do it every single day. It is not difficult. It just takes some skill and attention to detail.

My goal is not to see if I can get a decent final image straight out of the F3, I know that is possible, my goal is to get a little more dynamic range by using the very best combination of settings and techniques. Nothing wrong with that. If I had 12 stops, I'd want 14. If I had 14 I'd want 16. But the trade offs of using S-LOG have not proven worth it yet to me. To someone else is might be and I wouldn't try to talk them out of it, but it's not for me at this stage.

BTW, I have not seen your work, so don't take this personally, but a lot of people who THINK they can grade are actually really, really bad at it. Just as there are people who are really, really bad at shooting.

We can agree on one thing you said, "Good work always requires some time and effort". True. I just prefer to apply that time and effort in the field, and not sitting on my ass at a computer. I am a cameraman, DP, cinemagotgrapher, whatever you want to label me, but I am not a colorist and don't want to be. And I don't work on productions with a whole team of specialists who will do it for me later in the workflow. That isn't reality in the world most of us work in.

I guess we have two different approaches to get to the same destination, but please don't say that it can't be done without grading. That is nonsense.

[Leonard Levy]

Doug,
I don't mean to accuse you of sloppiness. I also try to be extremely careful while shooting to get a picture as close to perfect as I can. That's the way most people I'm friends with in San Francisco have been shooting video for years. But my experience is also that when you string together a whole of show of shots that looked "perfect" in the field they always need some finessing to actually match and flow together smoothly. If your stuff doesn't need it more power to you, but its atypical at best. My background in addition to shooting and being a gaffer was that I used to time film negative, so I'm very picky.

I agree that many editors are terrible at grading but the more typical problem in my experience is that they accept at face value decently shot footage, don't see matching problems and don't make many simple corrections that could produce a more beautiful and even result. Moreover I don't think you could come up with a single PP that would work for every shot but I assume you realize that already.

Re: shooting sLog: I don't think that the idea behind shooting sLog is be sloppy and not care about your exposures any more than it would be shooting film or a Red. In fact if you were using a DIT and viewing LUT's on set you could be as careful as shooting 709 w a paint box. But you would need to give up a lot of final control to post and that is an issue on a lower budget shoot. I'm not sold on Slog for my clients because many are loath to get into a lot of color correction also. I'm trying to come up with some pretty simple filter packs they can use to get in the ballpark. In the same spirit I am interested in what you come up with as far as a straight out of the camera look.

[Doug Jensen]

Leonard. don't worry I didn't think you were calling me sloppy.

It would be quite arrogant of me to think that all my footage was perfect and didn't occasionally need some correcting in post. It certainly does. But I make a distinction between the hardcore grading that is necessary and part of the workflow when shooting S-LOG (or some other flat look) -- and making an occasional tweak to fix something that wasn't quite right. There's a difference between grading and just touching stuff up a little when necessary.

I don't agree that is isn't possible to create a PP that covers almost everything. I don't like to change PP's on a shot by shot basis or project by project basis, and maybe that is why my footage is consistent and doesn't require a whole bunch of massaging in post? For the past 4.5 years I've been using just one PP with my EX1 and EX3; basically only PP with my F800 although it has evolved a little; and I mostly use one PP with the F3, but that one PP keeps evolving too. Eventually I know I will settle on one PP for the F3 that I will use for almost everything I shoot. I guarantee it.

[David Knaggs]

I think that there will be two different "markets" for video cameras going forward:

1) Those operators who work with expert colorists, and

2) Those who want to set their exact look before they shoot and only do minimal tweaking in post.

I'm in category 2. That's why a new camera like the Blackmagic isn't for me. I don't believe that it has any Picture Profiles (or scene files) that can be manipulated to give you the look you want. My impression is that it shoots flat and that's why they bundle the coloring software with it, so it can be expertly graded afterwards.

An expert colorist can make such footage look great, but anything less than expert grading looks incredibly fake to me.

I would like more dynamic range, though. That's why I'm following Doug's S-Log Picture Profile experiments with such keen interest. It will influence (for me) whether the next purchase is F3, C300 or the new FS700.

[Leonard Levy]

Doug, I'm wondering what you've been doing to the sLog to get the new look. I'm guessing mainly raise the gamma and the black stretch then add sat?
Does it respond in a more linear way to gamma changes than cinegamma which gets quite weird.

[Alister Chapman]

I'm definitely interested in your experiments Doug and of course once I get V1.4 I will be trying my own tests, it has to be done after all.

But I do have to wonder if it is really possible to create an good in camera look that has 13 stops of DR? This not a reflection on the way you shoot Doug, just wondering if it really is asking for too much.

The problem of course being that if you take 13 stops and squash it in to the typical 7 stops that most TV's and monitors can deal with, it will look flat and it will almost certainly look un-natural as you are compressing the image. The issue with using a Log curve to do this is that the highlights are compressed, with S-Log everything above about 60% is getting quite strongly compresses and above 75% very highly compressed. Compare that to a cinegamma where the compression doesn't really start to get significant until you get to about 80% and on paper at least that suggests that with the log curve your going to struggle to get skin tones to look natural and bright scenes are going to look very flat.

While you can add in some negative black gamma to make the image look more contrasty, this isn't going to change the highlight performance and with no other way to modify the top end of the curve your kind of stuck.
Anyway, perhaps you'll prove me wrong Doug, but I'm not convinced you'll get a useful benefit using the log curve for your way of working. I think that the cinegammas are a good compromise between extra dynamic range over standard without the image washing out.

If you do choose to shoot log and grade, the advantage you (or the colourist) has is to be able to selectively work on parts of the image, so you can just bring the highlights down while leaving everything else intact, which is something that just can't be done in camera, not yet at least.

[Leonard Levy]

Alister,
I'm curious why it is easier to spread out all 13 stops in post (back to 709) than in the camera itself. In my grading experiments with Slog I've been first de-logging using a curve which by itself would push the highlights into over exposure , then I'm adding a shadow highlight filter to additionally compress the highlights ( and also to finesse the shadows but that's less in need).

So technically why is it so much harder to achieve in camera? It would certainly be difficult to get the balance exactly right on set and maybe more dangerous. I assume that's kind of like what the cine gamma does already but not as well.

[Alister Chapman]

The problem is that the settings available in the F3 don't allow you to make any changes to the shape of the S-Log curve other than a bit of black stretch which only raises or lowers the bottom of the curve. So your stuck with the same highlight handling.

In post you have much finer control over which parts of the image you adjust and by how much. A proper grading tool can work on very specific tonal ranges giving you much greater control than you could ever have or perhaps want on a camera. One issue is that if you start adjusting specific ranges relative to the others within the same image is that it can be very hard to un-pick that look if you later decide you don't like it. For me what I want is a neutral image so I can pick my look in post.
I'm working with F65's at the moment that have 16Bit linear recording and it's just staggering what you can pull out of a shot that is so unbelievably over exposed it looks almost totally white. As the recording is linear there is no highlight compression. As it's 16bit one stop has over 4000 grey levels, so even if it is so overexposed that the entire scene is in the top one stop, you still have 4 times as much brightness information in that one stop than 10 bit S-Log has in it's entire range.

[Leonard Levy]

Thanks,
So if I'm understanding you right there is no inherent technical reason you couldn't manipulate or paint the SLog DR in camera even to an 8bit recording, but our you're suggesting 2 reasons against it.

1- Its just hard to do in camera as even when it becomes another gamma in the new firmware you just don't have that many parameters available to control in the F3.
I assume though that a good 3D LUT made with sophisticated software would theoretically be able to do that. Of course the F3 doesn't accept 3D LUT's so you'd need to record to something else.

2. But the 2nd objection is that even if you could record a sophisticated LUT then the curve would have so much compression in very specific parts of the curve that you would really be stuck with it and have very little ability to change it in post. I'm guessing less ability than you currently have with 709 or cinegamma?

Makes sense to me. Thanks

[Leonard Levy]

Thanks once again Alister,

So if I'm understanding you right there is no inherent technical reason you couldn't manipulate or paint the SLog DR in camera even to an 8bit recording, but you're suggesting 2 reasons against it.

1- Its just hard to do in camera as even when it becomes another gamma in the new firmware you just don't have that many parameters available to control in the F3.
I assume though that a good 3D LUT made with sophisticated software might theoretically be able to do that though. Of course the F3 doesn't accept 3D LUT's so you'd need to record to something else.

2. But the 2nd objection is that even if you could record a sophisticated LUT then the curve would have so much compression in very specific parts of the curve that you would really be stuck with it and have very little ability to change it in post. I'm guessing less ability than you currently have with 709 or cinegamma?

Makes sense to me, although I am also very interested to see what Doug comes up with. Maybe you can get something that's halfway there, similar to the Canon log curve.