FS5 S-log tips and gotchas

[Nigel Davey]

Thus far I've only used profile 1 (cine) on the FS5 and Doug Jensen's Vortex-FS5-Pro cine profile. I have never used log before on the FS5 or any other camera.

This weekend I intend to go out and shoot some test footage in S-log to a) get some experience and b) work out a good workflow. I would really appreciate tips, advice and a heads up on the gotchas before I attempt this. Keep in mind I'm starting from scratch, so need to start/learn from the basics upwards.

I suspect the two areas I need to really get my head around are achieving correct exposure and monitoring on camera. So advice about ISO's (I tend to use dB right now in cine), zebras (with recommended levels), S-log 2 or 3, etc, are gratefully received.

I know Alister Chapman covered a lot of this in his FS5 introduction video last year. So I will go and watch that. But nonetheless folks have had 6+ months of real world experience since then, so I'm interested in your thoughts.

In case it's relevant, I intend to go and shoot test footage at a old ruined church that was bombed in WWII. So there will be a fair amount of of grey stonework, green ivy, broken windows/frames, headstones and (hopefully) blue skies. In terms of lens options I have the Sony 18-105 kit lens, a Sony 35mm (f1.8), a Tokina 11-16mm (f2.8) and various Nikon glass (70-200, 18-55, 35, 55). I'll also take my slider, jib and sticks for support/moves.

I suspect my three main lens choices on the day, unless advised otherwise, will be the Sony 35, 18-105 and the Tokina 11-16.

[Dave Sperling]

My advice would be to do some simple s-log testing in your own office or living room before going out in the field. Put up a gray scale and color chart - have someone stand in - shoot at a variety of exposures - see what your parameters are when you start grading. My feeling is that s-log exposure will be somewhat personal based on your subjects and workflow, so try out a variety.
Also remember that when you are shooting a 'test' the idea is less about getting usable footage, and more about finding our what your optimum settings are. So unless you have time for numerous different/documentable options for each setup, don't try to do a test with something you are worried about capturing.

That said, occasionally a 'test' ends up being surprisingly usable. Not the same camera, but the idea here is informative. I was recently doing a shoot of scenes from a Broadway show. (They were doing the scenes specifically for us during the afternoon - creating press clips - and the lighting designer was there to adjust the lighting design for better video images.) We were using regular disc-based cameras for the shoot, but also had an A7s for some specialty shots. I'd adjusted a profile to closely match the big cameras, but we decided to shoot some test s-log footage (unmanned wide shots, since we were all operating on the big cameras at the time.) There were a lot of video projections on scrims as part of the set, and the big cameras were having trouble seeing them. We tried several different s-log exposures just to have some footage (not actually intended to be used) that the company could play with grading to learn more. Interestingly enough we found that what was theoretically a good couple of stops overexposed (faces recorded up around 80%) ended up giving them an image that worked beautifully for bringing up video projection details in the shadows with minimal noise -- while the compressed curve area of the overexposed faces was not that bad in the extreme wide shots. They were able to use some of it in the finished cut - because the important part was seeing the dim videos, and the s-log kept the faces gradable / with detail. Would I recommend exposing faces at 80% -- ? Absolutely not / Never. But our test bracketing found a loophole. also note that in grading they had to make their own custom curves rather than use a LUT, since the LUT would have probably placed the brightness on the faces somewhere in the light bulb range...
One thing to think about with s-log is where you need to expose, based on how you color correct the footage, to keep your shadows from being too noisy. You will only determine this with testing.

[Nigel Davey]

Thanks Dave. Some great advice that I will be following as I conduct my tests.

[Cliff Totten]

Important - SLOG-2 and SLOG-3 do not top out at the same IRE on your scopes/zebras.

Going from memory here but SLOG-2 will clip right at around 109-ish like some other profiles do. However, SLOG-3 clips way lower. I want to say it's 94 IRE? So check that and make sure that if you want your zebras to reveal what is clipping in your shot, you will need to lower them down. 94 IRE is roughly the brightest possible white.

The Fs5 seems to be OK these days, but older Sony SLOG camera's like the original A7s will often clip bight blue lights in the most bizarre way. Things like stage lights or police car light and more would have a nasty blue channel saturated halo around them. Again, I think Sony has fixed this on newer models like the RX10-II, A7s-II and FS5.

CT

[Nigel Davey]

Thanks Cliff. I think I'll mostly be using Slog 2, so will set zebras to 109.

[Cliff Totten]

Make sure to point to point at something like a bright light and deliberately blow it out pretty bad. Then, make sure your zebras are right at that clipping point. Also, you might be able to trust that the 109 IRE internal is exactly the same 109 IRE output over HDMI but you might want to verify that to make sure. I have not tested that myself with my FS5 yet but I'm guessing it should be exactly the same.

[Doug Jensen]

Originally Posted by Nigel Davey

Thanks Cliff. I think I'll mostly be using Slog 2, so will set zebras to 109.

109??? !!
I hope you realize that you'll be over-exposing by several stops above Sony's recommendation of 61% for whites. Even if you weren't shooting S-LOG, I can't really imagine any scenario where 109% zebras would be a good idea. Zebras should be used to help you get the correct exposure -- and not just to tell when things have gone over the cliff and are about to clip. That is not what zebras are for.

Personally, I think Sony's suggestion of 61% is too low, but I don't know anyone (who's work I respect) who would go over 75% on bright reflected whites. 109% is way over the top. S-LOG is NOT designed to have highlights pushed right up to the brink of clipping. I highly suggest that you follow Dave's advice and shoot some controlled real-world test footage with S-LOG and then bring it into DaVinci Resolve and go through the entire workflow -- all the way to delivery in order to ensure you've got a handle on it before you embark on an important shoot. S-LOG is not a workflow you should take lightly (pun) and requires a different way of shooting and post processing.

Also, you don't say in your post above whether you'll be shooting UHD or HD. If you're planning on doing UHD, I suggest you abandon all plans for S-LOG with your FS5 because the files will only be 8-bit . . . and 8-bit isn't good enough for quality grading of S-LOG. You'd be much better off baking in a WYSIWYG look with one of the other gamma choices and then only doing minor touch ups in post.

[Nigel Davey]

Thanks Doug re advice over zebras. I will only being doing tests in HD. Being perfectly honest the type of work I get is most suited to cine profiles for now. But nonetheless I want to find out what those extra stops of dynamic range can offer in a pinch.

Unfortunately I don't have Resolve. For PP I was intending to start with the S-log LUTs offered under Lumetri in Premiere and then go from there.

On a connected but different note I assume the better workflow with the FS5 is to use the Gamma Assist Display? ... although using zebras for setting exposure.

Also is there any value in using AE adjustment to achieve the recommended over exposure. Or is that already baked into the FS5 PP7?

Keep in mind I'm just starting out with log and have set aside the rest of today to read articles/watch tutorials.

[Cliff Totten]

Doug is absolutely right. SLOG requires ALLOT of practice. Actually, TWO types of practice:

1.) Shooting with it. Exposing SLOG is nothing like most of us are used to with "rec 707 ish" profiles. You will need to learn the term "expose to the right". So, you will over expose between a stop or two and then bring "down" your wave form in post. SLOG is only an archival technique designed for maximum dynamic range capture in a small rec 709 "bucket". It looks terrible but that's OK, it's only job is to "get it all in". You job is post is to expand it all back and reallocate the brightness where it "really" belongs. (in a very careful was of course)

2.) Grading it. This takes allot of practice too. If you over expose between 1-2 stops without clipping the important areas of your scenes, you bring your waveform "downward" in post and this will help to hide noise. You NEVER want to be in a situation in post where you need to bring levels "up" or adding gain. It's noise city! Bringing "down" is good, bringing "up" is bad. The idea is to capture a brighter image in the field that is not clipped, get home in post, adjust your whites and normalize in the "downward" direction.

Yup, normalizing SLOG-2 on an 8bit capture is very tough on those 8 bits. With only 255 steps per RGB channel, they do not that stretch well! However, if you expose very carefully in the field, you can get decent results. SLOG-3? It's soooo nice but you really need 10 bits for this.

Please do not shoot SLOG on a job if you have never done it! Do lots of testing first.

The best resource I know on SLOG exposure (and proper over exposure) is Alister Chapman. He is a well respected guy that speaks for Sony in many events.

www.XDCAM-User.com

http://www.xdcam-user.com/2015/12/de...eform-display/

http://www.xdcam-user.com/2014/08/ex...-and-exposure/


Also, on the FS5 with SLOG-2, only shoot at 0db/3200ISO. Dont add gain to you shot. If you need more brightness, add more light by opening your IRIS or slowing your shutter. (remember, gain/ISO is not "light") If you still cant get the brightness you need, turn off SLOG for that shot.

[Cliff Totten]

Originally Posted by Doug Jensen

109??? !!
I hope you realize that you'll be over-exposing by several stops above Sony's recommendation of 61% for whites. Even if you weren't shooting S-LOG, I can't really imagine any scenario where 109% zebras would be a good idea. Zebras should be used to help you get the correct exposure -- and not just to tell when things have gone over the cliff and are about to clip. That is not what zebras are for.

Personally, I think Sony's suggestion of 61% is too low, but I don't know anyone (who's work I respect) who would go over 75% on bright reflected whites. 109% is way over the top. S-LOG is NOT designed to have highlights pushed right up to the brink of clipping. I highly suggest that you follow Dave's advice and shoot some controlled real-world test footage with S-LOG and then bring it into DaVinci Resolve and go through the entire workflow -- all the way to delivery in order to ensure you've got a handle on it before you embark on an important shoot. S-LOG is not a workflow you should take lightly (pun) and requires a different way of shooting and post processing.

Also, you don't say in your post above whether you'll be shooting UHD or HD. If you're planning on doing UHD, I suggest you abandon all plans for S-LOG with your FS5 because the files will only be 8-bit . . . and 8-bit isn't good enough for quality grading of S-LOG. You'd be much better off baking in a WYSIWYG look with one of the other gamma choices and then only doing minor touch ups in post.

109 would be the actual clipping point that would make something unrecoverable in post. If you caught some highlights (like bright clouds) at 105 you could draw those back in range and normalize them in post. Once they break the 109 barrier, they are permanently destroyed. That's all I'm trying to say about using Zebras for that function.

Also, how we expose something is very scene or "subject" dependent. If there is black bear fighting a grey wolf under a shady tree, we worry less about burning our clouds or even the grass in the distant field. We gotta get that bear in the shadows over anything first. So we often "choose" what parts of our image to destroy and what parts to save. Even SLOG can't give us everything.

I think Alister's +1-2 stops over is a good rule of thumb and works very well for most scenes. He has some nice LUTS too that are specifically tailored for +1, +2 and +3 over exposure.(+3 is allot)

If you are stuck in a really high dynamic range scene with SLOG2 and you have everything that is important you from 109 and below, you have allot information to work with in post. (easily 13+ stops?) 109 is super hot and highlights can look squeezed up there.

How much post work do you want to do? People have to figure that out when they are shooting it. Everybody is different.

CT

[Nigel Davey]

Thanks Cliff. This afternoon I've watched both Doug Jensen's FS5 Master class on picture profiles and the hour long section (relating to cinegammas and slog) Alister Chapman gave on the FS5.

A couple of things stood out that I now how questions on. Doug prefers to use slog 2 in HD whereas Alister recommends slog 3 because it apparently grades better and doesn't generate a green tint when strectching levels.

Tomorrow I intend to shoot/test on a dessertedand ruined church (think English church with grey flint stone) and the weather forecast is good. Thus lots of light and possibly even clear blue skies. Since I intend (for now) to grade in Premiere CC by applying the relevant LUT and therefater tweaking by eye, most likely with Lumetri, can anyone advice whether slog 2 or 3 is the better option.

For exposure I think I'll try Alister's recommendation of setting zebras to 70 (for either slog 2/3) and popping a piece of white paper in front of the camera and tweaking exposure until the zebras appear. That should put the exposure over by 1-1.5 stops according to Alister. So I assume that will give me a starting point for peak white. I assume after that I'll need to get a feel for how things look on the screen with the Gamma Diplay Assist enabled and tweak exposure as it feels appropriate.

A couple of things I'd like some clarification on though. Firstly I assume the requirement to overexpose by 1-1.5 stops is not built into pp7, pp8 or pp9? In other words is it ok to let things look a stop or so overexposed on the monitor (with gamma assist enabled)?

Secondly how do I get zebras to precisely 70? My current understanding is they have a aperature setting (for bracketing) that the lowest you can set it to is 2, thus 1 either side of 70 in this case. Will that make any tangible difference?

Finally if I do go with slog 3, which version is potentially better for my church scenario pp8 or pp9?

Forgive me if these are daft questions, I've taken in quite a lot today and perhaps not fully grasped it all.

Also I genuinely appreciate the time folks are taking to try and steer me in the right direction.

[Cliff Totten]

You are not asking daft questions at all. They are excellent questions.

SLOG-3 is spectacular. It really is wonderful and it doesn't "curve" the way 2 does. 3 works in a way that more linear than 2. It sees deep into the shadows but it is brutal on an 8 bit codec. That 100mbp/s XAVC long GOP codec is a good one but SLOG 3 hammers it hard. I have gotten "OK" results with SLOG3 and 4:2:2 ProRes via HDMI but there is no escaping the 8 bit limitation and the color banding that it brings.

When I shoot SLOG, I mostly record to ProRes on my Shogon or Pix E5. ProRes is an intraframe CODEC that is pretty durable as far as compression artifacts go. Stretching your waveform in post will bring out compression artifacts too. XAVC-L is OK but ProRes holds noticeably better with SLOG to my eyes. (wont really help 8bit banding much)

I tend to use multiple Zebras. I have SLOG clipping point on the small camcorder screen and 70- 80 on my bigger monitor. Depends on what I'm shooting.

I do love Alister's +1 and +2 over exposed LUTS. I do have his +3 installed too. I donated on PayPal for them. They are well worth it. I tend to bounce back and forth radically between LUT and no LUT on my external monitors but almost never use Sony's camcorder monitor assist. Once you see that milky SLOG look, you start to get used to it!

When you shoot SLOG, you almost have to say in you mind "Hmmm...exactly how will I grade this exact thing here..." and do that will EVERY shot. The closer you get it right in the field, the less stress you will have in post.

Because SLOG-2 is easier on 8 bit, I stick mostly with that. It's a bit more tricky to grade than SLOG 3 but it's not as flat as 3 and uses almost all of the bits that 8 bit has to offer. When you see SLOG-3 on you waveform monitor, your darkest black is WAY up and your brightest white is WAY down. You gotta stretch is like silly putty to normalize it. It really needs 10 bit. (1024 steps per RGB channel...instead of 255)

SLOG took me literally a year of practice. I strongly recommend getting allot of test shots that are deliberately "BAD". Try shooting SLOG-2/3 on automatic and +/-0EV. Just let the camera decide your exposure. Lift up that video in post and see all that noise? Then shoot the same scenes +1 and +2 over. Bring it all down and see how different the grading result is? Deliberately UNDERexpose by -1 or -2 stops and see how bad the problems show up in your NLE. Scarry!

Shooting bad SLOG video and trying to grade it really is one of the best ways to understand SLOG and getting it right.

[Nigel Davey]

Thanks Cliff. I'll try a variety of combinations tomorrow. Might even shoot standard Rec709 or Doug Jensen's profile on the same scenes to see what I gain and lose in Post vs slog.

Out of interest now that you've mastered slog, roughly what percentage of your footage/projects do you use it in?

[Cliff Totten]

lol,...dunno if ANYbody can say they are an SLOG "master"...hehe. I'm just a regular guy that shot allot of bad video, READ ALLOT, shot even more and eventually figured out over time how to get consistently good SLOG video with. It taught me how to color grade much better as well. That's all. I'm nobody special!

I use it only for projects that I know I want to spend allot of time on. I'm going to Yellowstone in August. I'm taking my FS5 and a A7s-II rig. I'll probably shoot 90% in SLOG-2 on a PIX-E5 external recorder. 100% 8 bit UHD. (On the FS5, I'm choosing resolution over bit depth)

This is because I'm expecting to take allot time of carrying gear, carefully shooting and grading. For "normal" projects, I dont use SLOG. I do what Doug does. Get it close in camera, do light tweaks in post and get it out the door.

Here is the ironic thing about SLOG. You can capture all this dynamic range but at this moment, you really cant "deliver" that dynamic range! What are you going to do? You will capture more highlights and potentially more shadow detail but once you add all the contrast back in, shift the waveform down and normalize it to look good for typical rec709 televisions/monitors....you just tossed out all that extra range that you worked so hard to capture! Right?

This is what SLOG is, a "capture" and "archive" format. It's not a "delivery" format. What you CAN do is return to a project years from now, when rec2020 TV's are more popular and re-render a new file that DOES deliver all that range you captured.

SLOG just gives you some more options in post that you might not get with rec709. More information captured in your image that you can decide what to throw away or to keep later on down the road.

CT

[Nigel Davey]

It's interesting you say that Cliff. It ties in with something I've noticed in quite a few log clips I've watched online. Either they still look a bit milky or they look like you would get similar results from 709.

I'll still play with slog to see what is achievable in post. But now you've said what you have, I will definitely record a duplicate clip in 709 to compare it with.

On the plus side most of this years new 4K TV's seem to have HDR. Not sure if there is a universal standard yet though.