FS700 + Samurai?

[Leonard Levy]

I'm wondering how much advantage you get from coupling an FS700 with a Samurai.

You would get 422 + 10 bit recording but only from an 8 but stream coming in.

1- Will it noticeably improve the picture?

2 - Will this give you good green screen?

3 - Will recording at 10 bit even from an 8 bit stream help with grading?

4 - Will it help for, or be sufficient for recording SLog2?

5 - How would this combo compare to F3 or F5 or C300?

[Colin Elves]

Originally Posted by Leonard Levy

I'm wondering how much advantage you get from coupling an FS700 with a Samurai.

You would get 422 + 10 bit recording but only from an 8 but stream coming in.

1- Will it noticeably improve the picture?

2 - Will this give you good green screen?

3 - Will recording at 10 bit even from an 8 bit stream help with grading?

4 - Will it help for, or be sufficient for recording SLog2?

5 - How would this combo compare to F3 or F5 or C300?

Although I'm a cameraman and don't have much to do with the post I purchased a Samurai along with my FS700 last year, so I at least have some experience working with the two.

To answer your questions:

1. No, I don't think it will 'noticeably' improve the picutre. It improves it, but only when you start pixel peeping. However...

2. It definitely improves the quality of the keying. Moving from 420 to 422 is a pretty big jump in terms of colour resolution. Plus the better compression can help with complex moving objects.

3. Again, the 422 seems to help with the grading a bit - especially heavy grading, but it's definitely not the same as graing 10bit.

4. I've shot a lot with Slog2 on the Samurai in the last couple of weeks. I've done some quick and dirty grades and I'd say that Slog2 is useable, but tricky. The issue is with the lower end of the curve, the shadows - as it's only 8bit material then tonal changes in the shadows are covered by a very narrow range of data values - so you get a lot of noise, macro blocking and banding in those tones. ETTR seems to be quite effective when using Slog2 though - as you end up pulling the midtones to the highlights that have a lot more units tied to them. Relative to 420, however, I can't say as I haven't compared the two, but I suspect that the 422 helps quite a lot with these issues (ie they'd be worse on the AVCHD versions).

5. I shot the FS700 Slog2 side by side with a C300 on cLog and the FS700 had considerably more DR. If you get the exposure right, I think the graded images are going to be better. But the C300 stuff will grade a lot more easily as the Clog is built for 8bit. Currently my thinking is that FS700 profiles using the cinegammas might be a safer bet for the FS700 and compared to the C300, the C300 produces nicer images out of the box - but the FS700 is more flexible, because of the slow mo, the 12bit Raw option etc.

As compred to the F3 - and Slog F3 with a Samurai to record the true 10bit 422 out - the F3 will give better images. But the price difference between and F3 and FS700 means that you can, for the same money, get an FS700 and Odyssey 7Q - and then the output from the FS700 will be superior - as good as the 444 RGB output you get from the F3, if not better assuming Slog2 crams a bit more info in than Slog. Again, I think the FS700 is more flexible.

As compared to the F5. The F5 is better. It will do the same HFR 2K @ 240fps. The Raw you get is 16bit, rather than 12bit and it combines with the Sony R5 recorder properly and it records broadcast quality internally, without relying upon a flimsy BNC cable. However an F5 package, without the R5, will cost nearly twice as much as an FS700 + Odyssey.

So, ultimately, it all depends upon: a) how much money you have to spend, b) what you need the camera for and c) what you would spend the money you save buying an FS700 on.

Yours,

Colin Elves
DoP, London
Colin Elves - Director of Photography and Steadicam Operator, London, UK

[Leonard Levy]

Thanks Colin, very informative.

I own an F3 and a Samurai but am trying to figure out what to do for future proofing . Until the 7Q really comes out all the promised FS700 goodies are still vaporware, so I was wondering what good the Samureai can provide. Sounds like for green screen it suffices but not really for SLog2 yet.

[Leonard Levy]

a couple more questions if you're still out there Colin.

- I understand that with Slog, ETTR ( I had to look that one up) is not recommend because there is so much compression in the top end. Is SLog2 any different in that regard?

- How are you mounting the Samurai on the FS700 and how cumbersome is that?

Lenny

[Colin Elves]

Originally Posted by Leonard Levy

a couple more questions if you're still out there Colin.

- I understand that with Slog, ETTR ( I had to look that one up) is not recommend because there is so much compression in the top end. Is SLog2 any different in that regard?

- How are you mounting the Samurai on the FS700 and how cumbersome is that?

Lenny

The difficulty with Slog2 and the samurai is that you're having to cram a big contrast range into a tiny 8bit bucket of just 256 values. When you look at the Slog2 curve you can see that most of those bits are dedicated to values in the highlights (above 18% grey). So if you want to avoid problems associated with this log compression in the important mid range, you need to over expose everything then bring it down in the grade. Art Adams seems to say as much even when dealing with the raw: My Week Shooting 4K with an FS700 Prototype at DV Info Net

I'm using a cheap Chinese knock-off noga arm to mount the samurai either off the Camera's handle, or the handlebars on my half inch rails shoulder rig - I then use it instead of an EVF (although it's not that great a monitor!). In this set up it is in a useful place and sort of out the way (as it is in line with my head). I'm not sure how I'd mount it if I just wanted it out of the way.

[Leonard Levy]

Thanks again Colin. I did read the Art Adams article which was very informative as usual. My reading of it though was that he insisted he was ETTRing because he was recording RAW rather than compressed log especially in 8 bit.

"Generally it’s a bad idea to expose log footage to the right of the histogram because the nature of log formats is that they compress highlight detail, much the way our eyes do, to save space. Exposing to the right can push crucial detail into the highly compressed upper portion of the log curve. In the case of the FS700, though, you’re looking at raw data that has a log viewing LUT imposed on it, so exposing to the right is okay as the deck is still capturing raw without the log gamma encoding."

That's not to say you're wrong just that it doesn't seem to be what Art was saying. What you said about the curve also makes sense and you've been testing in the field. I wonder if the curve for sLog2 is much different than Slog. Alistair Chapman strongly warned against overexposing Slog when recording 10bit 422 to the Samurai, but the curves may be different. Interesting question.

Lenny

[Colin Elves]

Originally Posted by Leonard Levy

I wonder if the curve for sLog2 is much different than Slog. Alistair Chapman strongly warned against overexposing Slog when recording 10bit 422 to the Samurai, but the curves may be different. Interesting question.

Lenny

A comparison of Slog1 and Slog2 can be see here: Sony’s S-Log2 and Dynamic Range Percentages | CineTechnica

Looking at the curve it looks like Slog2 compresses the midtones a bit more, and this is confirmed by the table, as it places 18% grey at 32IRE vs 38IRE in slog.

Converting the 10bit code values in the same chart to 8bit you can see that the slog2 will use values 22 to 86 (44 values) for shadow detail (below 18% grey), values 86 to 145 (60values) for highlight details (up to 90% white) and values 145-256 (111 values) for the super whites.

And therein lies the problem with slog2 in 8bit: you're using just 1/6 of your data to record half of the 'important' scene data while half the values are dedicated to highlights above 100% reflectance! There just isn't enough bits of data dedicated to the shadow tones to record shadow detail with a nice gentle graduation to black, leading to a lot of banding. You do, however, get to cram a huge amount of extra highlight range above 90% white (about 3 1/2 stops) thanks to the judicious use of those 111 values above it - although they are also a little crushed.

On top of this you have the added problem that a lot of this shadow detail (the stuff that contributes to this large 14 stop range) is close to the noise floor making the footage quite noisy (personally I have similar issues with Cinegamma 4, which also uses these 'noisy' stops).

Because of these two issues, I've found it works better to overexpose by a stop or two: pulling the important shadow detail away from the noise and up into the part of the curve with a little more data values dedicated to tonal changes. In effect adjusting the camera's ISO setting to 500 from 2000. Although this means your shadows get crushed (when you pull them back down) the detail you do get is less noisy and has better gradations. The extra DR seems to remain, but you get a bit more in the shadows and lose it from the highlights. So, instead of having 7 stops of range under 18% grey and 7 stops over, you get 9 and 5 stops. There's actually a good visual representation of this on the Arri website: ARRI Group: ALEXA FAQ (under 'what happens when I change the Alexa EI setting?). Of course this means you lose a lot of this extra headroom that slog2 is supposed to give you, but, in practice you still seem to gain a fair bit relative to a properly exposed Cinegamma 1 (for some reason)

As I understand it, Alistair's argument for not over-exposing Slog is that you risk putting skin stones (by which he means Caucasian skin tones) up into the compressed highlight portion of the curve - meaning that when you bring those tons back down in post you lose a lot of the tonal range you would ordinarily expect, so it looks odd.

I'm guessing the same would be true of Slog2, although the extra half a stop of highlight range relative to slog1 possibly means that there is a bit more room to manouvre perhaps. Either way, over exposing slog2 seems to work for me, at least in the 8bit world.

Of course, you're still going to have problems because you're trying to cram 14stops of range into just 256 values, but these seem to be hidden better if you shift the extra stops into the supers whites where you'd expect the viewer to be less aware of tonal gradations anyway (as human vision is able to discern tonal changes in the shadows more than in highlights - hence 18% reflectance being the middle tone in a range form 0-100% reflectance!)

However, if you're working on a scene with a high contrast range detail you need to capture (e.g. A bright sunny day, or an interior day scene with an open window) an over exposed slog2 could be really helpful. But if you're dealing with a more narrow set of ranges you're much better off with a Cinegamma.

If course, a lot of this is just guesswork on my part! I'm still trying to get my head round it all.

Cheers,

Colin Elves
DP, london
Colin Elves - Director of Photography and Steadicam Operator, London, UK

[Leonard Levy]

Thanks Colin.
Your explanation and thinking seems very clear to me and makes sense. I would be interested to hear Alistair's take on the same subject.

[Colin Elves]

No problem. Although I'd take everything I say with a pinch of salt as I'm still trying to get my head round it all!