F55/F5 shooting 2 stops over zebra question - Allister

[Annen James]

I watched Allister's F55/F5 presentation at CineGear2013 video and loved it. Very informative. I am shooting an upcoming series and had a few questions on his advice of shooting 2 stops over in Raw, native ISO, and setting the EI-ISO to 800 on the F5. He mentioned that he then set your Zebras at 100%, seeing just a tiny amount to get proper exposure for landscapes and better grading.

My question is, with that Raw setting, will using zebras at 70% give a good exposure for faces? What is suggested for proper exposure. Thank you for your suggestion.

[Doug Jensen]

EDIT: I'm removing my comments because they were take the wrong way. Sorry if I offended anyone.
I was merely trying to get Annen to think about the issue of proper expsosure and figure out why 70%, or any other number, would be incorrect to use across the board for all faces. I can either just tell people stuff, or I can try to get them to figure it out themselves so perhaps they will better understand WHY and actually learn something instead of just following instructions someone else told them. If it didn't come across that way, I apologize.

Cees, the pet peeve I was referring to in my original post was not regarding people asking the question. Just the opposite, if you don't ask about something, how will you learn?? And I don't mind repeating my advice as many times as necessary to teach new students, or people who didn't understand it the first time around. My pet peeve is with people who promote the myth of using 70% for skintones. It's bad advice and that's what I meant by it being a pet peeve of mine.

Oh, and while I'm at it, what's the deal with the personal attack on me? Your comments are directed at how I said what I said, and not on the topic at hand. I think Annen would be more interested in hearing your advice for setting exposure on faces rather than commenting on my personal character or teaching methods.

[Cees van Kempen]

How can a sincere question be a pet peeve for someone who is a teacher? The fact that other people have asked the same question ore made the same mistake is not relevant. Because anyone who askes a sincere question does not necessarily know what other people have done/asked. And think of it this way: suppose you have to explain everything only once and everyone knows for ever? Where would that leave the market for teachers like you? It would sound more respectful to give your opinion without showing your annoyance anyway.

[Peter Corbett]

I agree. It's rude and dogmatic. I too am about to start shooting on my R5 and I'm open to all reasonable opinions regarding correctly exposing in raw.

[Annen James]

So I'm in a run and gun, reality tv world, I can't stop to pop off a light meter, or take a sport reading. What prey tell should one use to judge exposure for skin tone. I guess the 70% is just out of the question and absurd. Any advice?

[Leonard Levy]

Annan's initial question was a good one. I would be interested in seeing this conversation turn away from whether or how one chooses to handle zebras, to how best to expose sLog2.
I haven't een Alistair's presentation. Is he recommending generally overexposing sLog2 rather than using Sony's recommendation ( as I understand it 32% is neutral gray)? Is that on the web somewhere. I know with original sLog he recommending against overexposing .
On the FS700 forum Colin Elves talks about experimenting with overexposing sLog2 and finds his shadow detail is much cleaner and he's not losing much highlight. He was recording 8bit to the internal codec though.

By the way I use zebra's all the time, but just as a rough idea that my exposures are in the ballpark when I am running and gunning. I set them at 60 though finding 70 to high for caucasians in HD. Its a very useful tool as long as its used judiciously realizing that every face is different and every scene may want a different exposure. By the way my use of zebras at 60 has been mostly for 709 gammas not log and even cinegamas I tend to sink a little lower.

The thing about zebras is that they are always just a guide. How much should you allow on a face, where on the face are you seeing them , etc.? Its never clearcut anyway.

[Doug Jensen]

That exactly why I would never use zebras on faces. People can justify it any way they want to, but zebras on faces ultimately just comes down to guessing, and I prefer to be more precise than that. Exposure should be clearcut.

[Leonard Levy]

Duly noted, but after 35 years I know how to expose. We all have our own ways of working.

Can anyone explain why Alister suggested exposing 2 stops over for Raw? Is his presentation on the web anywhere? I'm guessing 70% would be a little high in any case.

[Dave Sperling]

My guess is that overexposure is suggested because with linear (16bit RAW) data, half of your potential grading differentiation (numbers) go into the brightest stop of exposure, then the next brightest stop takes half the remaining grading numbers, etc. Thus the 2 brightest stops of highlights take up 75% of your grading numbers... your 3 brightest stops take up 87.5% of your grading numbers... Your really dark areas have fewer numbers to work with, and are tha parts that really benefit from the overexposure.

You aren't dealing with the high end roll-off compression of the log formats, so overexposing won't send your faces into a 'numerically over-compressed' range where they would be difficult to time. In fact, the overexposure gives you more timing number granularity to work with...

But --
If you have a brighrtness range in your subject that extends all the way up to the maximum accepted by the sensor, then obviously you shouldn't be overexposing if you want to keep detail in your highlights.

However, if your subject doesn't have super-bright highlights, some overexposing will slide the exposure data up the numeric scale to an area where there are more numbers available for grading. Plus of course you will have more grading options for you dark areas (where you have far fewer numbers to work with.

About a month ago I attended a Cine-Alta meeting in NYC which dealt with grading, and as the colorist worked with the 4K RAW linear data we came to the conclusion that some overexposure with timing down in post seemed to be a good thing based on the footage we saw, giving us more control in post, assuming that the overexposure was not pushing us against the highlight detail limit on either the sensor or the recording.

However, working with exposure in this way is somewhat counterintuitive for those of us who are used to trying to get the exposure/output exactly the way we want to see it. Also, using this kind of 'exposure by protecting highlights' could easily make your proxy files very uneven -- and what if you have a bright window behind an actor in one direction (need to protect the highlights) and a dark brick wall behind the other actor in your reverse (trying to get smoother details in the dark areas)? Would you really have time on set to create custom LUTs for each shot?

[Leonard Levy]

Thanks Dave,

I think I'm finally beginning to get this. Am I right then that exposing RAW is very similar to shooting with the RED, where you don't worry about compression in the highlights because there isn't any. You are more concerned with noise in the darks and consequently ETTR (expose to the right) is a general rule of thumb. Thus one of the main exposure tools is the histogram - not my favorite tool in general but great for showing what how bright your highlights are.

As you note of course you always have to use your judgement exposing a full scene. Often its better to find a consistent exposure to assist the final grading and consistency of the scene.

I'm also interested though in how people are exposing Slog2. Here you aren't dealing with RAW so you do have compression in the highlights. However with mid-grey at 30% it seems like it would be easy to get noisy darks. I know for original sLog Alister Chapman recommended against raising the exposure too much because the compression in the highlights can look artificial - esp on faces, but even on general imagery.
Is it as bad in sLog2? The noise would seem to be potential worse.

I've been conversing with Colin Elves on an FS700 thread where he has been testing sLog2 on the internal 8 bit recordings ( I know its heretical) and he's been using ETTR as a guide and hasn't noticed compression problems:

http://www.dvinfo.net/forum/sony-nxc...0-samurai.html

Lenny

[Alister Chapman]

Full details of my exposure method are here:

http://www.xdcam-user.com/2013/05/ex...mw-f5-and-raw/

The F5/F55 in raw mode have an abundance of headroom. If you shot at the native ISO and expose "correctly" then you will have good exposure, but your almost certainly not using the top two stops of the cameras available range.

If you use CineEI to reduce the brightness of the viewfinder image or video out then this encourages you to open up the iris. The raw recordings continue at the native ISO, so these become brighter as a result and you maximise the recording range. In addition when this footage then goes in to post you will often be grading down rather than up, so this means your noise will decrease. It also gives you greater shadow dynamic range, although that comes at the expense of highlight range.

This method is only really suitable for raw. You can use it with SLog2 but you will be pushing your mids well into the very highly compressed part of the curve and they won't look as good post grade as they would exposed normally.
What you have to understand with log is that the camera is rated at 2000 ISO (F5) when you use middle grey at 32%. An F5 at 2000 ISO is not noisy, whether you expose S-Log2 2000 ISO at 32% or Standard Gamma 2000 ISO at 40% the noise will be more or less the same after grading. Just because the SLog2 image may look dark it doesn't mean it's noisier, its just being recorded at a lower voltage level and once restored to "normal" levels will have the same noise as a normal gamma of the same ISO. Of course anyone that owns an F5 will know though that the standard gammas and Hypergammas are rated at 800 ISO, so exposed correctly these will show roughly half the noise of SLog2 exposed correctly.