Grading Eterna 400T for HD, SD and Web

[Joseph Owens]

Wouldn’t it be loverly if WYSIWYG?

Resolve pretty much operates in Rec709 by default, unless you deliberately go away from it by slipping a LUT in the mattebox. Internal computer graphics previews are operating in sRGB, which is similar, but not exactly the same as 709. That is where external SDI Video I/O interfaces like Decklink come in — and there are emulation modes included in many grade monitors that allow you to see your values as mapped to various colorspaces. That is the one big real advantage over graphics displays, plus of course accuracy and reliability.

Linearizing your dpx files in Resolve really is asking for the value scaling — 0-1023 or 64-940?

As far as gamma characteristic goes, depending on whether the dpx files were digitized log or not, I would be surprised if anyone went to the extent of customizing an antilog for a specific stock — that would have to be tweaked for every batch. Its a bit like the so-called film log curves that some digital cameras provide — they’re not log at all, but simply flattened out to try to avoid clip/crushing the high-and low-lights, and its “up to you” to eye-ball it out, and usually, with a scope, its not a challenge, and a hand-built S-curve is the most efficient path for that. Ultimately, make it look good – that’s what we do.

And as the manual and various tutorials point out, the “log” controls in DaVinci primaries aren’t specifically for reverse log operations — they’re just a remapping of the overall controls that are pinned at different control points, which you can slide around (pivot).

jPo

“I always pass on free advice — its never of any use to me” Oscar Wilde.

[Francesco Krattiger]

Hey Joseph

Thanks for the information! It certainly helps and makes me think I’m more or less on the right track.

Would you mind elaborating a bit on your remark regarding the value scaling?

I looked up rec 709 black and white levels and from what I can tell the 64-940 range corresponds to
the black and white points in rec 709.

Am I correct in assuming that I need to watch my scopes and keep shadows above 64, if they are to retain
detail, while doing the same with the highlights with respect to 940?

Thanks again.

Francesco

PS: The footage was in fact digitized 10-bit log.

[Toby Tomkins]

Try using an Alexa log c (important: film matrix ON) to rec709 LUT and grade before it (either load the LUT as the final node in a chain or load it as a node in track mode). This should get you nice results. LogC is the same as cineon log which is probably what your scans are.

I’m shooting my own tests on 10 35mm film stocks and 4 digital cinema cameras (Alexa, red, f65, c500). Off the back of that I’m creating various LUTs, some of which you might find useful. Let me know if you’re interested and I’ll keep you posted.

Toby

[Mike Most]

If you’re going to use an Arri website generated LUT with a film original, you want the matrix OFF. The matrices that Arri offers for those LUTS is based on the Alexa camera’s particular characteristics and is designed to restore proper saturation and specific color values when using that camera, and with the assumption that the in camera matrix was off while shooting. What you want is a LUT that does not incorporate a matrix because a film scan already has proper values (it’s not the result of a Bayer pattern capture system, at least not at the camera end).

Second, LogC is NOT identical to Cineon log. It is slightly different in overall gamma and in specific areas of the curve. It is very similar to Cineon log, and can usually be used interchangeably with it if color grading is being done, but it is NOT the same (as opposed to, say, RedlogFilm, which IS identical to Cineon nog…).

[Francesco Krattiger]

Hey Toby, thanks for your input. I’ll try that, modified by what Mike said.

I’d certainly be interested in what you come up with over the course of your test,
so please do keep me posted.

Francesco

[Francesco Krattiger]

Thanks for the clarification Mike – and thanks for writing that article in the first place.

Francesco

[Francesco Krattiger]

I’ve a follow-up question: Where do I control the “film matrix” setting?
I can’t seem to locate it.

Edit: Never mind, I figured it out.

[Joseph Owens]

[Francesco Krattiger] “I looked up rec 709 black and white levels and from what I can tell the 64-940 range corresponds to
the black and white points in rec 709.

Am I correct in assuming that I need to watch my scopes and keep shadows above 64, if they are to retain
detail, while doing the same with the highlights with respect to 940?”

No, not really. Your dpx scans, if scanned full scale, will correspond to 0-1024 RGB and you should check that Resolve is mapping the same scalar overall 0-100 IRE range — if your output preview is 709, then Resolve should be doing the work for you, but only if it is correctly interpreting the source files. You should be aiming for 0-100 709 if that is your ultimate output, even though in Y’CbCr those values are 64-940. That is the nature of so-called “linear video”. This will be very obvious in an outboard SDI scope, where you will see either “floating” or “very crushed” blacks, and a stunted white level. If there is a scalar mapping error, it is not a linear fix as gammas will be proportionally incorrect across the grey scale.

jPo

“I always pass on free advice — its never of any use to me” Oscar Wilde.

[Toby Tomkins]

I think you want film matrix on (or configure the lut in the arri lut builder as if film matrix was on during recording of an Alexa log image)…Think of it like this, if you want to have an Alexa log image similar to a film scan, you would record with film matrix on, as specified on the arri website;

“Log C (film matrix on) applies a color matrix that makes the resulting image resemble a film negative scanned on an ARRISCAN. While this option somewhat reduces the color gamut in contrast to Log C (film matrix off), it provides an easy and fast way for colorists who are used to scanned negative, thus speeding up color grading. It is also a great option when combining ALEXA images with film originated images. The same caveats as for Log C (film matrix off) also apply.”

Also from what I understand LogC is the same as cineon, here’s a brief history I found;

“LogC stands for Cineon Log and is not an ARRI invention it is a Kodak invention created to be able to capture film scans across a wide dynamic range.

https://en.wikipedia.org/wiki/Cineon (https://en.wikipedia.org/wiki/Cineon)

And

“Arri adopted the use of the Cineon specification as a means of encoding image data in a log format for the ARRISCAN film scanner. This was implemented to properly allow the digital representation of the dynamic range that film delivers in a 10 bit logarithmic file format where each stop of exposure is mapped in equal intervals of digital code values, thereby creating a digital negative. The Cineon Format was developed by Kodak in the mid 90’s for use in the Cineon Fim Scanner. ARRI adopted it for use in the ARRILaser ARRIScanner, D20/D21, and the Alexa.

Therefore the name LogC is essentially an abbreviation of Logarithm and Cineon referring to the way ARRI has chosen to implement it’s log encoding of image data in the Cineon Format Specification and the corresponding characteristic curve. Just like SLog is Log encoding based on Sony’s implementation of log in order to optimize their camera’s output; and LogF is Log encoding based on the Filmstream format for the same reason. They are all characteristic curves developed to take advantage of the technology inherent in the camera systems that they come from. You can think of it in the same way that different film stocks offer different characteristic curves and therefore offer more or less dynamic range and tonality then other film stocks.

Truth is there’s no perfect way to ‘linearise’ your files (FYI that term can mean many things), technically you could go to a scene referred colour spaces and then use an ODT for a given colourspace, but you’d still need an RRT which is subjective in itself. Experiment with curves (make them smooth!) and different LUTs and find what works best for this project. There is no one size fits all. You were ‘limited’ with film print LUTs before, where you had to grade in front of that LUT (so you could stay in log to deliver log files for the film print), but now if you’re delivering in rec709, you’re not finishing in log, so do what you want as long as it looks good.

From the sounds of it you’re in need of a colourist with their own suite. There’s plenty out there. Make a friend. Collaborate.

Toby

[Mike Most]

>>Think of it like this, if you want to have an Alexa log image similar to a film scan, you would record with film matrix on…

That would be the case if you were starting with an Alexa generated image. The original poster specified that he is starting with a film image. You don’t use a matrix designed for a specific Bayer pattern camera if your original material is not generated by that camera.

It is also absolutely true that LogC, while similar to Cineon, is NOT identical to it. Not everything you read on the Internet is true, including Wikipedia.

I stand by my original post.