FS7: The matrix of color space possibilities on the Sony F series cameras

starcentral

starcentral

Veteran
My latest article tests the matrix of color space and gamut possibilities on the Sony F series cameras in conduction with the Sony LOOKS (Sony 3D LUTs) – something in general everyone needs to know in terms of the difference in color space and gamut inconsistencies between shooting modes (CineEI and Custom Mode) – truly a fundamental problem with these cameras.

http://www.hingsberg.com/index.php/2015/04/sony-3d-luts-test/
 
Last edited:
While interesting and I applaud your efforts my first reaction is... meh. I don't view it as a "fundimental problem". Kodachrome and Fuji don't create accurate color either. For me the idea is to know how it alters the color to create the emotions of your story. Surreal reality can seem more real than reality. If you know what I mean.
 
Duke M. said:
While interesting and I applaud your efforts my first reaction is... meh. I don't view it as a "fundimental problem". Kodachrome and Fuji don't create accurate color either. For me the idea is to know how it alters the color to create the emotions of your story. Surreal reality can seem more real than reality. If you know what I mean.
Click to expand...

I believe color accuracy is very important for art.

Imagine you are shooting a vampire movie and you or the director want to use red, the color of blood, strategically through the film to evoke the right emotion at the right time. In other words the color red and it's shades is very important for your story.

For a given scene, perhaps set in a classic 19th century manor, you request that the art department have red furniture upholstery, walls covered with different shades of red textile and other red accents, just to find out in post that what you got is a room that looks decorated for Halloween because your camera rendered all the reds in different tones of orange.

Color accuracy is about control of the palette. Once you are sure you have captured the colors you need or want, if you decide you can then produce the surreal reality. At least you won't have any bad surprises. I get your point about film stock accuracy. But surely it makes things easier if your camera is accurate. And while there is always the argument that you can correct it in the grade, why would we want to work "in the dark"?

Color accuracy is one of the aspects which separates high-end cameras from low-end ones and one of the reasons expensive cameras costs more. There is certainly a reason for this. Unless you are shooting black and white color accuracy is as important as dynamic range and above resolution in my opinion.

Without wanting to sound artificially philosophical, how could Picasso mix his colors and make sure he was laying down the correct colors on his canvas if he couldn't trust the colors? If we understand that color is a powerful tool in story telling the ability and means to control it should be paramount.
 
Last edited:
I can agree there are worse problems in the world if that's what you mean.

But otherwise to me it's a problem because you have 3 camera's out in the marketplace now with two operating modes who's color space are completely different. The F55 can "fake" the slog3/s-gamut3.cine color using a matrix preset, but this matrix preset is only available on the F55 and not the F5 or the FS7.

Given there are fundamental operating differences between CineEI and Custom Mode, it means that if you needed to, or preferred to for whatever reason mix the two modes you'll be giving either yourself or someone else a headache further down the post pipeline.

Worse anyone new to using or renting these cameras won't even know about this stuff, and again, nuclear coloured footage ends up on people's hands because of what they didn't know and ultimately Sony gets a bad wrap for it.

I'm hoping something in v6 firmware addresses this, but haven't heard anything yet myself.
 
Again there can be reasons why you need features from one mode over the other and vice versa. We've beaten the "mode" discussion to death in the past so I don't want to get into that again.

But here's a thought: let's not forget that LC709-A was developed and released officially by Sony to "clone" the REC709 color look of the Alexa. My test shows that when you apply the appropriate 3D LUT to each color gamut slog combination they do not produce the same result.

Even if we decide to agree that color gamuts are like different film stocks (which I completely do appreciate, understand and respect) why doesn't S-Gamut/S-Log2 in both camera modes look the same?

All this is quite a problem to many.
 
Dennis,

I can't begin to give you answers to your issue, but I can suggest a few areas which you might consider if you want to dig deeper.

Custom mode and CineiEI are designed with two fundamentally different workflows in mind, and the processing involved (in method and order) seems to be geared to each.

For a CineEI, post-based workflow calculations have to be strict and consistent, with things like simple (3x3 matrices on scene linear pixel data) colour spaces / gamuts and functionally defined and reversible log transfer functions / gammas. Getting there from raw data takes processing though. Custom mode is for basically finished pictures where looking good is more important than the strictures of data consistency. I have no idea of the processing order, but it may very well be different from or more multi-layered than in CineEI mode. Both take processing and that is a finite resource - I'm not too surprised that the preset matrices have the word 'like' attached.

At the most basic level, to my understanding the S-Log2 and S-Log3 LC looks came out at different times and I don't think that Sony link between them on any forums etc. It could just be down to changing feelings or refining what looks good / appropriate over time. Maybe it's just that an updated version of the S-Log2 would fix things. You could trying baking your own up in Catalyst Browse or LUTCalc.

ACES aims to bring disparate camera material into a unified colour space / transfer function, ie ideally all cameras should end up matching reasonably well. The functions which import the material are the input device transforms or IDTs, and the ones which Sony supplied AMPAS are available here:
Sony Vendor Supplied ACES IDTs

even if you aren't a programmer, it's actually not that painful to look at these, as they are basically just a couple of lines of equation converting from log to scene linear, followed by a matrix multiplication going from S-Gamut to ACES colour space. Simple because the colour space is designed to be simple to hand things off to post consistently. That folder only includes S-Log/S-Gamut and S-Log2/S-Gamut, and looking at the S-Log2/S-Gamut ones you can see that there are daylight and tungsten IDTs. It's subtle, but the colour space matrix changes with colour temperature / white point for situations where precision is a big deal. If you check out Arri's IDTs, there are dozens, with different parameters for different temperatures and different ISOs - they also mention 'matrices for nitpickers' ;-). I can't find a publically-available S-Log3/S-Gamut3(.cine) IDT to compare, but the Sony S-Log3 technical summary does mention this:

Current S-Gamut needs color temperature setting to select color conversion matrix. But S-Gamut3 is optimized to combine these color conversion matrix. The conversion matrix does not depend on color temperature any more. The ACES IDT for S-Gamut3 becomes simpler than S-Gamut.
Click to expand...

As I understand it, the Arri Amira gets around processing complexities of ISO, colour temperature and all the CDL controls by effectively generating an MLUT on the fly rather than recalculating all effects on each pixel. I could well be misinformed there - I can't offhand find the document which talked about that.

Canon's C300 IDTs are available on their website. I mention that one because whilst the transfer function (C-Log) is nice and simple and conventional, the colour space is far from consistent - it's tuned to look nice rather than be a simple gamut (Canon Cinema Gamut on the C500 is that). That makes the IDTs rather more complex and approximate. Canon's own C300 LUTs are all 1D. Custom mode may well behave more like this. The seemingly improved green-spike handling compared to early firmware on the F5/F55 doesn't disagree with that.

Anyway, long post and no answers, but if you want to jump down the rabbit hole - and it does seem to be something you have a strong interest in resolving maybe there are some things to look into. Also, I think it's important for anyone reading all this to remember that none of it is about errors or bugs, just trying to use cameras in something of an 'off-label' way for personal workflows; it's great that that is even possible,

Ben
 
Last edited:
You must log in or register to reply here.
Back
Top