Sigma Fp review and interview / Cinema DNG RAW

[paulinventome]

3 hours ago, CaptainHook said:

You can an idea by calculating the primaries from the matrices in the DNG since they describe XYZ to sensor space transforms and then plotting that. You would want to do it for both illuminants in the DNG metadata and take both into account. I haven't read every post as its hard to keep up, but I'm curious at to why you want to know?

 

I don't know what the native space of the fp is but we do get RAW from it so hopefully there's no colourspace transform and i want to know i am getting the best out of the camera. And i'm a geek and like to know.

I *believe* that Resolve by default  appears to be debayering and delivering 709 space and i *think* the sensor native space is greater than that - certainly with Sony sensors i've had in the past it is. There was a whole thing with RAW on the FS700 where 99% of people weren't doing it right and i worked with a colour scientist on a plug in for SpeedGrade that managed to do it right and got stunning images from that combo (IMHO FS700+RAW is still one of the nicest 4K RAW cameras if you do the right thing). The problem with most workflows was that Reds were out of gamut and being contaminated with negative green values in Resolve (this was v11/v12 maybe? I have screenshots of the scopes with negative values showing and so many comparisons!

I understand that the matrixes transform from sensor to XYZ and there's no guarantee whether the sensor can see any particular colour at any point and that the sensor is a spectral response not a defined gamut.

I also know that if i was able to shine full spectral light on the sensor i could probably record the sensor response. I know someone in Australia that does this for multiple cameras.

But within the Resolve ecosystem how do i get it to give me all the colour in a suitable large space so i can see?

If i turn off colour management and manually use the gamut in the camera RAW tab Resolve appears to be *scaling* the colour to fit P3 or 709. What i want to see is the scene looking the same in 709 and P3 except where there is a super saturated red (for example) and in P3 i want to see more tonality there.

I'm battling Resolve a little because i don't know 100% what it is doing behind the scenes and if i could get the DNGs into Nuke then i am familiar enough with that side to work this out but for some reason the sigma DNGs don't work!

thanks!
Paul

[CaptainHook]

On 4/1/2020 at 8:23 PM, paulinventome said:

But within the Resolve ecosystem how do i get it to give me all the colour in a suitable large space so i can see?

If you want "all the colour" with no transform, then with DNG's you can select "Blackmagic Design Film" for gamut/colour space (Colour Science Version/Gen 1) and as mentioned that is sensor space with no transform, only what you select for gamma. So it's the colour as the camera has captured it (I say camera rather than sensor since there are no doubt corrections applied before encoding the raw data). As I also mentioned though, this is not suitable for display and the expectation is you will transform it/grade it for monitoring purposes. I believe Digital Bolex recommended this workflow and then provided a LUT to transform from sensor to 709 for their camera back when they were still around. You will still be able to "view" the colour unmodified from the camera though. I'm just stressing (for the benefit of others) that the colour from a digital camera in it's native sensor space is not intended to be displayed this way, so you can't (shouldn't) really judge "hues, saturation", etc.

Some manufacturers like us (and many others) design a "working space" that is generally larger than 709/P3 and ideally a better starting place to manually grade from than sensor space but also not intended for final display. AFAIK Sigma has not done that so your options are either native sensor space or another documented colour space.

Quote

If i turn off colour management and manually use the gamut in the camera RAW tab Resolve appears to be *scaling* the colour to fit P3 or 709

I probably wouldn't say "scaling" myself, but yes Resolve will do a standard transform from (sensor to) XYZ to Rec.709 or P3. It's not clipped on the output of this step though so you can still recover data. This is what you would expect to happen if selecting 709 or P3 in the RAW tab.

Quote

What i want to see is the scene looking the same in 709 and P3 except where there is a super saturated red (for example) and in P3 i want to see more tonality there.

I'm assuming you are compensating for the change in 709/P3 on the display side here, but whether or not you'll see the differences you're expecting will be influenced all the way from the sensor response and what they do in camera, to right at the end on the display side and how well you can display P3 versus 709.

Quote

I'm battling Resolve a little because i don't know 100% what it is doing behind the scenes

I work in the camera team (different country to where the Resolve team are based) so my knowledge of the inner workings have come via discussions with them so I don't know 100% either as the code isn't visible to me, but we do develop Blackmagic RAW and the SDK in the camera team (Resolve uses the SDK almost the same as any 3rd party app does) and we share with them the camera colour science information we develop so that they can implement it into their pipelines (DNG and CST/RCM/etc). They don't need to do as much of that now though for our cameras since it's handled in the Blackmagic RAW SDK which we handle from the camera team side.

As for DNG processing, they for the most part follow the Adobe DNG spec when it comes to processing AFAIK https://www.adobe.com/content/dam/acom/en/products/photoshop/pdfs/dng_spec_1.4.0.0.pdf

You may also be interested in looking at the DNG SDK if you can understand code as that will give you an even more clearer idea of how DNG's are/should be interpreted.
https://www.adobe.com/support/downloads/dng/dng_sdk.html

[paulinventome]

On 4/2/2020 at 11:07 AM, CaptainHook said:

If you want "all the colour" with no transform, then with DNG's you can select "Blackmagic Design Film" for gamut/colour space (Colour Science Version/Gen 1) and as mentioned that is sensor space with no transform, only what you select for gamma. So it's the colour as the camera has captured it (I say camera rather than sensor since there are no doubt corrections applied before encoding the raw data). As I also mentioned though, this is not suitable for display and the expectation is you will transform it/grade it for monitoring purposes. I believe Digital Bolex recommended this workflow and then provided a LUT to transform from sensor to 709 for their camera back when they were still around. You will still be able to "view" the colour unmodified from the camera though. I'm just stressing (for the benefit of others) that the colour from a digital camera in it's native sensor space is not intended to be displayed this way, so you can't (shouldn't) really judge "hues, saturation", etc.

First just to say how much i appreciate you taking the time to answer these esoteric questions!

So i have a DNG which is a macbeth shot with a reference P3 display behind showing full saturation RG and B. So in these scene i hope i have pushed the sensor beyond 709 as a test.

Resolve is set to Davinci YRGB and not colour managed so i am now able to change the Camera RAW settings.

I set the color space to Black Magic Design and Gamma is Black Magic Design Film.

To start with my timeline is set to P3 and i am using the CIE scopes and the first image i enclosed shows what i see. Firstly i can see the response of the camera going beyond a primary triangle. So this is good. As you say the spectral response is not a well defined gamut and i think this display shows that. 

But my choice of timeline is P3. And it looks like that is working with the colours as they're hovering around 709 but this could be luck.

Changing time line to 709 gives a reduced gamut and like wise setting timeline to BMD Film gives an exploded gamut view.

So BMD Film is not clipping any colours.

The 4th is when i set everything to 709 so those original colours beyond Green and Red appear to be clamp or gamut mapped into 709.

So i *think* i am seeing a native gamut beyond 709 in the DNG. But applying the normal DNG route seems to clamp the colours. But i could just be reading these diagrams wrong.

Also with a BMDFilm in Camera RAW what should the timeline be set to and should i be converting BMD Film manually into a space?

I hope this makes sense, i've a feeling i might have lost the plot on the way...

cheers
Paul

 

 

[Devon]

On 3/31/2020 at 5:08 AM, paulinventome said:

Actually you don't really want to work in a huge colourspace because colour math goes wrong and some things are more difficult to do.  Extreme examples here:

https://ninedegreesbelow.com/photography/unbounded-srgb-as-universal-working-space.html

Thanks Paul! Thankful for you all here to help clear up questions! 

One more question regarding your link; I can’t seem to find out what the heck “unbounded sRGB” is. Is it a Linux color space? Interpolated sRGB? And why would anyone interpret a RAW file, and immediately convert to sRGB for editing??? 

So confused 😭

[paulinventome]

On 4/4/2020 at 5:00 AM, Devon said:

One more question regarding your link; I can’t seem to find out what the heck “unbounded sRGB” is. Is it a Linux color space? Interpolated sRGB? And why would anyone interpret a RAW file, and immediately convert to sRGB for editing??? 

Hi Devon, i missed this sorry.

Unbound RGB is a floating point version of the colours where they have no upper and lower limits (well infinity or max(float) really). If a colour is described in a finite set of numbers 0...255 then it is bound between those values but you also need the concept of being able to say, what Red is 255? That's where a colourspace comes in, each colourspace defines a different 'Redness'. So P3 has a more saturated Red than 709. There are many mathematical operations that need bounds otherwise the math fails - Divide for example, whereas addition can work on unbound. There's a more in-depth explanation here with pictures too!

https://ninedegreesbelow.com/photography/unbounded-srgb-divide-blend-mode.html

Hope that helps?

cheers
Paul

[Devon]

3 hours ago, paulinventome said:

Unbound RGB is a floating point version of the colours where they have no upper and lower limits (well infinity or max(float) really). If a colour is described in a finite set of numbers 0...255 then it is bound between those values but you also need the concept of being able to say, what Red is 255? That's where a colourspace comes in, each colourspace defines a different 'Redness'. So P3 has a more saturated Red than 709. There are many mathematical operations that need bounds otherwise the math fails - Divide for example, whereas addition can work on unbound.

I think I’m getting there! Thanks so much Paul! I’ll keep reading about color management and  scene referred workflows!

[CaptainHook]

On 4/4/2020 at 12:45 AM, paulinventome said:

So i *think* i am seeing a native gamut beyond 709 in the DNG. But applying the normal DNG route seems to clamp the colours. But i could just be reading these diagrams wrong.

Its to be expected with a gamut transform to not see colours outside of that gamut, that's sort of the point of it. How you map colours from outside the target gamut into it is another thing entirely. I'd be wary of what the CIE scopes show though.

Quote

Also with a BMDFilm in Camera RAW what should the timeline be set to and should i be converting BMD Film manually into a space?

The timeline space sets what space the data is transformed into for processing through nodes etc in RCM. Personally I would leave this as 709/2.4 because the tools were originally designed to work that way and I'm used to how they respond but I also don't really use RCM. I would probably not use the CIE scopes for actual grading either and just use the vectorscope otherwise it may be confusing.

Yes you need to convert BMDFilm Gen 1 into a display/output space but unless you have information from Sigma on how to do it you won't be able to do it in a "technically correct" way in Resolve and will have to manually grade it.

[Andrew Reid]

It is a shame to users battling Resolve here with this camera.

If Resolve is out there as a serious NLE to replace Premiere, and it is... I frankly expect universal compatibility and I expect far more from Blackmagic to work with companies like Sigma in providing that better level of recognition for their RAW files.

We all know why they don't.

Blackmagic clearly sees RAW on the Fp as a threat to their camera division.

This thread is getting really bogged down by the way with long running and incredibly dry technical stuff.

I would like to see the conversation get more broad as there is a lot more to the Sigma Fp than just Resolve.

Maybe it's time to split the topic into one for the camera, and one for Cinema DNG in Resolve.

[Chris Whitten]

For the record, I have been very happy using the FP in Resolve, either using the Cinema DNG default or BMD Film setting in the camera drop down on the colour page.

I actually haven't encountered anything jarring or defective. It's been super easy to drop the footage in, grade it and export.

If you are going to stress test Resolve with your FP footage, then yes, I'm sure you can find fault. But just as an editor and colour grader for normally balanced footage out of the FP, using 12bit or 10bit 4K I think it's absolutely fine.

In order to get the nasty blotchy colour shift we were discussing earlier I had to take perfectly well balanced footage and increase the gamma and gain way beyond anything I would need.

 

[rawshooter]

47 minutes ago, Andrew Reid said:

It is a shame to users battling Resolve here with this camera.

If Resolve is out there as a serious NLE to replace Premiere, and it is... I frankly expect universal compatibility and I expect far more from Blackmagic to work with companies like Sigma in providing that better level of recognition for their RAW files.

Naah, it's the opposite - Sigma doesn't provide a Log profile and corresponding color space yet (we have to wait till firmware 2.0 in summer), and doesn't provide any LUTs, Aces IDTs or other transformation functions for the fp's CinemaDNG so that in the meantime, we're limited to grading the material in Rec709 although the camera has a wider gamut/more color.

It's likely because Sigma doesn't have any previous experience with video and digital cinema while the stills photography world still lives in sRGB (=almost the same as Rec709) or Adobe RGB at best, and thinks that paper prints (with their 5 stops dynamic range) are the Nirvana of image quality...

[Chris Whitten]

Phase One has also never supported a Sigma camera in Capture One.

They are niche products unfortunately.

[rawshooter]

Btw., the limitations and guesswork regarding the right color spaces for the CinemaDNG recorded by the Sigma fp apply to any software that can process CinemaDNG...

[Andrew Reid]

Resolve was made for Cinema DNG.

Of course it works.

Blackmagic are implying it is a pain to use the Sigma Fp's raw with Resolve. Of course it isn't.

They just don't really want it to be a thing or to support it in any way. Doesn't take a genius to work out why.

[rawshooter]

32 minutes ago, Andrew Reid said:

Resolve was made for Cinema DNG.

Of course it works.

Blackmagic are implying it is a pain to use the Sigma Fp's raw with Resolve. Of course it isn't.

They just don't really want it to be a thing or to support it in any way. Doesn't take a genius to work out why.

With all due respect, the matter is more complicated:

DNG (whether stills DNG or CinemaDNG) contains internal color matrixes. If a camera manufacturer doesn't provide raw specifications - like Sigma unfortunately does - then Raw developing software (no matter whether it's Lightroom, RawTherapee, Resolve, After Effects...) just interprets the raw DNG data based on the internal color matrixes. That is not optimal but works good enough for photographers who usually work in the sRGB or in the AdobeRGB color space.

However, as film people, we're now used to work in wider color spaces - making proper use of the 10bit/OLED/DCP etc. screens that increasingly replace legacy 8bit/Rec709 HDTV - such as Aces, P3, Rec2020. 

However, this is the point where DNG without additional camera manufacturer support fails, and where NO software, no matter whether Resolve or anything else, can properly figure out how the RAW color values should be interpreted.

So, in this case, it's upon Sigma to step up and help manufacturers across the board (no matter whether Adobe for AfterEffects, Blackmagic for Resolve, Assimilate for Scratch, Apple for FCPX...) and share information about the fp's native color space if they want to promote the fp as a serious cinema camera.

Otherwise, we can't really make use of the full color spectrum captured by the fp, but will be limited by the Rec709 color space even when grading RAW.

[paulinventome]

7 hours ago, rawshooter said:

Otherwise, we can't really make use of the full color spectrum captured by the fp, but will be limited by the Rec709 color space even when grading RAW.

I'm with you on all this but i don't know what you need sigma to supply though...

The two matrixes inside each DNG handle sensor to XYZ conversion for two illuminants. The resulting XYZ colours are not really bound by a colourspace as such but clearly only those colours the sensor sees will be recorded. So the edges will not be straight, there's no colour gamut as such, just a spectral response.

But the issue (if there is one) is Resolve taking the XYZ and *then* turning it into a bound colourspace, 709. And really i still can't quite work out whether Resolve will only turn DNGs into 709 or not. So as you say, for film people, P3 is way more important than 709. And i am not 100% convinced the Resolve workflow is opening up that. But i might be wrong as it's proving quite difficult to test (those CIE scopes do appear to be a bit all over the place, and i don't think are reliable). 

As i mentioned before i would normal Nuke it but at the moment the sigma DNGs for some reason don't want to open.

Now i could be missing gaps here, so i am wondering what sigma should supply?

cheers
Paul

[Andrew Reid]

No this is not the point where DNG "fails".

I am happy to grade in sRGB just as I am happy to grade any sRGB 14bit RAW photo file in Adobe Camera Raw.

I never open a RAW photo in Lightroom and think "damn this limited colour gamut, it looks terrible, what a failure..."

You guys really need to come into the real world and stop bitching.

[CaptainHook]

22 hours ago, Andrew Reid said:

Blackmagic are implying it is a pain to use the Sigma Fp's raw with Resolve. Of course it isn't.

I'm sorry but where is "Blackmagic" implying that?

If it's me, my only intention is to help some understand how to get the best out of the DNG workflow in Resolve with Sigma FP files. I'm here on my my own time including the weekend to help people, and if my intention was otherwise it would be better for me to let people stumble instead of help and not waste my time. We work with companies like Sigma all the time and I personally think the Sigma FP is a great little camera and told the product manager so a couple of times in person.

I will leave you to it. People know where to find me if they need help.

[paulinventome]

14 hours ago, Andrew Reid said:

You guys really need to come into the real world and stop bitching.

I think that's pretty unfair without knowing what we do.

Most of my work is P3 based. Ironically the vast majority of apple devices are P3 screens, there's lots of work targeting those screens. 2020 is taking hold in TVs and is becoming more valid and of course HDR is not sRGB. Any VFX work is not done in sRGB. Any professional workflow is not sRGB in my experience.

All we're doing is taking quarantine time and being geeks about our cameras. Just like to know right. My main camera is a Red, not a sigma fp, but the sigma is really nice for what is it. It's just some fun and along the way we all learn some new things (i hope!)

cheers
Paul

[Andrew Reid]

Most people come to learn more broadly about the camera, but we've had pages and pages of this now and it's really dragging on. It should be a separate topic.

CaptainHook please forgive me, it's nothing personal. I just absolutely hate the company you work for.

[Lars Steenhoff]

15 hours ago, Andrew Reid said:

No this is not the point where DNG "fails".

I am happy to grade in sRGB just as I am happy to grade any sRGB 14bit RAW photo file in Adobe Camera Raw.

I never open a RAW photo in Lightroom and think "damn this limited colour gamut, it looks terrible, what a failure..."

You guys really need to come into the real world and stop bitching.

Hey Andrew I'm actually happy with all the technical talk here, because with the input technical users give to sigma,

It is improving things in the firmware, like the flickering and the exposure preview problem.

Please let us bitch a bit more until those issues are all sorted. 😎

And yes lets make separate tread about resolve DNG no problem with that.