Sigma Fp review and interview / Cinema DNG RAW

[paulinventome]

On 10/12/2019 at 8:53 PM, Lars Steenhoff said:

It seems like you can manually switch between crop and full frame.

We will see if they downsample the raw. ( you can downsample raw, just downsample each channel separate )

or if they use some form of binning

You can't really downsample RAW, the reconstruction method don't work as well as you'd get aliasing in your downsampled channels and loose the ability to interpolate the missing pixels surely? 

6K -> 4K Binning, not sure how that would work without introducing issues. Keep all the greens and dump some other colours? Seems like a lot of work when that processing work could be better spent dumping the data out. During the design they knew the sensor size so why not ensure the rest of the camera can handle the datarates, even if it was a little bigger? Sigma have no video camera line to eat into, they're perfectly positioned to disrupt.

I can't believe that a few weeks away from supposed delivery the most basic question of whether all the movie modes are crop isn't answered!

cheers
Paul

[Brian Williams]

Here's something of a review, with samples-

https://www.photographyblog.com/reviews/sigma_fp_review/hands_on

The videos look like crap, but again, pretty sure its the photographer. Wish they'd shot it in raw.

 

Also, this- 

There was clearly some sort of event today or yesterday, so guessing more footage and reviews will start to trickle out any day now.

[BTM_Pix]

39 minutes ago, Brian Williams said:

Here's something of a review, with samples-

https://www.photographyblog.com/reviews/sigma_fp_review/hands_on

 

What would posses anyone with any semblance of quality control of their output to publish images like these ?

I get that there is a race to get things out first amongst those type of review sites but fuck me.

As for the video, the least said the better.

[Sid]

I thought the video was quite interesting as an indication of rolling shutter and also the helicopter rotors at a distance.  I expected to see a real "swirly-bird" but was pleasantly surprised.  Anyone else have any views on rolling shutter based on that ?

[Alt Shoo]

https://***URL not allowed***/sigma-fp-camera-first-impression-sample-footage-and-lab-test-results/

 

 

 

[jase]

Sounds rather disappointing, especially if you take into consideration that cinemad5d reviews tend to be very "friendly". EIS only when shooting MOV, yet MOV only has 8 stops? Moire? Mh, at this stage I wont be an early adopter although I really would like to love this camera. Hopefully upcoming firmware updates address some of those issues.

[Lars Steenhoff]

24p 4k uncompressed cdng

at full frame and crop

Its still impressive!

you take a 45 mm lens and you also have a 67 mm at the same time.

I can shoot most things I want with just camera and that lens.

Add a small wide angle and your set

 

[Andrew Reid]

1 hour ago, Lars Steenhoff said:

24p 4k uncompressed cdng

at full frame and crop

Its still impressive!

you take a 45 mm lens and you also have a 67 mm at the same time.

I can shoot most things I want with just camera and that lens.

Add a small wide angle and your set

 

Full frame 4K raw for $1800 in small stills camera.

Unique. What other cameras do that?!

It's like an official Magic Lantern with bullet proof reliability.

Very happy about the price, it's Sigma being typically competitive there.

[seku]

yep, and it even seems to have ok-ish rolling shutter @ 20ms. Maybe i will get one and pair it with a bolex 1.5x or Kowa C35. Now we just need a l-mount to EF/Nikon adapter with integrated VariND, and this little cam will pack a hell of a punch.

The only thing i am worried about is the flickering issues, and wondering if the sensor is fully read out, or is line skipping/skipping around the sensor.

About the 300mbyte/sec CDng RAW: i plan to run it through Slimraw, at 1:3 to 1:7 compression. This gives me 5dmk3 RAW filesizes, at 4K! If RED won't allow compression incam, i will compress it out of cam :3

Edit: did i mention the losslessly or lossy CDng opens fine in Resolve?

[Lars Steenhoff]

You can download two cdng stills from cinema 5d , one 8 bit and one 12 bit.

I played with them opening the dng directly in photoshop raw and it looks amazing.

You can sharpen all you like or turn it to 0, you can change exposure and white balance, and even the adobe profiles look great for getting a quick grade

 the dynamic range is great !

And the best thing for me,  no compression, so really nice filmic grain structure.

[seku]

9 minutes ago, Lars Steenhoff said:

You can download two cdng stills from cinema 5d , one 8 bit and one 12 bit.

Yeah, reading that comment thread eagerly! i cannot wait to get home and work on the still in Resolve!

[Lars Steenhoff]

Also samsung t5  ssd 2 terabyte is only 299 usd.

And after you get home with the footage you can use slimraw to compress the dngs.

https://www.slimraw.com/

You can use the ssd directly as the source and compress while you copy to another drive.

 

 

[Llaasseerr]

I took a look at the CDNGs in Resolve just with a basic ACES setup and they both seem pretty decent - even the 8 bit one which is interesting, because an 8 bit linear image should be practically useless. Maybe it's got a log curve under the hood? That lakeside image has a lot of dynamic range variation between highlight and shadow. It might just be the right kind of image to camouflage potential issues. I'll try a non-ACES workflow where I would just interpret to Alexa LogC/AWG. From there you would write out ProRes or something.

I like cinema5d, but if I understood their review correctly they seem to be saying the camera's internal picture profiles are influencing the raw recording. I don't see how that is possible, those profiles would be only baked into the h264s. They are also advocating for a log profile to interpret the raw image as well as an option for the baked h264 footage. Resolve should do a decent job of translating the DNG frames based on metadata, then if you want you can convert to your chosen log profile with CST nodes or similar. If Sigma did create a log profile and gamut, and put out a white paper then that would be nice, but I'm assuming the DNG metadata currently there isn't garbage because the images seem to look okay.

A known, published log profile and gamut is essential though when recording raw for monitoring over HDMI - assuming you can add a custom LUT to your monitor. That way, you can get a decent "one light" match to what you will eventually see in Resolve after ingesting your raw footage and applying your LUT.

 

[cpc]

38 minutes ago, Llaasseerr said:

I took a look at the CDNGs in Resolve just with a basic ACES setup and they both seem pretty decent - even the 8 bit one which is interesting, because an 8 bit linear image should be practically useless. Maybe it's got a log curve under the hood?

It is not linear. I haven't looked at the exact curve, but it does non-linear companding for the 8-bit raw. The 12-bit image is linear.

38 minutes ago, Llaasseerr said:

I like cinema5d, but if I understood their review correctly they seem to be saying the camera's internal picture profiles are influencing the raw recording. I don't see how that is possible, those profiles would be only baked into the h264s.

The DNG spec allows for color tables to be applied on the developed image. The Sigma sample images do include such tables. No idea if they replicate the internal picture profiles though. AFAIK, only Adobe Camera Raw based software (e.g. Photoshop) honors these. Unless Resolve has gained support for these tables (I am on an old Resolve version), it is very likely that the cinema5d review is mistaken on this point.

[Llaasseerr]

1 hour ago, cpc said:

It is not linear. I haven't looked at the exact curve, but it does non-linear companding for the 8-bit raw. The 12-bit image is linear.

The DNG spec allows for color tables to be applied on the developed image. The Sigma sample images do include such tables. No idea if they replicate the internal picture profiles though. AFAIK, only Adobe Camera Raw based software (e.g. Photoshop) honors these. Unless Resolve has gained support for these tables (I am on an old Resolve version), it is very likely that the cinema5d review is mistaken on this point.

Thanks for clarifying there is some kind of curve on the 8 bit image - nice to know they thought about that!

I'm not sure what you mean as far as color tables in DNG (will need to give the spec another look) but I'm mainly referring to Resolve interpreting colour matrices stored as DNG metadata - which it does. So that should be the crux of the colour transform decisions it's making on the raw image.

Resolve actually does the best mainstream job at interpreting a Cinema DNG image because it puts it into a high dynamic range space with no highlight clipping if you have the Resolve project set up correctly, and it exposes it more or less correctly for middle grey.

[seku]

2 hours ago, Llaasseerr said:

I like cinema5d, but if I understood their review correctly they seem to be saying the camera's internal picture profiles are influencing the raw recording. I don't see how that is possible, those profiles would be only baked into the h264s. They are also advocating for a log profile to interpret the raw image as well as an option for the baked h264 footage. Resolve should do a decent job of translating the DNG frames based on metadata, then if you want you can convert to your chosen log profile with CST nodes or similar. If Sigma did create a log profile and gamut, and put out a white paper then that would be nice, but I'm assuming the DNG metadata currently there isn't garbage because the images seem to look oka

 

Iirc they just put the RAW stuff into an SRGB/gamma 2.6. Which seems quite ... unusual. whenever i do raw stuff in resolve, i go for blackmagic film, as flat as possible, then onwards to cineon and other log stuff. as i understand they tried to emulate the RAW to something they can compare to their other measurements.

For me, RAW needs special treatment : ETTR, normalisation, and then noise reduction (i love neatvideo for that). it's difficult to compare the fp's dynamic range, because all other cams lower the noise floor already. you need to do your part in RAW yourself.

[Llaasseerr]

46 minutes ago, seku said:

Iirc they just put the RAW stuff into an SRGB/gamma 2.6. Which seems quite ... unusual. whenever i do raw stuff in resolve, i go for blackmagic film, as flat as possible, then onwards to cineon and other log stuff. as i understand they tried to emulate the RAW to something they can compare to their other measurements.

For me, RAW needs special treatment : ETTR, normalisation, and then noise reduction (i love neatvideo for that). it's difficult to compare the fp's dynamic range, because all other cams lower the noise floor already. you need to do your part in RAW yourself.

My "raw workflow" is just expose with a light meter, also get a grey card and then import to Resolve. In the raw settings I may use highlight reconstruction and do an exposure adjustment. Generally from there, if not working in ACES I do a CST to log (Cineon or Alexa LogC) then put the PFE lut on it to see how it will look before any log space grading.

Edit: to be clear the PFE lut is the last thing in the chain but I'll put it on before grading (time-wise, not in the node graph). This gives me a nice quick one-light and should match the on-set monitor.

I didn't see anything about these images that would make me deviate from that. They seem like regular raw images to me once they are in Resolve, but I'll have to see when more becomes available.

[buggz]

Maybe Fast CinemaDNG Processor would be good to use?

https://www.fastcinemadng.com/

I first heard of it today.

[Lars Steenhoff]

8 hours ago, buggz said:

Maybe Fast CinemaDNG Processor would be good to use?

https://www.fastcinemadng.com/

I first heard of it today.

If you have a Nvidia card, yes you can use it

[paulinventome]

14 hours ago, cpc said:

It is not linear. I haven't looked at the exact curve, but it does non-linear companding for the 8-bit raw. The 12-bit image is linear.

 

Yes, i delved a little deeper and the 8bit DNGs have a linearisation table attached to them.  I've been opening these up in RAW Digger which gets me the values before any debayering of colour work. I need to see if i can get the table out but it's a good sign.

Having said that by the time the image is debayered into a working colourspace that 8 bit source of thee RGGB channels is tonally more spread meaning that effectively the end result will have more tonality than 8 bit implies, say compared to an 8 bit movie. I hope that makes sense?

Paul