Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

CF Weapon ARRI XT common colourspace comparison - seeking advice

Hi Brett, I know the focus is linear light, but if you are going all out try all 3 OLPFs and explore things beyond "just" what we consider the Base ISO. Specifically ISO 1280 with the Standard and Low Light OLPF. There's something interesting around that area. Testing the usable ISO range on both of these cameras reveals a lot of the story.
 
Hi Phil, great to get your input on the thread.

This test is a 'one OLPF' visual comparison so I chose not to do all three filters (even though I had access to them) because the multiplication for this test (2 cameras x 2 workflows) would mean 16 images per scene for 4 'cameras'.

And since this is the standard filter, it's the logical choice (and is garnering a lot of positive comments).
 
Understood. Standard is a good baseline. It is indeed interesting to see the variance in how the highlight roll-off/retention at each ISO differs from the 3. Averaging the noise is a good way to do sort of a side-by-side comparison and that's one of the interesting stories I think with these two systems. Looking forward to seeing your results.
 
Phil Holland said:
Understood. Standard is a good baseline. It is indeed interesting to see the variance in how the highlight roll-off/retention at each ISO differs from the 3. Averaging the noise is a good way to do sort of a side-by-side comparison and that's one of the interesting stories I think with these two systems. Looking forward to seeing your results.
Click to expand...

Yeah cheers Phil :-)

Could you elaborate about noise averaging, perhaps it's something I can include in the comparison.
 
Great, never seen a test on the color space. outstanding ;)
Waiting ..
for the rest!
As for Phil, please i hope to see at least a test with ALexa 800 iso and Weapon to 1000/1280 ISO, this usually fit very well on set and not . Otherwise we would always be the same result where ALexa wins on highlights and Weapon will be less noisy with the OLPF standards.
Another point would be to add A.D.D to see what happens.
It would be appropriate to share the original files so that everyone can try and get your same result.
 
Brett Harrison said:
Starting with debayers in 16bit-HDR ACES primaries with gamma curve 1.0 (linear) and arriving at 4 outputs per scene:
RED DATA -> RED TRANSFORMS
RED DATA -> ARRI TRANSFORMS
ARRI DATA -> RED TRANSFORMS
ARRI DATA -> ARRI TRANSFORMS​
Click to expand...


Once you have gone to ACES, or scene referred linear light, why use different LUTs for each camera's final output? The same LUT or RRT should be used for both so they are on equal ground. Could be RED, ARRI or some other RRT. Software like www.colorfront.com would be perfect for this.
 
Stacey Spears said:
Once you have gone to ACES, or scene referred linear light, why use different LUTs for each camera's final output? The same LUT or RRT should be used for both so they are on equal ground. Could be RED, ARRI or some other RRT. Software like www.colorfront.com would be perfect for this.
[/FONT][/COLOR]
Click to expand...

Hi Stacey,

With the 4-way comparison you do in fact get the same LUT for each camera x2 involving both manufacturers' LUTs.

The purpose is to see what the contrast curve, applied to its native colour space, as well as the other's, looks like for the native data, as well as the other's.

If this is daft let me know.

I'm also considering flowing everything through the one gamut mapping to BT/ITU 1886 in Mistika after the wide gamut/log step, to keep things a bit simpler and just do a 2-way comparison, if that makes any sense, still using the LogCtoVideo and RG4 contrast curves (1D LUTS) as a final step - that way everything stays floating point until the contrast is applied.

Would love to get your thoughts.
 
Last edited:
:lurk5::lurk5::lurk5:
 
Hey All,

I'm working away at this comparison between gigs, and it's gotten a bit grander in the making to really show the difference between the two cameras with all the variables: camera, log gamut, rec709 gamut, log curve, display curve - which is a lot of variables to wrangle once they've multiplied. But I think it will be revealing to see the differences, which I am cross fading from and to in video.

Here's another preview, I'll leave you to guess what's what.

U3EvqYL.jpg


This will be out well before I jump on a plane for NAB so there's not too much longer to wait!
 
Excited to see the results as well, thanks Brett!

When I did my Scarlet-W vs. Panasonic VariCam, the hardest part was indeed trying to determine which would be the best way to prepare the footage and display them for a fair comparison.
 
Hello dear patient Red Users on this thread!

I am in Las Vegas and finally finishing up this test in my hotel room rather than going to NAB, because I was booked on an 10 day away TVC that ended 2 days after I was originally going to fly! Not a bad problem to have.

Here's a little preview as I put the finishing touches on the post (which will be its own thread).

https://vimeo.com/163315107
 
Another update dear constant Red readers.

I managed to meet with RED and ARRI during NAB - each kindly took a look at the test. I felt I should check in with the manufacturers as it's a political subject potentially and I figure any comments/criticism before the fact would only improve how I eventually release the results. I also am aware of some negative attention from previous comparisons so I wanted to curtail that potential as much as possible.

This all means that it will be another week or so before the test is released! Sorry to those who have prompted me. The upside is that the results will be more comprehensive and also that feedback from various parties will be channeled into a more comprehensive test in the near future!

Cheers,
Brett.
 
Hello

I used to provide development feedback to SGO and UniColour (the Mistika Colour Space/Gamma transform effect) and ACES support were 2 of the last things that I provided information and feedback for, as well as tested on real-life jobs when grading using a Mistika (although I use plenty of other grading platforms as well as Mistika).

I hate looking like an idiot so I tested the hell out off both the debayer effects as well as the UniColour effect. Most of the information was available within the camera vendor's SDKs. I got the ACES debayer option added to the Mistika R3Dparameters effect (the initial debayer effect for .r3d files) and then contacted and confirmed that the tranform was correct ACES with Graham, as well as comparing the Mistika debayer to ACES result against the same thing from REDCineX, and then Fusion (just to throw a 3rd party in there as well to triple-check).

My development feedback ended over a year ago (anyone who was at NAB and IBC can confirm that) but I still use Mistikas quite a lot and I've never had a problem matching the results from other software. I also grade on Baselight and if you set debayer to 'AUTO' in Truelight Colour Spaces you'll find that Baselight will debayer RED directly to ACES (what it considers the RED debayer option to contain the maximum amount of information) and then applies the transform to any other colour space desired from there. The reason being is that digital cameras are all linear light under the hood - they don't 'capture' Log, that's applied later. So an ACES debayer means 'what the camera captured' in terms of Gamma/Contrast (no change in other words) and the native RED camera gamut fits entirely within the ACES gamut (so it's a none destructive gamut mapping. Just like Rec709 -> DCIP3 is non-destructive due to the fact that all of the colours possible in Rec709 or possible in DCIP3).

So the ACES transform from RED is 'correct' (at least it was last time I ran tests). The other transforms that are both non-destructive and well-documented are the Sony RAW debayer and transform flavours (SLog3/SGamut3) and Arri Wide Gamut LogC. Also the Canon CLog/CLog2 & Canon Cinema Gamut but they're not the subject here. Either way the reversible Curves/Gamuts are LogC/SLog3/CLog/Cineon (Cineon non-destructive and reversible without losing information only in Mistika I believe. 'True' Cineon, PQ and DCIP3 curves/gammas all contain a clip if you check out the oh-so-exciting White Papers on the subject).

I graded a UHD job last year that contained RED Epic (non-Dragon sensor) .r3d, Alexa .ari, and Sony F55 RAW .mxf. I used the native gamut of each (chroma is easy to match, it's the gamma/contrast that's the pain in the arse) but used UniColour to move all 3 into a consistent gamma/contrast. It doesn't matter whether I graded on Linear-Light/SLog3/LogC/Cineon-REDLogFilm, the point is that I could have chosen any of them and they would have matched. The same for matching the gamuts. The only information that is in contention is how to get 'the most' out of each RAW file. I didn't care about mapping the gamuts - I just graded them to match, which was easy as it was almost entirely down to a Gain chroma manipulation and Saturation on top of that. If I had wanted to be pedantic/accurate then I could have sent them all to the ACES gamut as well then mapped them from there (I wouldn't grade using the ACES primaries - personal experience has proven that it's extra stress for no gain and RGB grading tools hate it, particularly Hue tools). The other option would be to debayer the Sony to SGamut3, the Arri to Arri Wide Gamut and the RED to ACES, then map the ACES Arri & RED to SGamut3. That is totally pedantic 'for the sake of this arguement' logic but there are some areas where the Wide Gamut and SGamut3 primaries go outside of the ACES primaries. These colours are outside of the human visual range, and well outside of Rec2020's limits but, if the point of this thread is to be 'totally accurate' then that's the way that I would do it. You can then map the Sony and Arri to ACES using UniColour. It contains a Bradford matrix so that would result in the most accurate gamut mapping.

Almost all of the above is mostly academic. Just debayer RED to ACES Linear, Arri to Wide Gamut LogC (it's 12 bit so the RAW does't contain the Linear Light - the LogC is essentially working like an 'Image Zip file'), and Sony RAW to SGamut3 Linear, and then use UniColour to take them all to a unified colour space of your choosing.

Don't use the Arri 'Rec709' or the Sony SGamut3.cine. Both contain a 'Beauty Transform' for lack of a better term. They're not true transforms in terms of 'What the camera saw'. Much like transforms like DragonColour2/REDGamma4 they are transforms that the manufacturer created whose purpose is to 'Make The Camera Look Good'. Not to meet SMPTE standards, etc.

Clear as mud? Good. I got half-way through that and even I started going colour-mad :drool:
 
You must log in or register to reply here.
Back
Top