Just to share what I saw, below are three stills from Ursa Mini 4.6K footage captured to HD ProRes444 12bit.
It's an awful shot that was never meant for anything, but I chose it because it includes a lot of blue items that illustrate well the issue I finally noticed after JB pointed it out.
In the matching grade, beyond setting contrast and saturation, I adjusted the hue v hue and hue v sat curves to get as close as possible to the look of the Color Space Transdform + Arri REC709 LUT combo. The real noticable difference is the luminance of blue. The only way I could see to match that exactly would have been to adjust the hue v lum curve, but that added a lot of noise to the image so I didn't do it. Interestingly, the Color Space Transdform + Arri REC709 LUT combo also adds noise to the image. So, it seems to me that the Color Space Transdform + Arri REC709 LUT combo is doing things to the image that I'd otherwise not like and would not choose to do in a manual grade.
Is this the problem with the Color Space Transdform + Arri REC709 LUT combo that you are describing JB?
If you'd like to see it motion so that the noise is apparent, here's a link to download the original file and graded clips, plus .drx files of the grades to load into Resolve to test on other Ursa Mini 4.6K Film v3 files.
Ok, i finally have some time to look at the shot. nice raining day both outdoor here in LA and in the image.
no evident problems in the raw. a bit on the contrast side, no clip of sort, no breaking of color, on the saturated side: I'm not used to the tonal distribution of the Ursa, so I assume that it is fine.
First test, do the transformation from [blackmagic design film v3/film log] to [709 gamma 2.4].
the result is pleasing, I'm noticing a clumps of color in the cyan side (straight line) and it is the cyan in the window. very high and saturated. It looks like that particular shade of cyan is hitting the boundaries of the color space. I open the log image (offset) by 90, looking at the waveform: indeed the math create two lobs in the R and B channel that go under [0] line. nothing unexpected for pure math.
I add another color transformation to see if it is just a 709 limit or the data is really clipped somehow: [709 gamma 2.4]to [blackmagic design film v3/film log] and i get the original image with no clip/crush/loss. The math works (as expected).
BM LUT; signal log straight to BM LUT: this will map the tonal response as for the recommended BM workflow. No clipping, nice tonal curve, good colors, nice toe and shoulder, no noise in the image.
This, is the BM color science at it's minimum:
sensor capture light
[AD] --> raw container (or log container, whichever is used in this case)
apply the recommended tonal curve from BM.
Nothing strange, this is how BM decided that the camera should look and respond to light plus how the final image should be mapped to the display device, both mathematically and artistically (the toe/shoulder part.)
Now, i will transform the BM to the alexa using the color transform and to 709 as for the other workflow:
[blackmagic design film v3/film log] to [alexa wide/log]+ another node+[alexa wide/log] to [709 gamma 2.4].
This will look identical to the [blackmagic design film v3/film log] to [709 gamma 2.4] done earlier.
In fact if i add the [709 gamma 2.4]to [blackmagic design film v3/film log] from there I get the original image.
Then i will do what the original idea was:
[blackmagic design film v3/film log] to [alexa wide/log]+ color corrector+alexa lut (the standard provided by arri): the image looks nice and smooth, no issue. The workflow works. I will call this [alexa lut] for short
Now, does the [alexa] lut match the [BM lut]?
No. and it should not be a surprise.If you decide to use the [ALEXA LUT] workflow, what you're trying to say is: if i have an alexa camera and i point it to the image, using the standard ALEXA color science (
and how they decide to map it to the final device) it will look like that.
If someone has an alexa and a ursa, you can try to put it side by side and do the double check, when both are used at the optima setting, similar lenses, you can do either:
Ursa->color->UrsaLUT
and you will use the BM device representation, in this case you will want
ALEXA->color trasform to URSA log-> color-Ursa LUT
will make the ALEA behave like an ursa.
The opposite if you want to mimic the alexa workflow:
ALEXA->color->alexa lut
Ursa->color trasform to ALEXA log-> color-alexa lut.
and both will behave as for the Alexa color science.
w.