next thread »

[Anders Holck]

Just a heads up that Davinci Resolve has a nasty bug interpreting ARRIRAW when the built in ND filter is engaged. I have not seen it described in detail on the forum yet, so just wanted to inform you about it.
The issue is what looks like an incorrect sensor profile used: basically setting the raw whitebalance slider above 3000 kelvin is gradually destroying green frequency response. Don't know if resolve uses the dual illuminant technique of Adobe, but it does seem to be correct at around 2900 kelvin and becomes fully corrupt at 6500 kelvin. So one theory would be that the D65 illuminant profile is wrong. Why it occurs only with the ND engaged, I don't know. But it's a total change in color processing, not just a slight color shift due to the filter.

Here is a Macbeth test chart that I mailed to Blackmagic support, but you can grab it if you want to see how the error looks. First take is with the filter engaged, the second is without. Reference Prores444 QTs are also included.

http://clients.redrental.dk/files/anders/arriraw_alexastudio.zip

You all might want to advise your rental clients about the issue if they are shooting ARRIRAW on the Alexa Studio and are planning to conform in Resolve.

[cameratechdre]

I wonder if you go through ALC, and activate the ND correction.... awww but you can't render out to .ari (just double checked). Would be nice if you could render out to .ari files from the ALC.

[Michael Borenstein]

Anders,

Thanks for the heads up.  I will forward this up the chain to see if we can put some pressure on the resolve people to get this fixed.

Regards
Mike

[joseph]

Hello Anders--

Could I ask you to please replicate your experiment, using grabs to the SD card in .ari format, and post the results?

Many thanks.
—joseph

[Anders Holck]

I will do so tomorrow and post the results. I got the same results on some codex onboard material I shot a few months back, where we ended up using ARC. The ones above was shot using the Gemini 4:4:4

[Anders Holck]

Here are two .ARI files grabbed from the camera:
http://clients.redrental.dk/files/anders/arriraw_grab.zip

It does show the same fault inside Davinci

ND On:


ND Off:

[joseph]

Well, that's certainly consistent with what you've been saying. Could you send me the raw files? jgoldstone (that first letter is a lower-case J) blah arri.com, replace the blah with an '@' and you're there.

On my side, last Friday I did two grabs on a Studio next door. Here are some ways of processing them. The ones that are slightly out of focus are the shots with the ND (room light + a very small TrueBlue meant that I was very open).

Resolve 8.2.2 (no ND in path):




Resolve 8.2.2 (ND in path, aperture opened way up):



ARC 2.3.0b20 (no ND in path):



ARC 2.3.0b20 (ND in path, aperture opened way up):



So the residual difference in Resolve 8 seems comparable to that of the ARC.

We just installed Resolve 9 next door and I will try and experiment with it tomorrow. If you can email me the .ari files corresponding to the images you are showing above, I will try them out here, compare your metadata with my metadata, etc. And I will reply to your email enclosing the .ari files that I got from my grab and you can check them out. (I'll send you the .ari attachments in two separate messages to try and stay below your mail system's attachment limit.)

Many thanks--
--joseph

[Anders Holck]

Weird....
I have send you the link in a mail.
Our Studio is out this week, so my access is limited.
Thanks for looking into this

Best
Anders

[joseph]

If you take our frame with the ND in it, and override the CCT (in ARRI 'codec settings') to be anything up to and including 5100 it gradually changes color. If you go past 5100 to 5600, then wham, it gets weird.

I will pass this on (and all four images via DropBox) to the Resolve color science guy, who is probably up to his eyebrows in IBC prep right now, but whatever.

Note that I never got around to seeing if I could reproduce the same CCT 'cliff' in Resolve 9.

Thanks for sending those images, they really helped narrow this down.

Best--
--joseph

[joseph]

Hello Anders--

I received email back from the Resolve team. Using the images supplied, they reproduced the bug, found its cause, and fixed it. The fix will be in the final, non-beta release of Resolve 9.0 (which he says will be "very soon").

They don't anticipate back-porting this bugfix into 8.x, so if getting the right color with a Studio ND in place is required, and you must use Resolve 8.x, I'd suggest using something like the ARRIRAWConverter in its GUI [Mac only] or command-line [lots and lots of platforms] form.

Best--
--joseph

[Anders Holck]

Hi Joseph.
Great thanks.
I'm very glad this will get fixed in the 9.0 release build. So thank you very much for investigating and getting the Resolve team on the issue and fixing it. Thanks!

PS. I also have a resolve 9 bug to file regarding Phantom Cine file RAW conversion, can you get it through to them if a mail a report to you?

[joseph]

Hello Anders--

Sorry for the delay, I don't normally read the forums unless Michael brings something specifically to my attention.

No, my credibility with the Resolve team is pretty much limited to the ALEXA (or anything to do with ACES). I've never even seen a Phantom, much less know anything about its workflow. So many products, so little time...

Best regards--
--joseph

ps: two words for IBC attendees to look up on Google: Puccini Bomboni