New RED Color Science - Test Pilot Feedback Thread

[Roberto Lequeux]

Will the 5500k rating of the sensor be re-evaluated or is it still a correct value? From what I've gathered in this thread so far it seems that though there is an immense leap forward in the blue channel Mysterium still likes blue light better, right? Is the 5500k the sensor's rating or a rating for the whole system?

 

[David Mullen ASC]

The natural sensitivity for all silicon sensors is biased for 5000K light or so, that's not going to change. The only issue is whether you can get the noise level low enough so that 3200K balance looks acceptable, noise-wise.

 

[Roberto Lequeux]

I see. So perhaps the 5500k rating might be dropped a bit closer to closer to 5000k now.

Thanks for the answer.

 

[Sanjin Jukic]

Eau de Toilette bottles at tungsten 75W

Eau de Toilette bottles at tungsten 75W

Eau de Toilette bottles at tungsten light 75W

Shot on RED1 with Canon 50mm FDL f/1.2 @ f/2, ISO320, pick WB in RA 3.72B = 2542K, Tungsten light bulb 75W, RED Space out

4K JPEG>>>

Shot on RED1 with Canon 50mm FDL f/1.2 @ f/2, ISO320, pick WB in RA 3.72B = 2456K, Tungsten light bulb 75W, RED Space out

4K JPEG>>>

Shot on RED1 with Canon 50mm FDL f/1.2 @ f/2, ISO320, pick WB in RA 3.72B = 2416K, Tungsten light bulb 75W, RED Space out

4K JPEG>>>

Shot on RED1 with Canon 50mm FDL f/1.2 @ f/2, ISO320, pick WB in RA 3.72B = 2392K, Tungsten light bulb 75W, RED Space out

4K JPEG>>>

 

[Paul Leeming]

Another set that needs no comments...

Peter, I notice you have different colour spaces set between 3.60 and 3.72B RedAlerts. Is this Red's request for testing? Strangely though, I would expect RedSpace to be the one with brighter, oversaturated tones, yet it clearly gives a more natural,tonally correct image than Rec709!

Thanks for all the examples, I can't wait to have this propagated through the whole SDK and app chain (RedCine, Adobe plugin etc).

Cheers,

Paul

 

[Irri]

Paul,

There was no request for specific colour spaces as far as I know. I used to use redspace as the colour space when using REC709 gamma as Peter is doing in that example. However, with the new Red Alert the REC709 colours are much nicer. None of the strange colours that you see in the old version of Red Alert, such as those posted above.

 

[Peter Majtan]

Paul, I did not notice it myself. I was trying to do minimum changes in RA. I simply opened the two versions of RA, opened the same file in both, selected the same exact spot for white-balance, applied the white-balance and exported the TIFF's. I will re-run the same thing and make sure to change the color space to be the same...

peter

 

[Peter Majtan]

Paul - I corrected the color space and here's the result:

So very little difference (as compared to the mismatched color-spaces)

Ockare Sama Desu!!!

Peter

 

[Peter Majtan]

I was also intrigued what is going on, so I set the 360 RA to the same exact setting as the 372 and things got much worse (for the 360):

This is some serious woodoo...

Graeme - what was the exact brand of whiskey You drink??? :devil:

Peter

 

[Carl Dieker]

Great days to be a RED shooter. I don't know if this is the place to comment on this, but I wonder how stable the new 18 beta build feels like.

/Calle

 

[Birns and Sawyer]

Looks like REDspace migt be getting a tad less of a curve applied on it based on Peter's revised RA screen grabs.

If so, that might make REDspace a bit more of an option, as I currently think that REC709, R 1.0, G 0.9, B 0.8 and Saturation 0.8 gives a much better look than Redspace.

Can't wait to get ahold of this.

 

[Gregory Leno]

So does the RED Color Science positively affect keying with tungsten light sources?

Has anyone tested this out?

Greg

 

[Paul Leeming]

Paul - I corrected the color space and here's the result:

So very little difference (as compared to the mismatched color-spaces)

Ockare Sama Desu!!!

Peter

Otsukaresamadeshita!!! :beer:

Thanks for the update Peter, identical settings actually show even greater improvements!!

The difference is frankly amazing! Unbelievable that this is all down to the algorithms used to process the RAW data....

....my hat's off to you Graeme, you are the MASTER!! :thumbsup:

I shot some tungsten heavy test footage today for NHK and would love to try out this new build (or the RedAlert portion I guess) on it. Hope it's made a release package for the rest of us soon (especially in RedCine so I can use it on my PC also).

Cheers from Tokyo,

Paul

 

[Irri]

Has anyone else compared build 18 RA 3.72b with build 17 RA 3.6 yet?

Seeing as this is the difference everyone thinks they are seeing here, it's odd that no one seems to have compared them.

 

[Jonas Rejman]

Sweet mother of what is good and pure!

 

[Laco Zamba]

Thats great. Thanks, Graeme.

But I cannot resist: why it was not so good from the begining?

OK, just joking.

 

[Graeme Nattress]

Laco - we have all learned an awful lot in the past couple of years.

Graeme

 

[Mark L. Pederson]

Does this affect the H proxy workflow, or only if you render things out?

There will need to be an updated QT component - right now only on render.

 

[Laco Zamba]

Laco - we have all learned an awful lot in the past couple of years.

Graeme

Me too, thanks to you and RED.

 

[Patrick Tresch]

The natural sensitivity for all silicon sensors is biased for 5000K light or so, that's not going to change. The only issue is whether you can get the noise level low enough so that 3200K balance looks acceptable, noise-wise.

I'm working with a Nikon D2x (old of 4 years) and the color science used by Nikon make it possible to shoot at 2900k or 5500k without "great" differences in the blue channel. I was surprised as I come from a RED "CMO background"...

BTW, what RED achieved with this new debayer algorithm is wonderful.

Thanks.

Patrick