Web Analytics

Colour Reproduction of Large Format Digital Cinema Cameras - METExpo


Pawel Achtel, ACS
 

Just a quick note and invitation to anyone interested.

 

I will be presenting an ACS branded workshop at the METExpo (former SMPTE) at ICC Sydney on Friday, 19 July 11:00 am.

This one hour session will cover some general colour science background information and will be followed by a presentation of test results of the following four cameras:

 

·       Arri Alexa LF

·       Sony Venice

·       RED Monstro VV

·       Panavision DXL2

 

It is not going to be a competition of which camera performed best or worst (although you will be able to make your own judgements), but more of an insight of what colour response and accuracy we can expect from a modern digital cinema camera and how to get most out of it. Lots of interesting/unexpected observations.

 

Registrations: www.metexpo.com.au

 

I look forward to see you there.

 

Kind Regards,

 

Pawel Achtel ACS B.Eng.(Hons) M.Sc.

producer and director

 

ACHTEL PTY LIMITED, ABN 52 134 895 417

Website: www.achtel.com

Mobile: 040 747 2747 (overseas: +61 4 0747 2747)

Mail: PO BOX 557, Rockdale, NSW 2216, Australia

Address: RA 913 Coles Bay Rd., Coles Bay, TAS 7215, Australia

Location: S 42° 0'14.40"S, E 148°14'47.13"

Email: Pawel.Achtel@...

Facebook: facebook.com/PawelAchtel

Twitter: twitter.com/PawelAchtel

Skype: Pawel.Achtel

 


Pawel Achtel, ACS
 

 

Justin Johnson wrote:

 

This following movie shows how super saturated colors fall outside of various color spaces when debayered at different WB settings. 

https://www.dropbox.com/s/f08latv4u3cp78x/CameraPlotting.mp4?dl=0

 

Justin, that’s very interesting and consistent with some of my test results.

 

My methodology was very similar too: I used fibre optic illuminator and a slit to achieve collimated light source. Monochromator/spectrograph (ruled diffraction grating) was used to produce spectrum of monochromatic light that was directly projected onto the sensor’s surface.

 

However, I used Rec 2020 as the output transform (using each vendor’s native software), so I didn’t notice colours going outside the space. Interesting observation, but maybe a bug in the debayering software you used, not sure. It would be interesting to know your workflow.  

 

In theory, monochromatic locus should be invariant to colour temperature. This, clearly doesn’t seem to be the case, as you have demonstrated in your tests. That’s one of the common findings.

 

Another common observation is that there are (very) significant colour shifts in all of the RED cameras that I tested, including DXL2.

 

I would be interested to hear more about your workflow. You are welcome to contact me directly or on CML and see if we can get to the bottom of it.

 

Kind Regards,

 

Pawel Achtel ACS B.Eng.(Hons) M.Sc.

“Sharp to the Edge”

 

ACHTEL PTY LIMITED, ABN 52 134 895 417

Website: www.achtel.com

Mobile: 040 747 2747 (overseas: +61 4 0747 2747)

Mail: PO BOX 557, Rockdale, NSW 2216, Australia

Address: RA 913 Coles Bay Rd., Coles Bay, TAS 7215, Australia

Location: S 42° 0'14.40"S, E 148°14'47.13"

Email: Pawel.Achtel@...

Facebook: facebook.com/PawelAchtel

Twitter: twitter.com/PawelAchtel

Skype: Pawel.Achtel

 

 


Pawel Achtel, ACS
 

Oliver Schietinger wrote:

 

“I was hoping you could still share with me what you meant by color shifting in the Red Monstro and DXL2 cameras. (I assume you used IPP2 (Red Wide GamutRGB/ Log3G10 mode and corresponding post settings. )”

Hi Oliver, no problem.

 

Yes, you assumed correctly: I used IPP2 (RED Wide Gamut RGB/ Log3G10 gamma) pipeline in RedCine Pro with output transform set to Rec 2020.

Our source material involved projection of monochromatic light onto the sensor’s surface. As we know, monochromatic colours, by definition, are single wavelength light and cannot be recreated by mixing other colours. They are also invariant to colour temperature as they fall on the outer curved boundary is the spectral locus. These last two statements don’t hold true with RED Monstro. Below is the result:

 

 

As you can see red monochromatic colours are reproduced as magenta (which is NOT a monochromatic colour). The vector of colour shift is quite significant compared to other cameras that I have tested.

 

While there is always scope for error, I have performed this test several times and the settings in Red Cine Pro were independently reviewed by at least three experienced cinematographers. In previous post Justin Johnson’s test was entirely independent and it shows consistent results too.

 

Kind Regards,

 

Pawel Achtel ACS B.Eng.(Hons) M.Sc.

“Sharp to the Edge”

 

ACHTEL PTY LIMITED, ABN 52 134 895 417

Website: www.achtel.com

Mobile: 040 747 2747 (overseas: +61 4 0747 2747)

Mail: PO BOX 557, Rockdale, NSW 2216, Australia

Address: RA 913 Coles Bay Rd., Coles Bay, TAS 7215, Australia

Location: S 42° 0'14.40"S, E 148°14'47.13"

Email: Pawel.Achtel@...

Facebook: facebook.com/PawelAchtel

Twitter: twitter.com/PawelAchtel

Skype: Pawel.Achtel

 


Justin Johnson <johnson322@...>
 

 

 

 

 

However, I used Rec 2020 as the output transform (using each vendor’s native software), so I didn’t notice colours going outside the space. Interesting observation, but maybe a bug in the debayering software you used, not sure. It would be interesting to know your workflow.  

 
Both ARRI and RED crop (clamp) any colors that go outside of the chosen output colorspace during the debayer, except for ARC's conversion to ACES which doesn't crop any values and allows them to go negative into the float EXR container.  (Hence the artifacts seen in ARRI-ACES pipelines (purple fringing)).  Both RED and ARRI are currently working on methods for reigning in out-of-gamut colors, I assume with gamut mapping algorithms. 
 

 

 

In theory, monochromatic locus should be invariant to colour temperature. This, clearly doesn’t seem to be the case, as you have demonstrated in your tests. That’s one of the common findings.

 
ARRI has published their image processing pipeline (RDD 31:2014) where they describe the white balance specific multiplication and matrices used.  (These values are also included in the metadata of the ARRIRAW file) These push the position of the supersaturated values when WB is changed.  RED is obviously doing something similar but they don't publish the specifics.   I am not sure what other camera makers do. 
 
 

 

 

I would be interested to hear more about your workflow. You are welcome to contact me directly or on CML and see if we can get to the bottom of it.

 
I am using Nuke to generate these images by converting each pixel of an image into a point cloud in Yxy colorspace.   

 

Please sign your messages with name job and location