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 :)

RED Alert White Balance off on build 20 footage

Andrae Palmer

Andrae Palmer

Well-known member
Joined
Sep 1, 2007
Messages
2,480
Reaction score
1
Points
38
Age
52
Location
Fort Lauderdale, FL
Website
www.facetmedia.com
Still frames from footage recorded with Build 20. Made sure RSX file was deleted when switching in between RED Alert and REDCine. Kelvin was set at 5600, REDspace for both. I also tested on many other clips that were shot with Build 20 under 5600 outside in sunlight.... still the same issue. The interior was lit with sunlight for this still frame.

REDCine v3.18:
3750335593_f1571f1dc6_b.jpg


RED Alert v20.1.6:
3751173402_2f74b454d9_b.jpg


I thought the white balance issue should be the other way around. How comes RED Alert is so way off?
 
Can you tell me what the actual color temperature of the light that was illuminating your subject was?

Graeme
 
That's exactly what I would expect to see, given that Red Alert is using the new colour science, and RedCine is not. You should NOT expect them to give the same result for the same parameters.
 
Daylight here is usually measured around 4800k. I did a measure down at the RED offices in Lake Forest last week and it was 4900k. If you're setting the wb to 5600k when the light is not 5600k, then the warm appearance is totally correct and as expected. That is why I need to know what the measured color temperature of the light was.

Graeme
 
Graeme Nattress said:
Daylight here is usually measured around 4800k. I did a measure down at the RED offices in Lake Forest last week and it was 4900k. If you're setting the wb to 5600k when the light is not 5600k, then the warm appearance is totally correct and as expected. That is why I need to know what the measured color temperature of the light was.

Graeme
Click to expand...

Great point Nattress... perhaps sunlight in Florida is more towards 4800k. I didn't have a Color Meter with me and didn't use the auto setting on camera. I always used the daylight setting before without any issues. Still it seems strange to me that REDCine got it so right and RED Alert is way off. I now have to White Balance Build 20 footage in RED Alert to get all the clips correct. I have many more clips that were shot under very bright Florida sunlight and it's still the same issue.

Nick Shaw said:
That's exactly what I would expect to see, given that Red Alert is using the new colour science, and RedCine is not. You should NOT expect them to give the same result for the same parameters.
Click to expand...

From reading the Build 20 notes you would be lead to believe that REDcine would be the one with the incorrect white balance. It states that you will have to white balance Build 20 footage in REDCine. Currently I have to white balance clips in RED Alert... the other way around of what I would have expected.
 
The original kelvin calibration before build 20 was a bit rough. It was consistent, but rough. Now it's consistent, but accurate (usually within a 100k or so of the measured value, which is pretty darn good for a camera).

Remember, the data is raw, so it's entirely conceivable that the old calibration thought white was around 5600k, and the new around 4800k, and hence the visual difference in the apps. But what did it look like in camera where you'd set it to 5600k? Did it look warm their too?

If you can shoot a kodak grey card, and do an auto-wb off that, that'd be great. If you have a real macbeth chart (not one of the copies as their colors are way off) and can shoot that under your lighting, I can always take a quick look at the R3D here and see what is going on. If you want to do that, shoot the smallest possible R3D for me and give me a pm and I'll tell you where to send it. If you know how to set up interval record, that can shoot a 1 frame R3D that's pretty small to email.

Hope that helps,

Graeme
 
Graeme Nattress said:
The original kelvin calibration before build 20 was a bit rough. It was consistent, but rough. Now it's consistent, but accurate (usually within a 100k or so of the measured value, which is pretty darn good for a camera).

Remember, the data is raw, so it's entirely conceivable that the old calibration thought white was around 5600k, and the new around 4800k, and hence the visual difference in the apps. But what did it look like in camera where you'd set it to 5600k? Did it look warm their too?

If you can shoot a kodak grey card, and do an auto-wb off that, that'd be great. If you have a real macbeth chart (not one of the copies as their colors are way off) and can shoot that under your lighting, I can always take a quick look at the R3D here and see what is going on. If you want to do that, shoot the smallest possible R3D for me and give me a pm and I'll tell you where to send it. If you know how to set up interval record, that can shoot a 1 frame R3D that's pretty small to email.

Hope that helps,

Graeme
Click to expand...

Ok I'm heading outside right now and will white balance off an 18% Gray card. I will be right back.
 
Thanks! This kind of data is very useful for me as we go forwards!

Graeme
 
It certainly felt a 100 degrees warmer down there.... :-)

Graeme
 
Graeme Nattress said:
Thanks! This kind of data is very useful for me as we go forwards!

Graeme
Click to expand...


You are correct Graeme there is a big difference in the color temperature of daylight. I thought 5600k would have been pretty close... I'm learning that I will have to white balance each and every time. I used the 18% gray card and got a reading of 4578 Kelvin in bright sunlight with no cloud cover. With cloud cover its around the 5600k ball park.

RED Alert v20.1.6 color temperature at 5600k:
3752433244_6d9c130264_b.jpg


RED Alert v20.1.6 color temperature at 4578k:
3751646783_5047c08683_b.jpg


The funny thing is that I don't remember this type of cast on my older 5600k clips shot with non Build 20. More accuracy in Build 20?
 
Much more accuracy in B20 - that was one of the goals of the project.

Graeme
 
Graeme Nattress said:
Much more accuracy in B20 - that was one of the goals of the project.

Graeme
Click to expand...

Thanks for your help sir... I was surprised that the color temperature of sunlight varies so much by over a 1000 kelvins depending on the conditions outside.
 
Daylight color temp varies wildly and is hardly ever, so it would seem, the D65 it should be. It seems to be commonly in the 4800-5000k range when I've tested it, but I'm told can get very much higher under a storm cloud, but I've not been outside to test that just yet.

Graeme
 
If the real color temp of the light was around 4800k, but metadata was recorded at 5600k, then the new CS will display the image as warm. The old CS will display it as neutral because the old CS calibration thought that the real world 4800k was much nearer to 5600k.

Graeme
 
Graeme Nattress said:
If the real color temp of the light was around 4800k, but metadata was recorded at 5600k, then the new CS will display the image as warm. The old CS will display it as neutral because the old CS calibration thought that the real world 4800k was much nearer to 5600k.

Graeme
Click to expand...

thanks. gotcha.
 
Graeme Nattress said:
If the real color temp of the light was around 4800k, but metadata was recorded at 5600k, then the new CS will display the image as warm. The old CS will display it as neutral because the old CS calibration thought that the real world 4800k was much nearer to 5600k.

Graeme
Click to expand...

This makes a lot of sense... but why was the old CS calibration so way off?
 
Why was it off? Because it's darn hard to get it accurate :-) Take a look at Aperture and Lightroom (just because they're the stills raw converters I'm familiar with) and shoot something raw on, say, your Canon DLSR (again, just because I'm familiar with it), and measure the color temp with your temp meter. Shoot with RED One B20 and compare...

Graeme
 
Graeme Nattress said:
Why was it off? Because it's darn hard to get it accurate :-) Take a look at Aperture and Lightroom (just because they're the stills raw converters I'm familiar with) and shoot something raw on, say, your Canon DLSR (again, just because I'm familiar with it), and measure the color temp with your temp meter. Shoot with RED One B20 and compare...

Graeme
Click to expand...

I wish I knew more about the technicalities of Color Science to be aware of the issues faced.
 
You must log in or register to reply here.
Back
Top