Welcome to our community

Be a part of something great, join today!

Register 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 :)

The blue channel under 3200K vrs 5600K lighting...

J

Jim Arthurs

Well-known member
Joined
Dec 28, 2006
Messages
555
Reaction score
0
Points
0
Website
www.imageshoppe.com
Hi all... remember this question I had?

http://www.reduser.net/forum/showpost.php?p=117182&postcount=185

Well, thanks to Jon Firestone and Chris Bagley and RED #232 we had a chance to test the whole question about noise in the blue channel as it relates to the color temperature of the lighting. Since the sensor is daylight native it makes sense that it would produce the cleanest image in daylight, and that this might be an advantage when doing blue or green screen work.

We shot both blue and green screen footage, changing the lighting from tungsten to daylight. Obviously the same effect can be had by filtering at the lens, but we didn't have access to the proper filter in 4 by 5 so we did it with gelling the lights and ND...

We've just made available raw R3D clips of this test, along with some dpx's here...

http://www.reduser.net/forum/showthread.php?t=7451

But to cut to the chase, here's a 1:1 sized sample from both a blue and green screen, shot under both lighting conditions. It's very easy to see that if you can do it, light your key stages with daylight sources. Not only for blue screen work, but as you can see, the blue channel in the green screen plate benefits greatly.

http://ftp.datausa.com/imageshoppe/outgoing/RED_232/blue_channel_compare.png
 
This is the same thing with every video camera. I figured this out the hard way after lighting a greenscreen and a scene with all tungsten fixtures for an HVX shoot. The closer to daylight the better with a video sensor.
 
Slinnn said:
Wow, the difference is huge !
Did you expose both at 320 ISO ?

Thanks for doing this test. Maybe we'll have to keep using blue filters after all...

Nils.
Click to expand...

Yes, 35mm zeiss superspeed, f 2.8 and 320 rated on all....

Understand this is more important for keying than "ordinary" footage... I don't mind the natural look under tungsten in the least and wouldn't worry about it in the slightest in non-key situations. But knowledge is good, and if you know a thing, you understand a thing...

And while this might become one of those "oh yeah, of course" points that every RED Ninjia will tell a client as rote in a bit, until this test by myself and Jon Firestone it was unknown in fact, and if people read back to just a month ago the party line from folks with deep RAW experience in general was that there was little value to be had by properly color filtering to match the sensor.

Just sayin'.
 
Good to know

Good to know

Great post, thanks a lot the pictures do tell the real story
 
Wow thats such a huge difference it makes me think something might be broken with their white balance transform.


Were these processed in RedCine and RedAlert to compare?


You gelled the tungsten lights to go daylight right? How did you manage to keep the stop the same? did you move the lights in?
 
In this case, the screen lighting was the desired color temp + either 1/2 CTB for the bluescreen, or 1/2 Plus Green filtration for the green. Distance was such that there was no contamination on the subject lighting. Green (or blue) kino's would have been nice, but weren't in the budget.

RedCine on the PC only for processing, latest version.

We did white balance the camera to a white card and that's the meta data that was in the files (3003K and 5339K to be exact)... however, when I manually moved it back to either 3200 or 5600 there wasn't a noticable difference in the grain amount. Feel free to download the raw files and have a go... the dpx's will be back on line by evening (client upload right now)...
 
much more blue light in 5600K

much more blue light in 5600K

Slinnn said:
Wow, the difference is huge !
Did you expose both at 320 ISO ?
Click to expand...
I gather he's putting up raw R3D files, so you can check that directly.

As I understand it, there is roughly 5 times (that is 2.3 stops) more blue in 5600 K light as compared with 3200 K light, given the same overall exposure (that is, "L" is the same, in Lab color space).

With 2.3 stops less blue light, I would expect to see a noise difference in the blue channel. I expect this would hold true for any camera, by the way, unless you are far enough above the noise floor not to matter.

I used the color calculator at http://www.brucelindbloom.com/index.html?Equations.html
and assumed sRGB has a 2.2 gamma, to work back to get linear light.
 
It's been long known that for optimal keying use daylight to light. Using tungsten you're adding red to the green. You can see this by simply hooking up any camera to a vectorscope.

Changing WB to tungsten means effectively you're upping the gain in the blue channel. More noise will always result. Some cameras get aroung this by having an optical filter in front of the sensor. Does this really solve the issue, I'm not 100% certain. If you have plenty of light probably not but under low light you're then reducing the amount of light in the red and green channels so overall the gain has to go up.
Regardless of all this being well known before Red was even thought of it still pays to test.
 
Jim,

Thanks for posting the results. I do green screen fairly often with tungsten lights, so it's good to know.

BTW, did you get a chance to test the rolling shutter.
 
Yes, there must be a reason why (Photoshop) Camera RAW's Chroma NR defaults to 25%...

I find that enabling the Matrix in Red Alert adds a lot of artefacts when the tempreture goes close to 3200.

When comparing Red Alert to the other RAW converters I use (Mostly Lightroom/Camera RAW) I also find that the tempreture and tint slider seems to act weird when operating on a 3200 image compared to a 5600 image.

Could a tweak to the matrix lower the noise/artifacts here?
 
Anders Holck said:
Could a tweak to the matrix lower the noise/artifacts here?
Click to expand...

Hi Anders! I'm a total newbie to the world of RAW and have only started to learn REDCINE...

We're thrilled to be able to offer the raw files for any experimentation you'd like to do... my DPX files will be on-line later this evening, but even these are approximations compared to the RAW's...
 
None of this surprises me one bit, I was talking about this quite a few months ago in the footage forum.

This seems to apply to any RAW digital camera as far as I know. I figured it out by messing around with my Andromeda camera. I also started asking around about magenta filtering, as I found it further decreased my noise (its complex to say the least, but basically I strap on about 60 ccM and then manually white balance under daylight sources, even helps when shooting under tungsten). I'm told magenta filtering won't help the Red One, my personal theory is that the camera has a sleight amount of magenta filtering built in... Just look at any picture of the cameras sensor, its pretty darn magenta looking!

http://www.reduser.net/forum/attachment.php?attachmentid=2459&d=1199293388

But I could be wrong.

To say the least, this just goes to show that digital cinema has always had its finer points, points many people simply over looked.

Nice to see someone finally test this. Thanks!
 
Hey Jim,

Thanks very much for doing these tests. It's one of those cases where we all *know* about these issues but until someone does the test it's all speculation. It's great to see the actual results so that we can work around it and use the Red to it's full potential !

Cheers,
Seth.
 
To be fair, and to dispell some disinformation:

1) Not all cameras are 5600k native. Some are 3200.

2) Some cameras (most fullsize) use a filter wheel to physically transform some of the white balance before the electronic channel gain kicks in the do the rest.
 
Mike Prevette said:
1) Not all cameras are 5600k native. Some are 3200.
Click to expand...

Do you have any examples? I'm curious about it. It would seem like 3200 would be the optimal native white balance for cameras given how cheap tungsten lighting is. I don't personally know how you can change the cameras native white balance from a manufacturing point of view, but part of me thinks it has to with the type of filtering going on with each chip, for example, making the red filters used on a bayer patterned chip have more ND mixed into them, etc...
 
You must log in or register to reply here.
Back
Top