CMOS sensor

[Gavin Greenwalt]

But isn't that exactly what you do when you rate the camera at 800 ISO? I don't see a problem with that.

Well if I'm shooting a greenscreen I will expose for 320 since I know exactly what will be in frame and can ensure nothing clips. The less noise the better.

[PatrickFaith]

In the "old day's" I would look at the spectral sensitivity curves of the film, and make sure that would match up with the spectral emission curves of the light source (i.e. http://motion.kodak.com/motion/uploadedFiles/TI2647.pdf ) . It would be real nice to have spectral sensitivity curves of the RED sensors. Would help out a lot in using modern light sources, time lapses, etc.

[Scott Crawley]

So you light your green screen with daylight sources then.(?)

Well if I'm shooting a greenscreen I will expose for 320 since I know exactly what will be in frame and can ensure nothing clips. The less noise the better.

[paulherrin]

Scott, couple things.

My understanding is that all the color science in an image comes down to the white balance. For accurate colors, that seems to be the most important thing to set well. That's why it's the first setting. So, (I could be wrong) setting the white balance to an outrageous number and cranking other parameters down the line will result in objectively different processing and is probably not recommended. Because different light sources put out various amounts of red, green and blue light, the raw image data must be interpreted based on the color temperature of the light (check out all the cool graphs and stuff on google). That makes the starting point a good white balance.

There are really no rules, but 'true' daylight sources seem to be the best all-around way to go, especially when colors are important (they always are...) like in greenscreen work. Tungsten makes greenscreen (and anything) difficult because it is inherently blue-deficient plus cameras just have trouble picking up blue (generalization) - so on film or digital cameras you get a really noisy, underexposed blue channel - which doesn't always make for the cleanest colors, or keys. On the other hand, if you're doing underwater greenscreen, you might be fighting to get your red channel where it needs to be... but that's a different story.

Plants like daylight, humans like daylight, cameras like daylight... the sun is a very 'healthy' source of light and the reason we have life and can see and make pretty pictures in the first place. You can't go wrong with daylight. That doesn't mean you and I can't get decent results with tungstens or leds or sodium vapor or whatever - and many have. However, when color fidelity is important you should at least understand what 'kind' of light you are getting, how the camera sensor 'likes' it, and what results to expect in terms of postproduction - including benefits, caveats, and workarounds through the whole process. Testing is good.

When in doubt, try this simple equation. daylight = life :)
(and don't forget the smile at the end, it's very important to balance the equation)

[Scott Crawley]

Paul,

I am beginning to enjoy when you take something as technical and dry as color balancing and make is spiritual. ;-)

I have a pretty solid grasp of balancing mixed sources, and color balancing. Thanks for the effort. My questions for Graeme were intended to determine what best practices should be when it comes to choosing between digital manipulation and overall filtration. It was an attempt to verify if one was inherently preferred, or cleaner than the other. The answer appears to be that within the daylight-tungsten range it doesn't matter. No surprises there.

The question that interests me more relates to the RCX looks panel. When manipulating color temp, lift gamma and gain or the color sliders, are you in fact executing the same underlying -/+ gain operations or are they somehow different beyond the interface?

[paulherrin]

hehe, glad you enjoy it. my personal definition of spiritual does not include 'immaterial' or 'incorporeal' - first because it is generally unhelpful to negatively define something (positivity is always good), but mainly because i believe the nature of spiritual things are intrinsically meshed within the physical - therefore the distinction is perceptual and mostly only useful for communicating the simple idea that the distinction is not necessary.

all that aside, color - and the perception thereof - has carried special significance for thousands of years across every culture. we like to keep things straightforward and objective with color science and color theory (based on scientific method). i believe that color affects us in ways that may not ever be fully understood, and i do not believe it matters that we have awareness or understanding in order for them to affect us.

and even all of that aside, i really appreciate the science and research done on graeme's part to enable creativity :) love that signature quote... he's probably one of only a few people that can answer those types of questions precisely... however, i did get the impression that the operations are not exactly the same and that adjusting white balance, to some degree, sets the stage for all the other transforms down the line. from a programming perspective, however - it is possible some of those controls may fine tune the same information in the file, rather than being an additional adjustment - not sure. that's definitely one for the code guys and it may be too delicate a tasting of the secret sauce...

[Andrae Palmer]

Native color balance is 5000 Kelvin

Native color balance for the MYSTERIUM X sensor is 5,000 KELVIN, but it may be electronically compensated for any color temperature in the range 1,700 to 10,000 KELVIN. White balance preset values include Tungsten (3200K) and Daylight (5600K) light sources. The camera may also calculate a color neutralizing White Balance value on demand by imaging on a standard white or 18% grey card.

<--- Scarlet, RED One and RED EPIC manuals... search for "native color balance"

[David Mullen ASC]

All sensors, CCD or CMOS, are less sensitive to blue wavelengths and therefore generate cleaner signals in the blue channel in daylight color balance -- guess what, so is color negative film. The difference between tungsten and daylight balanced film stocks is just that in the daylight balanced stocks, they can make the blue sensitive layer less fast (and therefore less grainy) to compensate for having more blue to work with.

I think the confusion regarding digital cameras comes from the fact that for years, 3-CCD video cameras often used a clear internal filter for 3200K and then orange filters for daylight. This is mainly because the cameras were designed to make 3200K "white" (zero gain in each RGB channel), probably for the same reason why most people used tungsten-balanced film stocks -- generally in lower light levels, the color temp was closer to 3200K, and in brighter conditions, it was closer to daylight, so it makes sense to start with a tungsten balance and filter for daylight where you have the light levels to compensate for the filter factor. But how Sony, Panasonic, etc. got CCD's that naturally preferred more daylight to be balanced for tungsten, I'm not sure, but probably a combination of signal processing and the filters in front of each CCD.

I once did a test on the Genesis, which has RGB-striped CCD, and found that the blue channel was cleaner when the camera was set to daylight balance, as opposed to using an 85 filter. I'm sure the F35 is the same since it has the same sensor in it.

[Mike P.]

I once did a test on the Genesis, which has RGB-striped CCD, and found that the blue channel was cleaner when the camera was set to daylight balance, as opposed to using an 85 filter. I'm sure the F35 is the same since it has the same sensor in it.

Have you found the same characteristic to be true for Mysterium-X? Graeme et al at RED suggest you colour balance first in RCX as it makes for a cleaner image for looks processing, but it'd be nice to know if it'd be "cleaner" to always set the camera to 5600 (or 5000k) and then use the blue curve to compensate (under tungsten). Although something tells me that'll be the equivalent of a double-negative (ends up the same in practice...) It'd be nice to be able to "set it and forget it" ala ISO800, and know you've got the cleanest/broadest image possible by default. Plus it'd keep it simple for offboard/prores recorders too.