Home of Professional Cinematography since 1996
> RED & Lighting For The Native Sensor Kelvin Rating
Published : 13th April 2009
Been following the discussion re blueing up for the sensor on RED. I have now shot several green screen jobs on RED & since build 15 have felt no need to use daylight lighting or 80A or 80C etc. Testing of both 5600 & 3200 showed no appreciable difference, especially for keying. My only disclaimer here is that I have not tested blue screen, so can't comment on noise etc in that regard.
>> Testing of both 5600 & 3200 showed no appreciable difference, especially for keying.
I'll agree with that from a keying perspective, but not for noise.
I've posted my test here a couple of times, so I won't post it again, and it clearly shows decreased blue noise under tungsten light when using an 80D.
The test was shot in front of a green screen and the quality of the edges was unchanged by filtration.
And as Mark W. so eloquently stated (for free): you're not adding blue with a filter, you're only subtracting warmth (which includes red).
You're doing a partial white balance using optical tools in front of a sensor that likes daylight more than tungsten because of the way it sees blue (not well) and red (very well indeed).
Art Adams | dp
San Jose | CA | USA
>>Testing of both 5600 & 3200 showed no appreciable difference, especially for keying.
As a compositor, I have to say that I am surprised that you have not seen any difference in how the image keys. After doing some tests, with an object put in front a well lit green screen and making sure the subject has Motion Blur and DOF to be more in line with what happens in real life, and to push the keying process a bit, we found without a doubt that there is some differences in the keying. I will say that if pushed to get the shot done a comp artist can get it to work. But the image that was Tungsten, showed holes and chatter along the edge of the image, whereas on the daylight image the keys where very clean and showed very nice edges. We also did a range of exposures and raising it did improve the Tungsten a bit, but did not resolve all the issues. But when we used a 80B in this case it brought the image much closer in line to the results of the daylight lighting.
So since the RED camera is balanced for daylight it is very obvious to me that you want to work in Daylight as much as possible even when shooting Tungsten. We should also consider the colouring and texture of the subject as that can affect the outcome also. (i.e. reflectivity etc)
I also just want to note that it was said that no appreciable difference was seen. It could be said that what I saw could be noted at no appreciable difference, but if I have to do more work than I need to because of no filter, I vote for the filter.
VFX Supervisor | Compositor
Enigma Studios Inc
I think that we may be getting a little confused here about what works best with a RED as regards to CT.
As someone already pointed out when you are getting in to seriously low level, as you do shooting available light, blue filters on the camera will not help they'll just pull the red level down. As this is what there is most of at night it means that your overall level will drop considerably. There will be no advantage in the highlights but you will pull up noise in the shadows.
An example, shooting in the customs area of the Oresund bridge at 2am we were rating the camera at 500, using superspeeds wide open and only just getting an image. In a situation like this ANY coloured filter will make things worse.
On the subject of green screen I don't think that it matters what your foreground CT is as long as you're not clipping the image. What matters is the screen and you'll always be best off here using super green tubes.
Geoff Boyle FBKS
Mobile : +44 (0)7831 562877
>>On the subject of green screen I don't think that it matters what your foreground CT is as long as >>you're not clipping the image. What matters is the screen and you'll always be best off here using >>super green tubes.
Beat me to it... I have not tested green illumination for RED in the rigorous way I have tested many film stocks and other electronic cameras... but I was going to say that color temp is a different issue from matte extraction. extreme case being a silhouette, obviously.
The closer you can get to a monochromatic green, the better off you are going to be... pretty simple...you want a green screen that exposes the mostest on your green filtered photosites and the leastest on your red and green filtered photosites...but the Bayer patterned elephant in the room is the de-mosaicing that you have to do in order to create a color image which allows you to pull a key.
As we have found with other Bayer sensor cameras, WHICH algorithm or combination of algorithms you use in order to create a color image really affects how well you can extract the edge of the foreground elements.
I am not a compositor, and I don't play one on TV...but some day when I have a chance, I think I will take some clips that have been de-bayered several different ways and see which ones work best in which situations... we did that this spring for a Phantom job...the de-mosaicing (de-bayering) algorighm that gives you the best edges against the green screen might be ones that give you some of that rainbowesque ringing along some flyaway hairs... it might turn out that it is worth doing two different transcodes to DPX using two different de-bayering algorithms - one for the edges and one for the hair and foreground color.
why not? it's only time and money that you presumably saved shooting with a RED in the first place...
Mark H. Weingartner
LA-based VFX DP/Supervisor
I tried the 80 filter(large green screen cyc,) but after setting up/ lighting, I would have needed more lights and budget wouldn't allow for it, so I didn't use it. Compositor was happy w/ the results.
I did a test weeks before w/ blue screen, HMI's and CTB on tungsten fixtures- I left the camera at the native color temp-it looked very nice-I even put aluminum foil on the test actor's mouth(laughs) to simulate the singer's look(black skin with platinum grill) since I was also concerned w/ the RED's highlights under extreme contrast … But in the end we found a green cyc, had to go with that, balanced for tungsten. In any case, CTB on space lights didn't seem like a good idea anyway.
On the following shoot we had a white cyc, and just went w/ tungsten balance, same deal, not enough levels for the stop I wanted(camera at 320 ASA by the way)
I could try to use the filter again providing I have enough lighting next time.
>> On the subject of green screen I don't think that it matters what yourforeground CT is as long as >>you're not clipping the image. What matters is the screen and you'll always be best off here using >>super green tubes.
I agree also, if you are using green tubes with the green screen that will help as it evens out the screen and saturation levels will be good also and the levels on the GS also help in the keying process so if you lift from say 50IRE to 70IRE you can expect better results with the RED camera.. Our tests where just testing against Tungsten and Daylight on the GS. So depending on your method you can end up with different results all the time. The compositor may look at it and say, yup that works, cause they can just do some simple tweaks and all is good. In others if they had the misfortune of doing Blue screen with RED with Tungsten lighting they could be very unhappy.
Just as a side note "Keyers" themselves have some different methods and can produce slightly different results and so one comp artist may get fantastic results on flame and not so good using Keylight.
If you have a large GS area including the floor, and you are using tungsten lighting and it bounces onto the actor you can have spill that is harder to deal with than if it was of purer green.
I am sure you all know this, but I like to put it out there anyway.
VFX Supervisor | Compositor
Enigma Studios Inc
>>you're not adding blue with a filter, you're only subtracting warmth (which includes red). You're >>doing a partial white balance using optical tools in front of a sensor that likes daylight more than >>tungsten
And isn't this really made necessary by the limitations of storage? That if one just shot a tungsten scene with daylight color balance, the color channels would simply be out of alignment and thus would need greater bit depth to store them accurately, albeit inefficiently? Because going to tungsten balance is more or less just amplifying the blue channel to bring it more coincident with red and green?
Which of course makes it look normal, and also easier to store. Would HDR, maybe storing RAW as a variant of float TIFF, help?
Santa Monica, CA
>> You're doing a partial white balance using optical tools... isn't this made necessary by storage >>limits? That if one shot a tungsten scene using daylight sensor, the color channels would simply >>be out of alignment and would need greater bit depth to store?
[Some editing performed on lines above to fit with CML quoting limits.]
In the case of the RED ONE, it's more a limitation of the sensor itself. There are only about 8 stops between saturation and excess levels of noise in the shadows (note: I'm NOT saying the R1 has a dynamic range of 8 stops, only that its shadows get very noisy!
Actual dynamic range is rather a bit more than 8 stops).
If you shoot under tungsten, your blue channel will be a stop or two "underexposed" compared to red and green. If you then white balance it in post, you effectively boost blue two stops, and boost blue's noise two stops, too. (Balancing in-camera really does the same thing, since all you're doing is adding white-balance metadata to the raw file.)
There's plenty of bits in the raw data, it's just that the sensor noise is rather strong. Both for normal photography and for color keying it's visibly better to keep all the R1's channels "exposed to the right" as much as possible, to keep the sensor in its relatively narrow sweet spot between noise and clipping. Optically color- correcting the scene with an 80-something filter channel-balances the sensor to the scene and gets all the channels into their optimum exposure levels.
Adam Wilt / filmmaker, Meets The Eye / writer,
provideocoalition.com / Mt View CA USA
CTB filtering for the RedOne:
I love the idea of a filter purpose built to raise the color temperature/white point with as little light loss as possible. Several people have contended, some with a fair amount of test results to back up their assertions, that raising the CT makes for cleaner keys and better noise characteristics. Other have argued that when light levels are low it is better to keep every photon you can so that the red and green channels can provide grist to extrapolate the blue information.
From what I've seen there is an issue in terms of the "subjective noise awareness" factor of the viewer when the blue channel is starved and base level sensor patterning reveals. For this reason I advocate the "careful" use of blue filtering, especially in available light situations where the native CT may be less than 3000 Kelvin.
I would not give away a stop or more just to nail the Mysterium's native 5000K, but I would, for example, give away 1/3 stop to get to 4000K. In a perfect world I would carry 3 filters of differing strengths designed specifically to compensate for the lack of blue sensitivity inherent in the current crop of CMOS sensors.
It is my belief that Graeme's de-bayering algorithms perform best when the inequality of the color channels is more moderate. This advice is based on tweaking a lot of underexposed footy in Scratch in full res, full debayer with an eCinema reference monitor.
I am a big fan of Schneider filters and am stoked to see them jump in with CTB options. I believe the MIRED shift of traditional 80 series filtration is based upon assumptions that are overkill for the RedOne's Mysterium CMOS sensor/processing combo and results in more light loss than necessary. I consider it silicon chip color response compensation and prefer a filter that does exactly that.
Blair S. Paulsen