Magenta highlights, balancing the green channel

[Noah Yuan-Vogel]

This thread is about finding an in-camera solution for magenta highlights.

PROBLEM:
So I keep seeing lots of footage from various RED cameras where white turns out very magenta, and it is clear from looking at R3Ds in REDCINE-X that this is because it is common for the green channel to have a lower video value than expected at clip level, presumably because it clips before any other channel. These magenta highlights are noticeable in at most ISO settings but most problematic and increasingly bad at ISOs below ISO800, presumably because as ISO decreases, fewer stops of highlight detail is stretched over the same range, increasing the contrast of highlights in that range. I understand this is fixed by the DRX option which is available in REDCINE-X but does not appear to be available for in-camera looks.

I am concerned that the over-saturated green photosites might also be affecting dynamic range somewhat.

SOLUTION?:
After inspecting several R3Ds containing clipped highlights, I found that generally for all images tested a tint value of ~27.5 would put green clip levels about where they should be relative to the red and blue channels and that this value tended to work to fix highlights at all ISO settings and in various other look modes.

So my question is, might it make sense to leave an optical magenta filter on camera at all times? If a magenta filter is available that equates to -27.5 tint, it should bring the green channel back down to where it should be relative to blue and red, possibly increasing dynamic range by preventing the green channel from clipping early. It would also allow us to correct those magenta highlights in camera and see the corrected images while monitoring and even record corrected images requiring less work in post. The camera would just always be left with the tint setting set to 27.5 (or G gain ~1.4) to correct out the magenta filter in images, but it could otherwise be relatively transparent.

FURTHER BACKGROUND:
From what I've seen, a surprising amount of footage shot on RED cameras gets out without this issue being fixed, so either people are not aware of it, not aware that it is easily fixable, or otherwise ignorant to any need to fix it. I've seen magenta highlights in the SCCE tests, Contagion, many finished shorts ive seen online recently and in the past, among others.

As I work with RED increasingly more frequently, I find I really want a solution to this since it is a significant issue in my mind and I would love to be able to clearly say to my clients that this is not an issue because there is are a number of very practical solutions for it that work very well. Id rather it not be an issue I have to be overly-vigilant about in post, trying to make sure whatever assistant editor or DIT or PA is handling or transcoding the footage makes sure to apply DRX to every shot, especially considering that it is never clear what R3D look options work in what editing or grading software. What if they apply DRX and grade the native R3Ds in some software that doesnt support that look option? Is that possible? Better to not have to worry about it.

I will test this when I have the opportunity and post results but as I have not had my Scarlet delivered yet I am not yet a RED owner myself so it is not so easy to test until I have a camera available to me.

Does anyone know if the tint values on RED cameras equate to particular Green/Magenta Color Compensation values?
Does anyone know what optical filter would be closest to the equivalent of -27.5 tint in a RED camera?

 

[Marc Wielage]

What happens when you shoot a gray card exposed at 50%, or a white card exposed at 100% (just before clipping on the histograms)? Are those tinted in any way? I would think a problem this severe would show up all the time, not just in clipping.

I generally haven't found this to be a problem in the footage I see. On the rare occasions I do see contamination in the highlights, we just desaturate the highlights or do a soft clip on them in color-correction. Even film is not always 100% pure white when it clips, depending on developing, filtration, color temperature, exposure, and other issues.

 

[Sidney L. Plaut]

that is what the tool DRX is used for. - in redcineX

 

[Vico Martin]

Noha, I fight with the magenta from some M-Xs every day. I own a R1 and a R1MX, this only happen with the M-X.
Usually because some times Camera guys can´t do a WB On-Set, but some times even after it.
I can usually compensate it editing the metadata in post, but sometimes this purpple pollution affects the tones, even in the low lights. I notice that there are some MX cameras more problematic than others.

But will test your -27.5, thanks man.

 

[Ian Vertovec]

might it make sense to leave an optical magenta filter on camera at all times? If a magenta filter is available that equates to -27.5 tint, it should bring the green channel back down to where it should be relative to blue and red

This isn't the best solution as it would considerably homogenize that area of the light spectrum. Very close to skin tones. Instead, I would fix the issue with a color correction preset HSL qualification focused on low sat, high lum pixels.

 

[Noah Yuan-Vogel]

This isn't the best solution as it would considerably homogenize that area of the light spectrum. Very close to skin tones. Instead, I would fix the issue with a color correction preset HSL qualification focused on low sat, high lum pixels.

Interesting, are you saying it would significantly affect the red and blue channels in a negative way? In that the optical filter would have an effect on the image that would be complex to the extent that a green gain or tint adjustment would not compensate properly? My understanding was that this was a relatively common practice for some when shooting with the Viper camera, though it was never fully clear whether the effect was positive beyond correcting images in filmstream mode. As for HSL, as far as I know such corrections are not possible in camera. For post, DRX in RC-X generally does a good job of fixing the issue, but I'm aiming primarily for a fix that would work in camera and would not require that all images coming out of the camera be adjusted in post simply to correct consistently unnatural highlights. If it can be fixed in camera, that takes a step and worry out of the post process that frankly seems to be overlooked quite often and most certainly would be overlooked if people handling the footage are unfamiliar with recognizing or correcting the issue.

 

[Graeme Nattress]

We put DRX in the software for a reason - it's the best way to fix the issue. Optical filtering only works correctly under one particular colour temperature of light, and would completely screw up the camera colorimetry.

Graeme

 

[Chris Nuzzaco]

I used to battle this all the time with my modified DVX100-A Andromeda system. Even DSLR's have this issue, and its really because the sensors don't have a true optically set white balance. If you download Darktable and load a Nikon NEF file into it and select "passtrhough" white balance, you'll see what looks like green tinted Viper filmstream footage. I have successfully filtered optically, but like Graeme said, it's not very easy, and It will kill your sensitivity. In a nut shell, the magenta filtering would only work if you were shooting under a very high color temperature light. If you wanted to shoot under tungsten, you'd have to add a very dark blue filter in front of an already very dark magenta filter. I was using around 90ccm. That's pretty heavy.

 

[Noah Yuan-Vogel]

We put DRX in the software for a reason - it's the best way to fix the issue. Optical filtering only works correctly under one particular colour temperature of light, and would completely screw up the camera colorimetry.

Will DRX be available in camera? Could you explain in further why the colorimetry is designed with white as magenta and whether there is some benefit to this behavior?

Perhaps I'm not understanding correctly, but I expected that the clip point of each channel would be independent of any lighting and that tint is generally designed to correct for such isolated green/magenta color casts. Would a magenta optical filter affect the image in more than just the green channel in such a way that it would not properly compensate for using the tint/gain setting to balance the white clip point of the channels? Where in my assumptions have I gone wrong?

If using tint or green gain to compensate for color casts screws up colorimetry should we not be using it to correct for green or magenta spikes in fluorescent or led lighting? Or are there complexities in a magenta optical filter that I am not taking into account? Thanks to anyone who can help shed further light on this issue.

 

[Graeme Nattress]

Practically nothing in-camera is independent of lighting. Any filter ALWAYS effects more than what you wish it to effect.

Graeme

 

[Noah Yuan-Vogel]

Even a particular channel's white clip point? (My primary interest in this thread) There is no change one can make outside of the camera to change that, is there? I understand that optical filtration could compromise colorimetry in some way that has not yet been detailed or quantified, but to some extent it may be a worthwhile tradeoff if doing nothing leaves highlights compromised somewhat and if the filter has only some minor downside or the downside can be balanced by using a bit less filtration.

I appreciate people's contributions to this thread, but it is very easy to say optical filtration shouldn't be done for somewhat nonspecific reasons. It is good that we have DRX to fix it in post, but I often shoot jobs with very quick deadlines and where I hand off footage and have no control over it in post and fixes in post just are not an option, so post options are great for some but really not what I am after, so my aim is finding a solution for this issue in camera. It would be great if someone in a position to test for solutions could do so, but I will not be able to do so until mid February.

Graeme, is there some way to enable DRX in camera, or will there be? Or perhaps REDgamma3 will have balanced, white whites?

 

[Graeme Nattress]

"Even a particular channel's white clip point? " of course because you're altering what is "white". It also compromises sensitivity as well as colorimetry.

How does putting something "in-camera" help when all the R3D files go through our RAW development tools?

Graeme

 

[Noah Yuan-Vogel]

"Even a particular channel's white clip point? " of course because you're altering what is "white". It also compromises sensitivity as well as colorimetry.

I understand if what you mean is that it may not be practical to think of the camera as a system isolated from light since it is designed primarily to capture and interpret light in a meaningful way. On the other hand, if you do isolate it, arent there properties of the camera, white point being one, that can be important to consider and understand in isolation?

I would think the camera design must take into account extremes and as well as, and perhaps really before, subtle gradations. For example, black balance occurs without any light, and even with no photons collected, but it is still a very important part of the system. White clipping seems similar to me in the sense that it CAN be considered essentially independently when seeking to understand the system. It doesn't matter what filter is on the lens or what lights there happen to be if the image is 20 stops over exposed for example, yet its behavior at that point is still relevant since there are practical situations where most of the frame for some reason or another will be significantly over exposed. It seems that the color of light that is far far overexposed and far far underexposed as interpreted by the camera is one of the few things that the camera body alone is responsible for, since no matter what you do to your lights/scene/filters, that property will always be controlled by the properties of the camera's sensor and internal processing. That is why I expected the only way to fix it would be through adjustments internal to the camera, and the optical filter I suggested would just be used to compensate for those internal adjustments.

How does putting something "in-camera" help when all the R3D files go through our RAW development tools?

Most people judge and expose their images primarily with their on-camera monitoring options. Some people even record those images for greater ease in post in some circumstances. Being able to apply all features in-camera is critical to having accurate monitoring and recording in those cases.

It seems practical to consider in-camera as separate from post tools since they offer significantly differing feature sets despite being built on the same look system. There are some recording HD-SDI/HDMI output and bypassing RAW, there may be others using other manufacturer's tools based on RED's SDK that may or may not implement the same features, and others that may use REDs tools but only to export, not taking the time to fix or change anything. The more that can be done in-camera, the more DPs/CameraOps can rest easy that they saw images that looked the way they wanted and handed off to post those images and cannot be held responsible for issues that may arise with their images just because someone might have forgotten the step in post where they have to apply additional look features that were not available in camera.

 

[Brett Clements]

"Even a particular channel's white clip point? " of course because you're altering what is "white". It also compromises sensitivity as well as colorimetry.

How does putting something "in-camera" help when all the R3D files go through our RAW development tools?

Graeme

Hi Graeme. Goose me here; we should be white balancing off a 50% grey card EVERY time huh?

 

[Graeme Nattress]

An 18% grey card is great for white balance, as is the RED Cambook or DSC or other charts. I try to put them in each lighting setup as a record of what was going on. They don't need to be in every shot or used every time, but having that record can really help in post if something to do with colour isn't quite right.

You can decide on a wb setting for each setup, or you can have one for daylight and one for tungsten.

Graeme

 

[Matthew Love]

Speaking of Magenta. I just got off a job where our B camera RED ONE MX would go Magenta until we rebooted the camera, then it would be ok for a small time being.

The footage would come out looking like this, the 2nd image is one that I tried to play with but in the short amount of time I had couldn't really get it anywhere near acceptable in my eyes.

I have the R3Ds on another computer and when I can I will post them here as an R3D still for some of you to play with if you'd like.

 

[Graeme Nattress]

Camera issues like that are best put through tech support as they know the hardware better.

Graeme

 

[Vico Martin]

Speaking of Magenta. I just got off a job where our B camera RED ONE MX would go Magenta until we rebooted the camera, then it would be ok for a small time being.

The footage would come out looking like this, the 2nd image is one that I tried to play with but in the short amount of time I had couldn't really get it anywhere near acceptable in my eyes.

I have the R3Ds on another computer and when I can I will post them here as an R3D still for some of you to play with if you'd like.

This is the kind of mgenta I´d detect on some MXs; try re-blance on any not burned white on frame with RedCine-X or change to colorspace2.
I also noticed that not using the WB presets on camera avoids the problem, I mean set 5600º manually instead of "Daylight"; and much better White Balance in each location or light setup and then move the ºK manually.

Anyway I only notice this kind of "super-magenta" on some MX units. Waiting for the Redcolor3 to show how it handle this.

 

[Tommy Davis]

I apologize for rehashing an old post but after having a slight freak out I wanted to provide some more useful info for anyone who discovers this issue later in life and is searching for an answer. When doing some outside tests we realized our sky (and anything blown out) was registering a magenta color. After reproducing the problem we realized it was dependent upon what ISO we were using. If we go below ISO 400 the highlights begin to shift hue and it gets worse the lower the ISO you go.

We shot a few reproductions starting at 100, 200 and then at 800. We placed a 500w fresnel shined at a white wall and then placed a person against the wall as well for skin tone reference. For every single shot if we adjusted the source setting to 800 ISO (or anything above 400) the blow outs were pure white. However, each of these clips if dropped below 400 would reproduce the nasty hue. Looks like I will be shooting the recommended 800 ISO and putting lots of .ND outdoors as I don't want to rely on DRX for everything - though it is extremely helpful in a pinch.

 

[Noah Yuan-Vogel]

I apologize for rehashing an old post but after having a slight freak out I wanted to provide some more useful info for anyone who discovers this issue later in life and is searching for an answer. When doing some outside tests we realized our sky (and anything blown out) was registering a magenta color. After reproducing the problem we realized it was dependent upon what ISO we were using. If we go below ISO 400 the highlights begin to shift hue and it gets worse the lower the ISO you go.

We shot a few reproductions starting at 100, 200 and then at 800. We placed a 500w fresnel shined at a white wall and then placed a person against the wall as well for skin tone reference. For every single shot if we adjusted the source setting to 800 ISO (or anything above 400) the blow outs were pure white. However, each of these clips if dropped below 400 would reproduce the nasty hue. Looks like I will be shooting the recommended 800 ISO and putting lots of .ND outdoors as I don't want to rely on DRX for everything - though it is extremely helpful in a pinch.

Yes in my experience the magenta highlights get significantly worse as you lower your ISO, but I have found they still exist in ISO800 and above. I assume the magenta highlights get worse as you lower your ISO because lower ISOs decompress the highlight range, fitting fewer stops of dynamic range into the same "luma" values, which effectively increases contrast in highlights and the increase in contrast increases color saturation. In other words as you move your grey point down you stretch the highest highlights down and increase the difference in clip point of the color channels, and that difference is the reason the highlights look magenta in the first place.

Unfortunately even now years later I still see strange looking magenta highlights in movies/commercials/music videos all the time and know that they were shot on RED. Perhaps it's too much to assume even bigger budgets will fix their magenta highlights with DRX or grading? I still don't really understand why the camera is designed to render white as pink such that all footage has to be fixed in post if clipping is not completely avoided.