Using an 80/85 Filter to Balance

[Marc Wielage]

I was in no way involved with Tree of Life or New World, but from what I understand both of those films had some very odd stuff done with intentional overexposure and shooting mismatched stocks without filtration and then correcting in post. But Lubezki is at the forefront of pushing things in the grade and I think his approach is basically what you mentioned, pulling the correction filter when there's not enough light late in the day. Very similar idea to the 4x5 photographer who advised me to skip filtration both for color and reciprocity error and fix the drum scan in Photoshop.

Chivo is a sharp guy and totally gets what the rules are and when and where he can break them. Everybody else... not so much.

I've done the same kind of curve/denoise thing you're talking about, but it's a pain in the ass and takes time. A lot of post boils down to your available time and budget, and as long as the project can handle that, great. I don't deny that there are sometimes location issues that are just not possible to overcome, so mixed lighting happens. And I've seen some fairly hip, happening modern shows (like the Marvel shows, American Gods, and Legion) where they embraced mixed lighting and made it a part of the bizarre look of each scene. It can work as long as you're not expecting clean whites and natural fleshtones.

But I'm curious, when color is hard clipping (traffic lights, flares etc.), what approach do you take? All I can think of is somehow keying that area twice, once targeting the specific chunk of blown out color and pushing it to white, then soft keying the area around it with a mask and blowing it out a bit, too, and then using a global adjustment on the sat v Lum curve. But like.... beyond me... If you have any pointers about this I am curious because it's such an issue.

That would be telling! :blush5:

I'll give you a clue: It's possible to use the RGB Mixer in Resolve to compensate for clipped channel detail in some cases. It's a crutch, though, and is never going to be as good as getting the exposure right in the first place.

What do you mean by particularly when they're shot at 3200K though? Isn't the color temperature not baked into raw? This is a discussion that I think is really interesting because hard chroma clipping is the first thing that gives away digital to me but the discussion goes deeper than that into how various sources are rendered or even into how a sensor clips at various color temperatures and what kind of gain is used in each channel. Even color negative and slide film are different in this respect IMO and one reason digital and slide film have richer National Geographic colors than traditional color negative film used in cinema. Imo Arri is actually doing a great job emulating color negative, Red's current color is pretty good, too, but less slavishly color negative film-like. Have you seen Yedlin's infamous Nuke script demo where he shoots film and Alexa footage side-by-side and runs them through a Nuke script with no other correction? It's pretty amazing, the grain/color/halation/tonality is all really superbly matched but the biggest discrepancy is in how they render mercury vapor. It's a correctable difference on the hue v hue line to a perfect match, but that was the one thing the script doesn't procedurally at least get close. There are other discrepancies, including skin tones, but this is the one that stands out.

I think a lot of these cameras look kinda ugly at 3200°. It's even worse if you get 3200°K as a transcode that can't be changed from the Raw. And there are cameras that don't shoot Raw.

Being able to tweak exposure and color temperature is the biggest benefit of Raw for me. The problem with trying to compensate this with hue vs. hue or hue vs. sat is you can wind up tying yourself into a knot, with a kind of weird "stressed-out" picture that doesn't quite look natural. Every situation is different and it often requires a variety of tweaks to really nail it down, or at least get it to the point where it's reasonable.

 

[Anton Shavlik]

If you can avoid channel clipping you could also just rate the camera at a lower ISO when shooting in tungsten, no?

An 80A filter loses you two stops (or a little more) from the red side of the spectrum while leaving the blue side alone. This alone doesn't make the blue channel less noisy, it's only after you open your lens 2 stops to maintain the previous exposure that it gives the blue channel more light.

So long as you don't clip the red channel you could just open those two stops and forget filtration alltogether, and end up with a less noisy image overall than with filtration. Right?

I think the filters are designed to have a nice spectral curve and a linear effect on the light, so I can't imagine that the final image would be much different, even in that ketchup example. This thinking applies differently to film which doesn't respond in a linear way to light the way a silicon sensor does.



Also, science nerd here, sorry. It's not degrees kelvin, just kelvin. 3200K

 

[Björn Benckert]

If you can avoid channel clipping you could also just rate the camera at a lower ISO when shooting in tungsten, no?

An 80A filter loses you two stops (or a little more) from the red side of the spectrum while leaving the blue side alone. This alone doesn't make the blue channel less noisy, it's only after you open your lens 2 stops to maintain the previous exposure that it gives the blue channel more light.

So long as you don't clip the red channel you could just open those two stops and forget filtration alltogether, and end up with a less noisy image overall than with filtration. Right?

I think the filters are designed to have a nice spectral curve and a linear effect on the light, so I can't imagine that the final image would be much different, even in that ketchup example. This thinking applies differently to film which doesn't respond in a linear way to light the way a silicon sensor does.



Also, science nerd here, sorry. It's not degrees kelvin, just kelvin. 3200K

Correct, as long as you can open up to get good levels in all channels without clipping you are good. But in a ketchup scenario red channel will get a stop light before blue or green is even starting to fill up. So to get full "color resolution" in that ketchup you need to balance and then counter balance in post. To do so you could simply expose and shoot with your filter then shoot a chart and then renmove the filter stop down and shoot the same chart again. Then you got a reference for what the filter does and you got good amount of pixel color resolution in your ketchup.

Now in the above you can replace the word ketchup with bluescreen or any other high level R,G or B level objects in your scene that you find relevant.

We just did a parfume commercial where a girl is dancing "inside a ruby" and i had some discussions with the dp before the shoot if we should not counter balance in camera and he did not care a flying fuck about this theory of mine and tossed a few comments that possibly its needed on that red camera of mine that has so little lattitude, etc... And then explained the huge benefits of shooting alexa uncompressed raw... sure enough when all was done the material looked quite far from stellar and had all the kinds of trouble from low level b,g exposure. So its relevant for all csmeras, and it does not seam to mather if you have a top notch dp, some are still not grasping the concept of recording rgb.

 

[Matt W.]

You're a madman, Björn, but also probably a genius. That is fascinating and on such a higher level from where most DPs are working. My question is, what difference does this make in the 8 bit standard dynamic range deliverable that will make it to broadcast? How, quantitatively and subjectively, is the look of the ruby or ketchup bottle different than if a normal exposure without a filter were simply processed to push the saturation up to limits of legal broadcast. Is the concern superior tonality? A more saturated looking image because everything is nearing the peak saturation, similar to the tighter point cloud I noticed on the vectorscope for film vs digital (very very early Red)? It can't be a lack of chroma blocking (stop light, as you say) because neither exposure would be chroma clipping if both were exposed below chroma clipping in any channel.

I blame my own ignorance or maybe just my inability to visualize the difference in the final deliverable because it takes so many steps to get to the deliverable, even though I understand the theory. Compare John Seale overexposing the day for night in Mad Max (not that I loved how it looked, but I never like day for night) with this TV feature I worked on, also Alexa and with a rather high budget, where the DP shot day for night not only with an unnecessary and too strong 80A filter that muddied up the image horribly but also underexposing two stops for evening and four stops for night, literally block shooting scenes but stopping down two extra stops (nothing was even close to blowing out at normal exposure so that wasn't a concern, and you can imagine how four stops under plus a blue filter looked) and underexposing "night" so dark that shots had to be denoised just to lift faces to a normal exposure. Caucasian faces were far below 18% gray in the highlights. I see this stuff all the time. DPs shooting a flat image lit by one soft box and trying to hide it by underexposing many stops beyond what would be acceptable and the monitor looks good for client because it's too dark to see how flat it is and then post getting all the blame when the image is flat and noisy.

But if the final deliverable is limited by broadcast spec, what is the subjective visual advantage of filtering to ETTR with each channel (and I assume you advocate for exposing to the right whenever possible, something very few DPs I'd know would agree with, but which makes sense; the Dragon sensor looks best IMO around 320 ISO even if it's also fine at 2000).

 

[Björn Benckert]

You're a madman, Björn, but also probably a genius. That is fascinating and on such a higher level from where most DPs are working. My question is, what difference does this make in the 8 bit standard dynamic range deliverable that will make it to broadcast? How, quantitatively and subjectively, is the look of the ruby or ketchup bottle different than if a normal exposure without a filter were simply processed to push the saturation up to limits of legal broadcast. Is the concern superior tonality? A more saturated looking image because everything is nearing the peak saturation, similar to the tighter point cloud I noticed on the vectorscope for film vs digital (very very early Red)? It can't be a lack of chroma blocking (stop light, as you say) because neither exposure would be chroma clipping if both were exposed below chroma clipping in any channel.

I blame my own ignorance or maybe just my inability to visualize the difference in the final deliverable because it takes so many steps to get to the deliverable, even though I understand the theory. Compare John Seale overexposing the day for night in Mad Max (not that I loved how it looked, but I never like day for night) with this TV feature I worked on, also Alexa and with a rather high budget, where the DP shot day for night not only with an unnecessary and too strong 80A filter that muddied up the image horribly but also underexposing two stops for evening and four stops for night, literally block shooting scenes but stopping down two extra stops (nothing was even close to blowing out at normal exposure so that wasn't a concern, and you can imagine how four stops under plus a blue filter looked) and underexposing "night" so dark that shots had to be denoised just to lift faces to a normal exposure. Caucasian faces were far below 18% gray in the highlights. I see this stuff all the time. DPs shooting a flat image lit by one soft box and trying to hide it by underexposing many stops beyond what would be acceptable and the monitor looks good for client because it's too dark to see how flat it is and then post getting all the blame when the image is flat and noisy.

But if the final deliverable is limited by broadcast spec, what is the subjective visual advantage of filtering to ETTR with each channel (and I assume you advocate for exposing to the right whenever possible, something very few DPs I'd know would agree with, but which makes sense; the Dragon sensor looks best IMO around 320 ISO even if it's also fine at 2000).

For day for night shots you could benefit from shooting "muddy" as a clean picture but dark is not really how we see att night. filtering to make a resolution loss in that case is not a bad doing.

The finally delivery format does not matter so much. But what few see is when 1 or 2 channels are empty then thats actually much worse than cutting your overall exposure down by 1 or 2 thirds. Simply the sensor does not have anything in cup 2, and 3 or just very noisy info, data that is not correct. That will in the debayer be used to create the 3rd color chanel so also that last third chanel gets contaminated and less precise. So when you got that ( very blue underwater footage or such) then it will possibly look half ok if rendered blue as is but it will stil lack in resolution and have noise as the red and green is not used. So in such case if you tint towards red and open up accordingly and then drag kelvins to compensate back you get the same blue nice tone but with much more correct info for each pixels / better resolving resolution and better / lower noise. Thats no rocket science.

 

[Matt W.]

I'll leave the day for night discussion to others because I've never been fond of the technique in the first place.

Your description of the benefits of exposing each channel well makes perfect sense, particularly in the context of debayering.

 

[Blair S. Paulsen]

In the vast majority of cases, the best results come from shooting a reasonably balanced, well exposed image and "seasoning" it in post. That said, there are situations where shooting for post will yield better results, like Björn's ketchup example where a particular hero color needs to pop without looking oversaturated/unnatural.

I like to use the 1/4 CTB when the lighting conditions are sub-optimal - particularly under 3000K with low CRI and when underexposure is unavoidable. I have found that the loss of 1/2 stop is a worthwhile trade off for better balancing the color information available to the algorithm. Think of it as a "sturdier" digital negative for grading. YMMV.

Cheers - #19