2020 - Cameras and IR contamination discussion.

[Alex Lubensky]

Hey everyone, I've been on a test recently, where Sony Venice was presented (Sony is not that popular on our market, still no Venice for hire). It was set up against Alexa Mini LF, utilising Cooke S7 lenses. Looking into the image everyone noticed Venice had a magenta tint to black/dark clothing, while the charts where on point. Alexa had a light green tint to the wall behind the person (Venice grey wall was on point, just as in real life), which led me to suggestion - there's no IR Cut on Venice OLPF or sensor, while the Alexa has it. I've asked the AC to change the internal ND to an occasional 4x5,6 good old Tiffen one, with no IR cut - and my suggestion was right. It dramatically revealed the absense of the IR-Cut inside the Venice, while using 4x5,6 IRND of-course cut the IR, and presented some green tint to the image.

We didn't record any Alexa footage with occasional Tiffen ND's just because it was pretty much the same as with internal ND's.

My general question is - which cameras nowadays have the IR cut applied to the OLPF/Sensor (And, maybe, someone know is it the OLPF or the sensor tech?). Which cameras you should use with IRND, which are OK with old ND's from the film days?

I do know MX and Dragon sensors in DSMC1 bodies have no IR-cut on sensor/olpf.

 

[Nick Morrison]

I do know MX and Dragon sensors in DSMC1 bodies have no IR-cut on sensor/olpf.

I don't think is true/accurate.

Phil?

I thought all RED sensors from Dragon on up had IR cuts in them?

 

[Bob Gundu]

I don't think is true/accurate.

Phil?

I thought all RED sensors from Dragon on up had IR cuts in them?

Dragon in DSMC2 bodies was a big difference than DSMC1 when it comes to contamination.

 

[jake blackstone]

There are no cameras, that immune to IR contamination. I had seen plenty of contaminated images shot on Alexa. If you stack enough ND filters, any camera will fail the IR contamination test. It is imperative to know just how far you can push a given camera under given conditions. High temperature and elevation also can greatly affect the IR contamination.
Ok, I just looked at the images and I think, at worst, it extremely mild case of IR contamination, if any. I say that, because the earliest sign of IR contamination is the loss of saturation and channel separation. In this case, it seems, that Venice color chard actually is higher in saturation. It would be very helpful if both cameras were matched a bit closer, as right now it's almost impossible to say which is the culprit- the color mismatch or the IR contamination.

 

[Patrick Tresch]

M sensor had contamination on normal ND values (ND6 to 21). I'm not sure 100% but I think MX had it (MX was the frist sensor in the DSMC1 body). But Dragon definitly doesn't have any (in normal ND values ND9-21). It's not due to the body it is in, but due to the OLPF of the sensor.

 

[Alex Lubensky]

ND's used during the test where 1.2 and 1.8. That's a pretty common value. While I'd agree - the IR contamination is in OK range when you use internal ND's on Venice, it's still can be a surprise under some conditions. Wish I've rec the shot with my black jacket, it was way worse than the one we've recorded.

I own Scarlet MX and IR Contamination is there, you have to be careful for the ND you use. Most recent are fine, though.

I'd rather like to hear your experience, when you've got into trouble with IR.

 

[Phil Holland]

I don't think is true/accurate.

Phil?

I thought all RED sensors from Dragon on up had IR cuts in them?

There is absolutely an IR Cut/Block Filter in the Venice. The Macbeth (I guess I need to refer to these as ColorChecker Classics at some point) tells the entire story actually. Look at the variance of the Internal NDs in regards to the White Balance on the chart.

No digital cinema camera on the market made from different manufacturers matches each other out of the box.

That older Tiffen IRND is producing an obvious color cast. I need to stress that anybody using older dye based IRNDs is ignoring far, far better filtration via the nano-metal coatings. Way more precise and not a tinted green filter with wild variance. The older IRND concept is really a concept of "yesteryear".

If you want the best filtration, these are what you are after:
http://phfx.com/tools/redThing_REDDSMC2_NDFilterNotes/phfx_REDDSMC2_NDFilterNotes.jpg

I haven't tested a couple other good ones I know some professionals are using like the Arri FSNDs, but I suspect they are also likely to be added to this list once I get around to it. ALTHOUGH. Something does concern me in this test.


In the RED world part of the OLPF, which is really a Color Science Filter, you get various filter stacks and layers that provide different levels of IR and UV blocking, cutting, or absorption. The Skin Tone Highlight OLPF being the strongest of the bunch to protect from IR Contamination, but the Standard does a damn good job. It's a bit more complicated than that, but all the current OLPFs do a decent job.


My biggest note really is stop using dye based filtration. Which is tough considering the shear quantity of classic NDs out there in the rental pool. But as somebody fortunate to hand pick my Schneider and Tiffen NDs on a lightbox to ensure color stability and accuracy, I can tell you I picked mine out 15-30 copies each. Not everybody gets to do that. The dye stuff is much like baking a cake, though rather controlled, it still leads to slightly different results.

The newer standards of IRNDs (which they should have named something else to avoid confusion from the dyed green solutions) are superior when it comes to color accuracy. It's always important to test on the particular camera and even lens (oh hi coatings) combo to ensure expected results. But most of you for years have seen my ND tests and I am testing them on the usual suspects pretty often when it comes to cameras and glass.


ACES is a cool workflow, but you still will need to white balance correctly if you are looking for a closer match and control other variables that could create a larger difference. In fact, it's even deeper than that in many circumstances. On a good day ACES workflow gets cameras all into the same working space and ideally should be close.

It seems often that I come off negative on ACES and I need to make it clear I am not, it's actually similar to RED's IPP2 in some ways and about to be more so. It's a sensible way of doing things and will get much better and cooler versions 1.4-2.0. It's built on the Color Correction to Color Grade mentality where we previously would color correct to "match" all cameras used on a production to get them all in a similar spot before moving to color grading process. These days many people think that simply white balancing leads to a matched image and nothing could be further from the truth. There's reasons there are many patches on these color charts and color charts with way more patches as well as those designed to be more informative on the scopes. Even still, there's more layers to this sucker. When working on those production you would end up having stuff that would work for an individual camera and/or even scene then push that out as a LUT to make sure editors, VFX, post all had access to that correction and eventually the final look which either would come back through me or I would also create a hero LUT depending on where it was getting finished. Most big post houses are inclusive now, so it's usually your final destination.

Manufactures certainly do their best to color match their sensors on the calibration level, but where this gets really fun is grabbing 20-50 of whatever camera and putting them in a very controlled setup with the exact same lens and lighting.

Not to bore you all, but back when were were doing the earlier days of what would become larger photographic domes for human texture and model creation I noticed copy variation to the extreme on our 50mm primes. And at the time we could order a lot of them, similarly I had to create matched batches from I don't even remember how many lenses for those rigs. We would shoot charts for that as well, but it was much nicer to have a closer starting point than not and then deploy per camera/lens combination corrections. Example A cam always got this LUT, B cam always got this LUT, etc. Those rigs quickly got up to many, many cameras and you can imagine what a pain that was.

 

[Alex Lubensky]

Personally, I just use the Arri FSND, when I'm working with my preferred rental house, they are the most clean ND's I've used so far. Otherwise - just stick with what the other rental has. The IR cut on Venice seem to happen not on the OLPF, but on the internal ND (the last two clips prove that - If you stick a regular, old film-days ND into the mattebox, the result is a heavy IR contamined image). Of course the cameras are not balanced one to another, it's just an out of the box test (Venice was literally out of the box).

The thing is - Venice cuts the IR spectrum a little bit less with it's internal ND's than the Alexa Mini LF. I'm unable to do more tests right now, due to the virus situation, but I'm curious how different cameras react to IR on the OLPF - where do they place the cut and so on.

 

[Phil Holland]

Though things have improved vastly on this front, I do want to share what things were like back in the early RED Mysterium-X and Alexa Alev III days.

Most here saw my ND tests over the years, but IR contamination was really, really bad at stronger ND densities with these two sensors early on, I'm talking as little as a 0.9/3-stop ND it would begin to transform the color and have a negative impact on the image. It was early days honestly for both companies and things evolved for the better.

Here's an old frame showing the hardcore impact of the magenta IR color cast, upper right frame:

That particular frame was shot on a Leica-R 60mm f/2.8 Elmarit with a heavy backlight flare, essentially the worse case scenario. That piece ended up with a very stylized color grade, but you can see that it was mainly about correcting the magenta cast and adding actual black back into the mix.

To tackle these issues there were a few different ways. Common in the Arri world was to utilize an ND filter with a Hot Mirror to get better results. Eventually filter companies began searching for solutions, these first came in the form of an "IRND", which initially were dye based much like regular NDs with some sort of tint offset, which was never destined to be the correct solution, but rather a stop gap.

Here's an example of RED Mysterium-X in direct sunlight with if I recall correctly a Schneider Platinum IRND, which attempted to remove magenta by correcting towards the green side of things:

You can see the rather distinct and awful green cast because a tinted method would never account or correct things the right way. Also in that sample you can see my IR color correction, which was then used as a LUT. I created them back then per ND density strength and similarly did the same for other cameras as well to be used on various projects.


Anybody who's not new here knows my deep love for Dragon with RED launched that sensor in DSMC bodies. It improved color and dynamic range in a variety of directions. It provided much, much better IR protection with good old regular dye based NDs. However, a few companies and Mitomo, Formatt-Hitech Firecrest, etc came up with the newer sophisticated nano-rare earth-metal coatings to create much better and more consistent color well over standard dye NDs.

Here's my first public test of the Firecrests on Dragon, no color correction straight out of camera with at the time the only OLPF LLO:

That was before the days of the Standard OLPF, the Skin Tone Highlight OLPF was out, both do indeed provide better color and better IR protection over the Low Light Optimized OLPF. It was also before the Motion Mount which also corrected IR contamination. And it was also before IPP2 which provided better color overall. What is impressive though is even with the thinner IR cut on that uncorrected output of the Firecrests and Dragon, color was stable at really strong densities like 2.7 (9 stops). Most cinematography occurs between 0.3-1.8 stops of ND (1-6 stops), which sensors being more sensitive I certainly see 2.1 used often. I have modern filtration typically up to 3.0, or 10 stops, as well as a few stronger densities for special imaging needs.

So more or less in the RED world these are issues of the past, but I still do recommend using these newer style of IRNDs (which should have just been named something different and some companies do). I still keep my dye based filters when I need to do "weird stuff", but truthfully I'm rolling with the Tokina IRNDs or Firecrests 99% of the time in 2020. I continue to test all ND options to at least give you guys and essentially the entire industry advice on what works well. Here's stuff I aggressively tested:
http://phfx.com/tools/redThing_REDDSMC2_NDFilterNotes/phfx_REDDSMC2_NDFilterNotes.jpg

Keeping brand favoritism out of the equation. Some of those I like better than others, but all of those "work well" in my eyes.

As I mentioned I never had the ability to test a couple options like the Arri FSND, but people I trust who also have more discerning eyes seem to be happy with those too. When the world starts spinning again I'll grab a set to test out.