As cameras get more and more sensitive to light, infrared (IR) pollution becomes a bigger issue as stronger neutral density filters are used to maintain exposure. IR pollution ends up changing the color rendition of the entire image and can result in some VERY funky color shifts. (I have seen dark green leaves change to magenta for example.) These color shifts can be next to impossible to get rid of once they are recorded into the digital image. There are creative ways to get rid of this in post, but many of these solutions - like rotoscoping and painting out the problem - are cost prohibitive. Besides, a "problem" like this should have been corrected in the camera at the time of filming, not left to post. Always remember this mantra, and practices it daily: "Fix it in camera, NOT in post."
Knowing that IR pollution is a problem, camera manufactures are building in some IR blocking into their sensor design (Usually in the OLPF). Additionally, filter manufactures like Schneider Optics, are designing the next wave of neutral density filters to help address this problem. In fact, Schneider Optics just released their Platinum IRND's which offer the latest in IR blocking technology and dye's. These filters are designed to work with all digital cameras and have been especially tested with the Red Epic and Arri Alexa. When looking at IRND filtration, there are many options out there, all with various levels of success. Other options include Hot Mirrors and even dedicated IR filters. However, I have never been satisfied with either solution. As the name implies, Hot Mirros, are in fact mirrors. (Shocking I know...) And that creates its own unique issues on set with talent, lights, and reflections. Personally, I like keeping the number of filters to a minimum in the matte box to preserve as much detail as possible, and avoid internal reflections when working with multiple filters. So, any time I can get rid of a filter by combining it with another filter, that is the route I want to take.
Since Red recently released their Epic-X cameras and Schneider released their Platinum IRND filters, I wanted to put the two together and see how they stacked up in comparison to what I have used in the past. I have taken stills from the video above and created frames for easier comparison and evaluation. My thoughts, as well as links to the original R3D's are below. If you want to play with the frames yourself, you'll need the latest version of RedCine-X Pro which can be found here or here.
The Control Frame
This test was shot using the Epic-X along with a 100mm Red Pro Prime. My custom chart was lit using a tungsten Source 4, and using my Sekonic C-500 color meter, the light read 3,300k. The white balance was set according to the color meter to give a completely neutral / base reading to allow me to see what was actually going on in the image. If I had used the camera to white balance itself, I would have lost the comparison, and not had an impartial reading to reference. And the great thing about working with RAW files, is that I can adjust any of the settings later! Shutter speed was adjusted in-between shots to compensate for the exposure differences. That data can be found in the metadata of the R3D's and read in RedCine-X.
This chart consists of a Macbeth color chart and a selection of black cloth made from different materials as each one responds to IR light differently. This frame will be used as the reference frame for how the different fabrics should look without any IR pollution. Unfortunately, for this test I was not able to get a hold of a Formatt 0.3ND filter, so it is not in the comparison frame. Fortunately, at 0.3ND there shouldn't be much, if any IR pollution. However, theory and practice don't always line up. So if you are going to use a Formatt 0.3ND with the latest generation of digital cameras, I recommend that you test it out for yourself first. I have used a sub-selection of the chart above to create the following charts ...
Without IR
The first question I have when working with any digital camera is where does IR pollution start to be a problem? Based off of this comparison of filtration WITHOUT any IR blocking, I can see a slight color shift (a redding of the blacks) at 0.6 and 0.9. However, it is very a very small difference at 0.6 and just visible at 0.9. I am ultra critical of the final image & I want this addressed in camera so I'll be using IR filtration starting at 0.6ND. But it should be noted that at 0.6 & 0.9 most people will not notice it, and the shift is so minor that it can be adjusted / corrected in post, if for some reason it wasn't filtered correctly in camera. (Fix it IN CAMERA, NOT in post.) At 1.2ND and above, IR pollution becomes VERY noticable. It is at this level where colors are being dramatically effected, and where there will be noticeable issues in post. Unless that magenta look is the intended result ...
So my recommendation is to always use IR filtration starting at 0.3ND. Yes this is a conservative recommendation, but it will save time in post, and it will help keep the blacks as black as they should be.
0.3 IRND
At this level of ND there is not much going on with the image, which is a good thing as this is not very heavy filtration. If there were extreme differences, I would be surprised. But I do think it is important to point out a slight color shift that is happening in-between the two samples that have the white balance set according to the color meter. Check out the following vector scopes:
By comparing these two vector scopes I can see that there is a very slight shift towards green. And this difference is noticeable if you look for it in the sample above. It should also be noted that I have yet to come across an ND filter that is truly 100% neutral. All of the ND's I have worked with have had a slight color shift to them. In an ideal world, they would be 100% neutral, but that is not the world we live in. Instead, the goal is to be as neutral as possible. And with IRND to block out as much of, if not all of the IR pollution as possible. That pollution will show up in this test as a bias towards red & magenta. And that bias is not happening at all here, which is a good thing.
Now let's see what happens when we white balance this image in post:
That's what I like to see- there is almost no perceptual difference between the two! The vectors fall in almost exactly the same places on the scope. And when I look back at the reference frame the two images that have been white balanced in post are visually identical.
(By the way, because I am working with a RAW file, white balancing in post is the same as it would have been if I had used the camera to white balance during the shoot. So if you are working with a non-raw camera, I HIGHLY recommend that you white balance AFTER adding any ND filtration. This should remove any color bias the filter has in it. It will NOT remove IR polution, as that is a completely different beast you have to deal with.)
0.6 IRND
After evaluating this test, it appears that the Schneider filtration still has that slight green tinge to it, while the Formatt filter, even with the Rosco TruColor IR filter has a red tinge to it. Lets take a look at the vector scopes to confirm what is going on.
No Filtration / White Balance: Meter
Schneider Platinum 0.3 IRND / White Balance: Meter
By comparing these two vector scopes I can see that there is a very slight shift towards green. And this difference is noticeable if you look for it in the sample above. It should also be noted that I have yet to come across an ND filter that is truly 100% neutral. All of the ND's I have worked with have had a slight color shift to them. In an ideal world, they would be 100% neutral, but that is not the world we live in. Instead, the goal is to be as neutral as possible. And with IRND to block out as much of, if not all of the IR pollution as possible. That pollution will show up in this test as a bias towards red & magenta. And that bias is not happening at all here, which is a good thing.
Now let's see what happens when we white balance this image in post:
No Filtration / White Balance: Post
Schneider Platinum 0.3 IRND / White Balance: Post
That's what I like to see- there is almost no perceptual difference between the two! The vectors fall in almost exactly the same places on the scope. And when I look back at the reference frame the two images that have been white balanced in post are visually identical.
(By the way, because I am working with a RAW file, white balancing in post is the same as it would have been if I had used the camera to white balance during the shoot. So if you are working with a non-raw camera, I HIGHLY recommend that you white balance AFTER adding any ND filtration. This should remove any color bias the filter has in it. It will NOT remove IR polution, as that is a completely different beast you have to deal with.)
0.6 IRND
After evaluating this test, it appears that the Schneider filtration still has that slight green tinge to it, while the Formatt filter, even with the Rosco TruColor IR filter has a red tinge to it. Lets take a look at the vector scopes to confirm what is going on.
Schneider Platinum 0.6 IRND / White Balance: Meter
No Filtration / White Balance: Meter
Formatt 0.6ND + Rosco TruColor / White Balance: Meter
As you can see from these vector scopes, the Schneider filter does have a slight green bias while the Formatt has a slight yellow / red bias. But what happens when the image is white balanced?
Schneider Platinum 0.6 IRND / White Balance: Post
No Filtration / White Balance: Post
Formatt 0.6ND + Rosco TruColor / White Balance: Post
The good news is that the white balancing brings both images back to being very close to neutral. Neither filtration is vastly different here. And the visuals confirm it. It is interesting to note that the vector scope for the Schneider filtration yields a slightly more saturated image, while the Formatt filtration yields a slightly less saturated image.
0.9 IRND
Here is where the differences REALLY start to stand out between the filtration. Looking at the Schneider filter I can clearly see a green bias in the image, and a red bias in the Formatt filter. The vector scopes confirm what I'm seeing in the image. The Formatt pushes the entire image slightly toward red, where as the Schneider has a much stronger push towards green,
Schneider Platinum 0.9 IRND / White Balance: Meter
No Filtration / White Balance: Meter
Formatt 0.9ND + Rosco TruColor / White Balance: Meter
And now what happens when I white balance ...
Schneider Platinum 0.9 IRND / White Balance: Post
No Filtration / White Balance: Post
Formatt 0.9ND + Rosco TruColor / White Balance: Post
Again each of the images comes back to neutral and I am not seeing any presence of IR contamination in my image. The Formatt filter almost exactly matches the neutral image, with just a pinch of yellow to it. While the Schneider filter is not as close to neutral as the Formatt. The Schneider is also slightly desaturating the image.
1.2 IRND
Wow, now things are getting colorful! Visually speaking, when comparing the meter balanced image, the Schneider filter still has a green bias, but it doesn't feel any stronger then the 0.9ND. While the Formatt filter is noticeably more red / magenta then the 0.9ND. It looks like IR pollution is creeping into the image. And after white balancing in post the Schneider filter hold up well as the green cast has been removed, and there is no visible IR pollution. However, this is not the case with the Formatt + Rosco filter, the IR pollution can still be seen in the fact that the blacks are now brown. (They still have some red in them.)
What does the vector scope have to say about this?
Schneider Platinum 1.2 IRND / White Balance: Meter
No Filtration / White Balance: Meter
Formatt 0.9ND + Rosco TruColor / White Balance: Meter
What I saw visually is confirmed technically. The Format + Rosco filter has a strong red / magenta bias and the Schneider has a strong green bias. Now on to the post balanced results ...
Schneider Platinum 1.2 IRND / White Balance: Post
No Filtration / White Balance: Post
Formatt 1.2ND + Rosco TruColor / White Balance: Post
White balancing definitely helps bring these image back into alignment. The Schneider filter is less saturated, and still has a very slight bias towards green. Visually, it looks like a better match then the Formatt + TruColor. The Formatt + TruColor has an interesting push towards yellow and very slightly towards cyan. This is resulting in a brown coloration of what should be black in the image. So it looks like some IR is getting through ...
1.5 IRND
Now I am seeing very strong color shifts in the meter balanced image. The Schneider filter seems to be headed even more into the green and cyan direction and the Formatt + TruColor feels like it isn't even blocking IR. White balancing the image seems to bring the image back to neutral with the Schneider filter, where as with the Formatt + TruColor, the blacks are still brown, which is not something I'm excited about. What do the scopes say?
Schneider Platinum 1.5 IRND / White Balance: Meter
No Filtration / White Balance: Meter
Formatt 1.5ND + Rosco TruColor / White Balance: Meter
Yep, the meter balanced image reveals that the Schneider filter pushes towards green, and is starting to head towards cyan, where as the Formatt + TruColor is heading toward red and magenta.
Schneider Platinum 1.5 IRND / White Balance: Post
No Filtration / White Balance: Post
Formatt 1.5ND + Rosco TruColor / White Balance: Post
Once white balanced, the Schneider filter comes close to being neutral, however, it has lost some saturation. The Formatt + TruColor, is almost a dead match for the vector scope of the no filtration. There is a slight push towards yellow and magenta, but not much. The resulting image, however reveals a black that is tainted and is no longer black (IR Pollution)- something I do not find pleasing at all.
1.8 IRND
Last but not least the 1.8 ND. I think it is pretty amazing that we even need this type of ND. It speaks quite highly of the designers of this latest batch of cameras that I feel comfortable shooting at high ISO's like 800. But I digress ... At 1.8ND the differences really stand out. In the metered balanced image the Schneider is still favoring a green / cyan bias, and the Formatt + TruColor a red / magenta bias. While I'm not happy with either result, I prefer the Schneider over the Formatt + TruColor as it feels much closer to black. Once the images are white balanced, neither of them are perfectly neutral. However, again, I prefer the Schneider over the Formatt + TruColor. The Schneider still feels like it has a little bit of blue left in the blacks which I feel can be comfortably handled in the grade. Whereas with the Formatt + TruColor, the blacks have changed to a "bright" brown. That is much more disconcerting to me as that would be a much greater adjustment in the grade. It is not something that I would want to be fighting with in post.
Do the vector scopes reveal anything ...
Schneider Platinum 1.8 IRND / White Balance: Meter
No Filtration / White Balance: Meter
Formatt 1.8ND + Rosco TruColor / White Balance: Meter
And the scopes say ... yep the Schneider has a strong shift towards green / cyan, and the Formatt + TruColor has a strong shift towards red / magenta. Does it really go away with white balancing?
Schneider Platinum 1.8 IRND / White Balance: Post
No Filtration / White Balance: Post
Formatt 1.8ND + Rosco TruColor / White Balance: Post
Again we find the Schneider filter, once white balanced comes away desaturating the image, but not shifting it drastically at all. And the Formatt + TruColor keeps the scope pretty close to the scope of the no filter. On the vector scope the Formatt + TruColor is closer to neutral then the Schneider, and it does not desaturate the image as much. However, the resulting image reveals that the Schneider filters are doing a better job at keeping black its correct color by removing the IR pollution.
Conclusions:
- For those of you who are picky like me, play it safe and use IR filtration whenever using ND.
- For those of you who are less picky, you can feel comfortable not using IR filtration up to 0.9 ND. You WILL need it at 1.2 ND and above.
- No matter what brand of ND's you choose go with ONE brand so that the color shift is the same throughout the range of filtration. You want, and need to keep things consistent.
- White balance AFTER the ND has been put in the camera. This is a MUST on non-raw cameras, and post workflows, and it is HIGHLY recommended even on raw cameras as it will help communicate your intent, as well as make sure that any rushes or dailies that are produced look correct. Always remember and repeat this montra: "Fix it in camera, NOT in post."
- The new Schneider Platinum IRND's are doing a great job at keeping the IR pollution out of the image while producing nice blacks. At the stronger levels, some refinement will still be needed as there is a slight shift towards blue. But it is more then manageable, and preferable to the brown Formatt + TruColor.
- The Formatt + TruColor combo, while it worked on the Red One, doesn't seem to be holding up as well with the Epic. Personally, I wouldn't use it past 0.9 ND, but if you are less picky about your blacks actually being black, then you might find the 1.2 ND & 1.5 ND to be acceptable. The 1.8 ND combo is completely unacceptable to me.
I hope you have found this helpful as you consider ND filtration for your latest generation digital camera. Now I need to do the same with a daylight balanced source ...
Until Next Time - Get Out There And Shoot!
Ryan E. Walters, Cinematographer
Great test, very clear many thanks!!
ReplyDeleteSoren
Glad you found it helpful. :)
DeleteThanks Ryan for your test.
ReplyDeleteIs it correct, like stated in the Schneider website "Higher densities can be achieved by stacking two or more filters." that you can stack, say IRND1,8 and IRND1,2 to get 10 stop less light without "noticeable" color shift?
Thanks.
Patrick