Before we can look into ND filters, we’ll need to know what kind of ND filter to buy. The first important classification, one which certainly affects the price, is whether or not we need ‘hot mirror’ or IR ND filters.
Every camera has some form of IR (infrared) pollution. Many cameras have IR blocking filters right in front of the sensor to mitigate this (and to protect privacy). Under normal lighting conditions the infrared radiation passes on to the sensor along with visible light, and nobody notices anything.
However, when using ND filters (or even other filters that cut light to a certain extent, like polarizers, color filters, etc.), the visible light is cut, but the IR passes through.
This means, when you shoot in sunlight (or tungsten halogen, practically anything that emits heat and therefore, IR radiation) and use a ‘normal’ ND filter, the colors tend to shift due to the IR radiation. This is IR pollution, and the results are undesirable, as we shall soon see.
The test for IR pollution
The simplest test for this is to shoot outdoors in sunlight with green foliage, and under tungsten halogen lighting. If there is IR leaking the greens become brown, and some form of black clothing becomes brown. You can also use a color chart for this. This is what we’re going to do.
I have three kinds of ND filters, so I test all of them just to make sure it’s not any particular model but the camera itself. I have 4×4 ND filters, starting at 1 stop (ND 0.3) all the way to 10 stops (ND 3.0). I also have variable ND filters and threaded round ND filters. None of them are hot mirrors or IRND filters.
What are IR ND filters? These are filters (they don’t have to be ND, but can also be polarizing or others) that also have an IR blocking coating (or whatever) applied to them, so they serve double duty. Typically, they cut from between 680nm to 700nm (the beginning of IR radiation) and beyond. You can also buy pure IR blocking filters.
What’s the difference between IRND and Hot mirror? The end result is the same thing, both stop IR from reaching the sensor, but they differ in their technology. IRND filters, the most common kind, have dyes that “absorb” IR radiation, similar to color filters that absorb light of a particular color but lets the others through. On the other hand, hot mirrors are mirrors – they reflect IR radiation back completely. For this reason, they have mirror-like surfaces, and many say you shouldn’t stack hot mirror filters. The mirror side must face the scene, and be furthest away from the sensor if you’re stacking filters. It goes without saying that you must know which side of the filter is the mirror side, or if both sides are, etc.
One more thing, every ND filter shifts color to a certain extent. There have been recent tests with TrueND filters that show minimal color shift. I don’t know if it’s true (pun intended) or not, but both Tiffen and Schneider, the industry standards, have slight color shifts across the image. It is important to distinguish between a color shift and IR radiation. The latter only affects certain aspects of the image, while the former applies a tint across the entire image. A tint is easy to correct, but a color shift in certain regions is extremely hard.
So, there is no ‘formula’ to IRND filters. You must test each stop, each filter, each manufacturer, each technology and each sensor combination thoroughly! There is no easy way I’m afraid. Sometimes the results are surprising, especially with cameras that have their own IR blocking filter, such as the Sony A7s.
Variable ND Test
Some info: Variable ND filter, 6 stops total. Shot at f/16. Matrix metering exposed at middle grey in Standard Creative Style, Picture Profiles Off. ISO range from 100 to 4000. Auto White Balance.
Clearly, you don’t need any IR blocking up to 6 stops. That’s where my Var ND ends. However, note that there is a slight shift in the greens. Let’s dig deeper.
4×4 ND Filter Test
Here’s the same scene, but with 4×4 ND filters:
Info: Some of the color shifts are definitely due to the poor quality of the ND filters. They are exclusively resin, and not glass. All were metered with matrix metering, Auto White Balance.
How do we know the color casts are not caused by color shifts and IR radiation? To do this, I must also shoot the filters against a normal background with cool lighting. In this case, fluorescent:
What I’ve noticed is that stacking ND filters actually makes the image ‘bluer’ under fluorescent lighting. You can easily verify this with a color picker. This is one cause for the major ‘tinted’ color shifts, but of course, not IR pollution shifts. That comes next.
We test the same thing with Tungsten Halogen light. In this case, custom white balanced to 2800K:
Notes: The cloth on the left is pure 100% cotton. The cloth on the right is 95% Viscose (Rayon) with 5% Elastane (Spandex). The cloth on the back is synthetic photographic cloth. As you can see, cotton resists IR pollution a whole lot more than Rayon.
Quite clearly, you can see the effects of IR pollution in the Rayon even at 1.2 (4 stops). By 8 stops (1.2+1.2), the cotton and the chart are both affected. Some of this is due to the stacking of filters, as 2.7 (9 stops) is slightly better that 1.2+1.2 (8 stops).
Finally, we want to know – can IR pollution be corrected easily in post production? For this, I import these files into DaVinci Resolve and use the chart correction tool. Here are the results:
The results are surprising, and very important. You cannot easily eliminate IR pollution effects, especially in the blacks.
What do we learn?
If there is no sunlight or any IR radiation in the lighting plain ND filters will work just fine. But we need NDs most under sunlight, don’t we?
Under sunlight or tungsten lighting, there is a noticeable shift. If your talent is wearing synthetic clothing (which a large majority do), then you have no option but to use IRND or hot mirrors starting at 4 stops. If you’re shooting only daylight scenes, I’d say you can ‘get away with’ no IR blocking up to 6 stops, but clearly, the moment blacks become prominent, it starts to show.
You cannot easily correct it in a grading app, even if you shoot charts.
Bottom line, if you’re looking for professional results, you need IRND or hot mirror filters above 4 stops. It will save you a lot of time and agony in post.
This result has important repercussions, as we shall see in the next lesson.