Quick Look: Variable ND Filters
Variable NDs replace a boxful of filters, and allow smooth exposure changes... for a price.
By Adam Wilt | February 27, 2012
Formatt Hitech Multistop, Kenko Variable NDX, and Tiffen Variable ND filters on a Sony 24-70mmm f/2.8 Zeiss zoom.
The Sony NEX-FS100 has no internal neutral density filters, and its telescoping 18-200mm lens doesn't work well with matte boxes. Lens-mounted variable NDs are said to be the ideal solution: a single filter capable of 2 to 8 stops of brightness attenuation, thus replacing several conventional NDs.
Sony sent two variable NDs along with the FS100, and by sheer coincidence I had just ordered one myself, so I had a chance to try all three side by side. I've also explored one of them further on a PMW-EX1 and a DMC-GH2 EVIL (Electronic Viewfinder Interchangeable Lens) camera. Variable NDs are indeed useful, and are arguably the most important filter in your toolkit when shooting with electronically-controlled still-camera lenses—as long as you understand their peculiarities.
What They Are
Variable NDs consist of two polarizers, one atop the other. The front polarizer rotates, changing the alignment between them, typically going from a "min" position to a "max" position about 80-85 degrees away.
When their polarization axes are parallel at the "min" position, the filter cuts down light by about 1.3 stops. As you rotate the front of the filter, light transmission decreases, down to about 6-8 stops of additional light loss at the "max" position. In log-density terms, if we treat the "min" position as unfiltered (ND 0), that gives you a range all the way up to around ND 1.8 to ND 2.4.
The filter factor, or exposure change, isn't linear with angle of rotation. It takes about 45 degrees of rotation away from "min" before incoming light is reduced one stop, with additional stops of reduction coming with fewer and fewer further degrees of rotation:
Measured spacing of 1-stop adjustments in a quarter-turn of the Formatt Multistop. Tiffen and Kenko filters behave similarly.
Go past the labeled maximum position, all the way to 90 degrees of rotation, and the polarizer axes are orthogonal—completely crossed. In an ideal world, this configuration wouldn't let any light through: you'd have ND ?. In the real world, though, light squeezes through crossed polarizers in an interesting and angle-dependent way, as as you roll the filter past the 90-degree point you'll see a "rolling X" or "X bar" (Tiffen's term) artifact, as shown below.
In actual practice, there are of course two minima, 180 degrees apart, and two maxima, each at the midpoints between the minimum points. Some filters, like the Tiffen and Formatt examples, have a single min-max scale on a single quadrant of the filter; others, like the Kenko, quite sensibly label all four quadrants with min-max scales.
Since polarizers are used, the angle of the variable ND affects the rendering of polarized light, such as skies at 90 degrees from the sun; reflections in windows and water; glare bouncing off woodwork, floors, and the like; and any device with an LCD screen. If the frontmost polarizer in the variable ND is a linear polarizer, polarized light will be strongly affected, varying from full-intensity to almost total extinction depending on the rotation of that front polarizer. Some variable NDs used reversed circular polarizers for their frontmost filter; these have a much less dramatic impact on polarized light, showing only a gentle biasing of that light from blue to amber, depending on angle (samples below, but if you have a circular polarizer handy, you can see the effect yourself by looking through the filter "backwards" and rotating it).
Variable NDs vary wildly in price: 82mm variable NDs on B&H run from $50 to $600; on Amazon that same search pulls up filters from $16 to $500!
User reports vary equally wildly about which filters have unacceptable color shifts, too-limited adjustment ranges before "X bars" appear, or soften the image too much. Unfortunately I only had three variable NDs to test side-by-side, and only one to explore further. That's why this is a quick look instead of a review.
Sony sent me a 77mm Tiffen Variable ND ($189 at B&H) and a 77mm Kenko Variable NDX ($450 at B&H). I purchased a Formatt Hitech Multistop ($230 at B&H). I tried them all out on a Sony Alpha 24-70mm f/2.8 Zeiss zoom mounted on an NEX-FS100 LSS camcorder.
I set the camera on its daylight white-balance preset, and used zebras to set my exposure to a consistent point (the two whitest bars on the DSC Labs Chroma DuMonde grayscale chart went stripy; this overexposed the chart slightly, but let me line up an iris/shutter speed combo than gave me the widest ability to halve or double exposures to compensate for the effect of the filters).
For each variable ND, I measured exposure without the filter, then put the filter on at its "min" setting, and measured the exposure change needed to compensate for it. I then opened iris or shutter speed up a stop, and turned the front ring of the variable ND to bring exposure down to my setpoint, repeating this procedure all the way to ten stops of attenuation (ND 3.0).
Filter effects: Filter name vs. stops (and ND #s) of attenuation.
All three filters in their "min" position cut the light by 1.3 stops, and all had the same affect on color, causing a slight warming of the image compared to the no-filter case. And, up until the "X bar" creeps in, all hold the same color rendition very consistently.
All the filters appear to use reversed circular polarizers as their front filters; the screen of the iPad I used to display filter data never went appreciably darker than its surroundings (with a linear polarizer, or a circular polarizer used un-reversed, it would go completely black at the point where the camera's polarizer was at 90 degrees to the screen's polarizer).
The "X bar" appears at different positions in the different tests, but that's an accidental artifact, simply due to how the filters lined up when they were screwed into the lens.
The clip below shows you how the "X bar" got its name, as I roll the front of the filter through the 90-degree point, and also an extreme example of how overall filter alignment—not changing the angles of the two polarizers with respect top each other, but spinning the entire filter on its threads—affects the appearance of any differential darkening.
Varying the ND, and then spinning the entire filter.
The intensity and onset of the X bar, regardless of orientation, did seem to vary slightly between the filters (at least on this setup); the Formatt was usable perhaps a stop beyond the Kenko, which itself went maybe a stop beyond the Tiffen, though the actual point at which you'd draw the line as to their acceptability is up to you.
The filters also had different impacts on tonal-scale rendering and image crispness. The Tiffen was the most neutral; everything from deep shadows to specular highlights came though unscathed.
with (RGB) and without (CMY) Tiffen Variable ND filter.
In these grabs of the Adobe Premiere Pro 'scopes window, I've superimposed the unfiltered scene over the filtered scene, reversing the colors of the 'scopes for the unfiltered scene. While the overall waveform hasn't changed much (the levels are just a bit higher as the magenta-to-green WFM separation shows, but that was due to my imprecision in setting exposure), the color shift is indicated by the slight movement on the vectorscope towards the upper left, and the strong reduction in the blue channel (yellow without filter, blue with filter) in the RGB parade.
The Kenko filter performed very similarly, but there's a slight bit of flare on specular highlights; the reflections off the chrome on the cars blooms a couple of pixels more with the Kenko than without (the Tiffen showed no blooming at all).
WFM, vectorscope, and RGB parade with (RGB) and without (CMY) Formatt Multistop filter.
The Formatt shows a noticeable compression of the tonal scale: shadows are lifted and highlights compressed to get the same overall exposure (though the color changes are almost identical to the other two filters). It's almost as if the Formatt was acting as a contrast-reduction filter as well as an ND.
Next: deeper exploration; the secret weapon; pros / cons / cautions.
Get articles like this in your inbox: Sign Up