ND Filter Quality: What Actually Matters and How We Test It

There is a lot of interesting marketing in the ND filter world.

Optical glass. Cinema glass. HD glass. Nano coatings. 10 layers. 28 layers. Ultra-low dispersion. Professional grade.

Some of those words describe real things. The problem is that they often tell you surprisingly little about whether the filter in front of your camera is actually any good.

An ND filter has a fairly simple job. It should reduce the amount of light entering the camera by a known amount, without changing the colour, adding reflections, degrading the image, getting in the way of the lens, or breaking the first time the camera gets knocked around.

There are a lot of questions we can ask about this:

• Is the filter actually neutral?
• Is an ND16 really an ND16?
• How much reflection and flare does it add?
• What are the coatings actually doing?
• What is the filter made from?
• Is it strong enough for the camera it is going onto?
• Does the frame cause vignetting or interfere with other equipment?

We make ND filters, so we're not exactly a neutral party here. That's one reason we think measurements are more useful than marketing claims.

Where something can be measured objectively, that's what we try to do.

1. A neutral density filter should actually be neutral

The "neutral" part of neutral density is important.

A perfect ND filter would reduce every visible wavelength by exactly the same amount. Red, green, blue and everything in between would all be attenuated equally.

Real filters are never absolutely perfect, but a good one should get very close.

If a filter transmits more red than blue, for example, the image gets warmer. If it suppresses green disproportionately, you can end up with a magenta cast.

You can correct quite a lot of colour cast with white balance or in post, but it is much better if the filter does not create it in the first place. This matters even more if you're switching between ND8, ND16 and ND32 during a shoot. You don't want each filter giving your footage a slightly different colour that you then have to match later.

A good ND filter should work like a dimmer switch. Turn the light down, but don't change the colour of the room. If switching from ND8 to ND16 also makes the image warmer, greener or more magenta, the filter is doing something it shouldn't be doing.

How we test it

We use a scanning spectrophotometer.

It sounds more intimidating than it really is. The machine shines light through the filter and measures how much gets through at different wavelengths.

Human vision covers roughly 400 to 700 nm for practical purposes. Violet and blue are on the left of the graph, green is in the middle, and red is on the right.

Instead of looking at the filter and saying "that seems pretty neutral", we get a graph.

The flatter the transmission curve is across the visible spectrum, the more evenly the filter is treating different colours.

Here is an example from some of our earlier testing. Both ND16 filters were measured on the same spectrophotometer under the same conditions.

The difference is pretty obvious. The PolarPro filter transmitted roughly three times as much light at the red end of the spectrum as it did at the violet end. That produces a noticeable warm colour cast.

The Camera Butter filter stays much flatter across the visible spectrum, which means the different colours are being reduced much more evenly.

Our current filters have even improved since that test.  Here are measurements from an ND8, ND16, ND32 and ND64 from our current range.

The curves sit at different heights because they're supposed to: An ND64 should transmit much less light than an ND8. What we're looking at for colour neutrality is the shape of each line as it travels from violet through red. Ideally, it stays as level as possible.

That is one of the reasons I'm skeptical of terms like "cinema grade" when there is no measurement behind them. Colour neutrality is quite easy to measure.

We also supply existing filter formats for machine vision and industrial imaging applications, where measured density and spectral performance can be particularly important.

Where cost comes into it

Tighter spectral tolerances cost money. Producing to a loose specification is cheaper than specifying tighter tolerances, checking production samples, rejecting material that falls outside them, and continuing to verify later production runs. The difference is almost impossible to see in a product photograph, but it can be very obvious in the measurements.

2. Is an ND16 actually an ND16?

Colour neutrality is only half of the job.

The filter also needs to be the density printed on the frame.

The ND numbers aren't arbitrary. They tell you roughly what fraction of the original light should get through:

ND8 should transmit 1/8 of the light, or 12.5%.

ND16 should transmit 1/16, or 6.25%.

ND32 should transmit 1/32, or 3.125%.

ND64 should transmit 1/64, or about 1.56%.

In photographic terms, those correspond to 3, 4, 5 and 6 stops.

Each additional stop cuts the amount of transmitted light in half.

So if an ND16 is actually letting through enough light to behave like an ND10, it still makes the image darker, but the number printed on the filter isn't telling you the truth.

That becomes important when you're deliberately choosing a shutter speed.

If your exposure calculation says you need four stops of reduction, you want the four-stop filter to actually provide four stops.

It also matters when you switch filters. Moving from ND16 to ND32 should give you approximately one additional stop. If the densities are wandering around, the numbers printed on the filters become suggestions rather than useful measurements.

Getting the ND value right starts during the coating process. We use 16 ND layers on each side of the glass, 32 in total, and build the attenuation up in carefully controlled steps rather than relying on one thick layer to hit the target.

The coating material, number of layers, thickness of each layer, absorption-layer thickness and coating time are all adjusted together to reach the required density. That is why a layer count by itself does not tell you much. The number only matters as part of a controlled coating recipe.

The spectrophotometer tells us whether that recipe actually landed where it was supposed to.

How we test

The spectrophotometer also tells us how much light is actually passing through the filter.

Using 550.5 nm, near the middle of the visible spectrum, as a reference point, these were the results from the four Camera Butter filters we tested:

Filter	Rated	Ideal transmission	Measured transmission	Measured density	Error
ND8	3 stops	12.50%	11.51%	3.12 stops	+0.12
ND16	4 stops	6.25%	6.50%	3.94 stops	-0.06
ND32	5 stops	3.13%	2.97%	5.08 stops	+0.08
ND64	6 stops	1.56%	1.61%	5.96 stops	-0.04

The worst of the four was only 0.12 stop away from its rated density.

In practical terms, if the frame says ND32, it's behaving very close to an ND32.

That isn't something you can simply assume with every filter.

One independent drone-filter test found a filter marked ND16 that actually behaved more like roughly an ND10. A later comparison of 54 filters from several manufacturers found significant density errors across multiple brands, with only one product line consistently matching its stated ratings.

A separate test of fifteen 10-stop photographic ND filters found that accuracy varied between products as well. The author considered filters within roughly a third of a stop acceptable, while several fell farther from their nominal value.

Against that background, having all four of these Camera Butter samples within about one tenth of a stop is a result we are proud of:  (dataset)

We also do camera-based checks of colour and density. The spectrophotometer tells us what is happening to the light itself; putting the filter on a camera gives us a useful real-world cross-check of what the imaging system actually sees.

Where cost comes into it

Again, tolerance is expensive.

The cheapest option is to accept a wide density tolerance, price the filter low, and sell as many as possible.

Testing filters does not make a bad filter better. It tells you which filters don't meet the specification so you don't sell them.

That distinction matters.

3. How the ND is made

There are several ways to make a filter reduce light.

One common approach is absorptive ND glass, such as SCHOTT's NG family. The material itself absorbs light, with the final density determined largely by the composition and thickness of the glass.

Another approach starts with a clear substrate and creates the ND attenuation with coatings.

We used absorptive SCHOTT glass in some of our earlier filters, partly because it was considerably less expensive. It can be a perfectly legitimate way to make an ND filter, but we found some important tradeoffs. The density was less consistent than we wanted, and some filters had fairly obvious colour casts.

That isn't necessarily a manufacturing defect. Absorptive ND glass itself can have uneven transmission across the visible spectrum. The example below shows several commercially available absorptive ND glass formulations, all at the same 2 mm thickness. Although they are intended for neutral-density applications, their transmission changes substantially with wavelength.

With our current filters, we separate the two jobs. We choose the glass for its physical properties, then create the ND attenuation with a carefully controlled coating. That lets us use Gorilla Glass for durability while tuning the optical performance independently.

It costs more. We use 16 ND layers on each side of the glass, 32 in total, with the coating material, layer thicknesses, absorption layer and coating time adjusted together to reach the target. It also requires tighter process control and measurement of the finished filter.

The advantage is control. Our current ND8, ND16, ND32 and ND64 samples were all within about 0.12 stop of their rated density, while maintaining much flatter transmission across the visible spectrum than the absorptive filters we used previously.

A less expensive absorptive filter is not automatically a bad filter, and a coated filter is not automatically a good one. The important question is still the same:

What does the finished filter actually do to the light?

4. "28-layer coating" doesn't tell you very much

This is probably my favourite specification in filter marketing.

You'll regularly see things like:

16-layer coating
24-layer coating
28-layer nano coating

The bigger number is clearly supposed to sound better.

Multilayer optical coatings are real, and the number of layers can matter. But the number by itself tells you surprisingly little.

For one thing, what are those layers actually doing?

Are they ND attenuation layers? Anti-reflection layers? Hard coatings? Hydrophobic or oleophobic layers? Are they on one side of the glass or both? Is “28 layers” 28 on each side, or 14 + 14?

Without that information, the number is almost meaningless.

Performance also depends on what the layers are made from, their exact thicknesses and refractive indices, how accurately they are deposited, the wavelengths they are designed for, and how tightly the process is controlled.

Saying a filter has 28 coating layers is a bit like telling you a car engine contains 600 parts.

Maybe it does... But you still don't know if it is a good engine.

What we actually care about

For an action-camera ND filter, we care about:

• colour neutrality
• accurate attenuation
• low reflection
• resistance to flare and ghosting
• water and oil repellency
• durability
• adhesion to the glass

Those are things you can actually see or measure.

“28 layers” describes how something may have been made. It does not tell you how well it works.

Where cost comes into it

There are inexpensive ways to coat glass and expensive ways to coat glass.

The coating materials, deposition accuracy, layer thickness, temperature, process control and inspection all affect the finished result.

Our coatings cost considerably more than the less expensive options available to us. We use them because they give us better control over the optical result and durability.

And this comes back to the same theme as the rest of this article:

If somebody tells you how many coating layers are on a filter but gives you no information about transmission, neutrality, reflectivity or durability, they haven't really told you much about the filter.

5. The inside surface and filter design matter too

This is something almost nobody thinks about until they have owned a twist-on action-camera filter.

Glass has two sides. The outside gets rain, mud, spray and fingerprints, so everybody expects that side to have good coatings. The camera-facing side is easier to ignore, even though it sits immediately in front of another optical surface.

Every uncoated air-to-glass boundary reflects some light. For ordinary glass, roughly four percent can be reflected at each uncoated surface. Anti-reflection coatings reduce those reflections and the flare, ghosting and loss of contrast they can cause.

Put two pieces of glass close together and you have also created more opportunities for light to bounce between them. This is one reason we generally prefer filters that replace the existing lens cover or mount directly in its place, rather than slip-on filters that add another piece of glass in front of it. Fewer unnecessary optical surfaces is usually a better starting point.

On many of our filters, we coat the camera-facing surface as well. That helps reduce reflections between the filter and the camera, particularly with a bright light source in or near the frame.

It also solves a much less glamorous problem: fingerprints.

Anyone who has touched the inside of a GoPro twist-on filter knows how difficult it can be to clean the bottom of a tiny recessed filter. An easy-clean coating on that surface makes a surprisingly big difference.

It is one of those features that isn't very exciting on a specification sheet and becomes quite exciting the first time you have to clean one.

Where cost comes into it

Coating both surfaces costs more than coating one, and the inside coating is almost invisible when someone is comparing filters online.

That makes it an easy place to save money, especially because coatings make up a large part of the cost of an ND filter.

We would rather spend the money there than add another number to the front of the box.

6. What kind of glass is it?

"Optical glass" is another term that gets used as though it were a quality grade, but it isn't very helpful by itself.

There are many types and grades of optical glass, with very different physical and optical properties.  Conventional optical glass can make an excellent ND filter, although some types are relatively brittle, particularly when used in thin sections.

For many Camera Butter filters we use Corning Gorilla Glass because we are putting the filter on an action camera.

The camera may end up on a mountain bike, an FPV drone, a race car, a surfboard, a helmet or at the bottom of a backpack full of camera equipment.  Strength is part of the optical design when the optic lives in these places.

Our Black Diamond filters use 2mm Gorilla Glass where the design allows it, and then we apply the optical treatment to that substrate.

That gives us the physical properties we want while the measured transmission tells us whether the finished filter is doing its optical job properly.

The hammer test

We have a video comparing Black Diamond glass with some other action-camera filter glass using a hammer.

It is not a laboratory fracture-strength test, however it is a fairly good demonstration of why we care about the glass.

Where cost comes into it

Gorilla Glass costs us much more than ordinary filter glass.  It is also significantly more difficult to work with and cut.

It would be easier to use something cheaper. We use it because these aren't filters designed to spend their lives in padded cinema cases.

7. Why the glass thickness matters

Thickness is mostly a mechanical decision.

For a flat piece of the same material, bending stiffness increases roughly with the cube of its thickness. In simple terms, doubling the thickness can make the glass about eight times as stiff in bending, not merely twice as stiff.

That doesn't mean a 2 mm filter is eight times harder to break. Glass fracture also depends on things like edge condition, mounting and how it is hit. But thickness can make a surprisingly large difference to how much the filter flexes and how well it survives real-world abuse.

Thicker glass also adds weight, material and cost, so there is a tradeoff.

For an action camera, where filters are small to begin with, we have generally chosen to put more emphasis on durability than on saving a fraction of a gram.

And thicker only helps if the optical performance is still good. That is why we measure the finished filter, not just specify a thicker piece of glass and assume everything worked out.

8. Why aluminum frames matter

A lot of action camera filters use plastic frames. Plastic has some obvious advantages. It is inexpensive, light, and can be moulded very efficiently in large quantities.

We use anodized aircraft aluminum on many of our filters because it gives us a rigid frame without needing much material.

That matters in several ways:

The frame protects the edge of the glass, which is one of the most vulnerable parts of any piece of glass.

It also keeps the mounting features rigid. Tabs, twist-lock features and narrow sections of the frame are less likely to flex or deform.

And because aluminum is stiff, we can keep the frame relatively compact.

That last part becomes especially obvious in FPV drones:

9. FPV is an extreme example of why mechanical design matters

Camera Butter actually got its start designing filters for FPV racing drones.

It is a fairly unforgiving place to learn how to make a camera filter:

FPV drones can exceed 100 km/h and regularly hit trees, rocks, concrete and the ground. We were designing filters for cameras mounted to machines that were expected to crash, then keep flying.

That environment shaped the way we design filters.

We learned very early that good optical performance is only half the job. The glass has to survive. The frame has to protect its edges. The filter has to stay attached. And everything still has to fit into an extremely small space without causing vignetting.

A DJI O4 Pro camera (a common FPV drone camera), for example, may sit inside a carbon-fibre or molded frame with only a few millimetres of clearance. A filter can have excellent glass and beautiful coatings and still be useless if its frame is one millimetre too wide.

Most action cameras will never experience anything close to what an FPV filter goes through. But designing for that environment gave us a very useful starting point when we began making filters for GoPro, DJI Action and other cameras.

We learned how to make filters strong because our first customers were crashing them into things.

There is some independent evidence that the approach works.

A comparison by skkyFPV tested eight DJI O4 Pro filter systems from DJI, iFlight, Camera Butter, BetaFPV, GEPRC, Flywoo and Freewell. They compared ND options, fit, vignetting, flare, weight and case usability.

Camera Butter received the highest overall score in their comparison, including maximum scores for minimal vignetting and minimal flare.

We did not sponsor the test or supply the filters - they purchased the filters themselves, and we did not know the comparison was being done until after the testing was finished.

That is exactly why I think the result is useful.

Watch the skkyFPV comparison
See the full test results (PDF)

10. Vignetting

Action camera lenses have extremely wide fields of view, which means the filter frame has to stay clear of the optical path right into the corners. A frame can look perfectly reasonable on the camera and still intrude into the image. That's vignetting.

Conventional camera lenses usually give designers considerably more physical space around the optical path. Action cameras give us almost none.

It is a little like trying to stand beside a wide-angle group photo. You can be surprisingly far off to the side and still end up in the picture.

What we check

When we develop a filter, we check it with the relevant field-of-view modes and camera configuration.

We also look at the physical things people actually put around the camera.

That can include:

• waterproof or dive housings
• cages
• protective mounts
• FPV frames
• other common accessories

Fit itself usually isn't an interesting problem anymore. If a GoPro filter doesn't fit a GoPro, something has gone very wrong.

Clearance, vignetting and accessory compatibility are where the mechanical design needs attention.

Where cost comes into it

Making something smaller while keeping it strong is often harder than simply making it bigger. That is one of the reasons frame material, machining and tolerances matter.

The customer mostly sees a black ring around a piece of glass, but there can be quite a lot of engineering, time, testing and attention to detail hiding in that black ring.

11. Testing is part of the product

There is one part of making an ND filter that never appears in the product photograph: checking whether the finished filter actually does what it says.

Specifications can look convincing on paper. A filter may be described as optical glass, four stops, colour neutral, with 28 coating layers. All of those things may be true, but they still don't tell you how the finished filter actually performs.

If we are selling something whose entire job is to control light, I want to know what it is doing to the light. That is why we test finished filters rather than relying on specifications alone. We verify the optical performance with measurements, then check the things instruments cannot tell us, such as vignetting, reflections, accessory compatibility and real-world behaviour on the camera.

You'll find plenty of filters advertised with terms like "cinema grade", "optical glass", "perfect colour neutrality" and increasingly large coating-layer counts. What is much harder to find is a transmission curve, a measured ND value or even a stated tolerance.

The materials and coating technology matter, but they are inputs. Measurement tells you what the finished filter actually does.

You don't need a spectrophotometer to sell ND filters. You need one if you want to know exactly what you're selling.

12. Why some filters cost more

Two ND filters can look almost identical in a photograph: a black frame, a dark piece of glass and a logo. The manufacturing cost behind them can still be very different.

Some of that difference comes from obvious things such as the glass and frame material. Some comes from things you can't see at all, including tighter manufacturing tolerances, more complex coatings, treating both sides of the glass, inspection and testing.

Part of the filter	Lower-cost approach	What we prioritize
ND attenuation	Use nominal or generic material specifications	Measured density and colour neutrality
Coatings	Basic treatment or emphasize layer count	Finished optical performance and durability
Camera-facing surface	Leave untreated	Coat where reflections and cleaning matter
Glass	Conventional lower-cost substrate	Gorilla Glass where durability matters
Glass thickness	Use thinner material	More thickness where strength justifies it
Frame	Molded plastic	Aluminum where rigidity and protection matter
Dimensions	Optimize primarily for manufacturing simplicity	Keep the filter out of the image and compatible with real equipment
QC	Rely on nominal specifications	Measure finished filters

None of these decisions guarantees quality by itself. Aluminum doesn't automatically make a good filter, expensive glass can still have a poor coating, and a perfectly neutral filter can still be the wrong density.

What matters is how all of those decisions work together in the finished product. That is why a specification such as "28-layer optical coating" is much less useful to me than an actual transmission curve.

So what makes a good ND filter?

For us, the list is fairly straightforward. A good ND filter should be neutral, reasonably close to its rated density, introduce as little flare and reflection as possible, stay out of the image, and work properly with the camera and accessories it was designed for.

It should also be practical to use and, because these are action cameras, durable enough to survive the kind of treatment action cameras actually receive.

Some of those qualities can be measured with a spectrophotometer. Others require cameras, test charts, bright lights and real-world use.

And occasionally a hammer is useful too.

If you're not sure which ND strength or filter set makes sense for your camera, use our ND Filter Chooser to narrow it down.