IR Cut Filters and Why They Exist

Generic engineering material, applicable to any module from any supplier.

For a separate reference on team workflow and work records, see here.

A silicon image sensor responds to light well beyond what a human eye sees, into the near infrared. Left unfiltered, that response contaminates every colour in the image — which is why a filter sits in front of almost every camera intended for visible-light work.

What happens without one

Colours shift. Infrared reaches all three colour channels of a Bayer sensor roughly equally, which desaturates and washes the image toward grey.

Dark fabrics photograph pale. Many black dyes are transparent in the near infrared, so black clothing and printed black surfaces can appear light grey or brown.

Further background is available from Electronics Weekly.

Vegetation goes bright. Chlorophyll reflects near infrared strongly, which is the basis of infrared photography and is a defect where the task is colour assessment.

And focus can shift, since a lens does not necessarily bring infrared and visible wavelengths to the same plane.

None of this is subtle, and all of it is invisible if the module is only ever tested under illumination with little infrared content.

The three options and what each is for

An infrared cut filter. Blocks the near infrared and passes visible light. The default for anything doing colour work under daylight or incandescent sources.

A dual-band filter. Passes visible light and a narrow band in the near infrared, typically around a specific illuminator wavelength. For systems that work in visible light by day and switch to infrared illumination in darkness — the filter admits the illuminator and rejects the rest.

No filter. For deliberate near-infrared work, where the infrared response is the point: penetrating certain materials, seeing under infrared illumination, or imaging phenomena that only appear outside the visible range.

The three are physically different parts and the choice is made at order time, which is why module families carry it as a position in the variant code.

Choosing between them

Ask what the illumination will be.

Daylight or white LED, colour matters: infrared cut.

Covert or night operation with an infrared illuminator: dual band, matched to the illuminator wavelength.

Deliberate infrared imaging: no filter, and the lens should be one that focuses acceptably at the working wavelength.

And where the answer is "both, at different times", dual band is the compromise that exists precisely for it — with the trade that visible-light colour accuracy is slightly worse than with a full cut filter.

The mistake worth avoiding

Testing under fluorescent or white LED light and deploying under daylight or incandescent.

Fluorescent and many white LEDs emit little near infrared. Incandescent lamps and sunlight emit a great deal.

Which means an unfiltered module can look entirely acceptable on a bench and fail outdoors — and the failure is a colour shift that looks like a white balance problem and is not.

Test under the actual illumination, which for this particular question is the only test that means anything.

Where the filter sits

Either as a coating on a glass plate in the optical path, or as a coating on the lens itself.

Which matters for two reasons.

A separate filter can in principle be removed — and doing so changes the optical path length slightly, which shifts focus and is not simply a matter of taking a part out.

And a filter is a surface. It attracts dust and fingerprints like any other, and it sits close enough to the sensor that contamination on it is visible in the image.

Switchable filters

Some cameras carry a mechanically switched filter that moves out of the path at night.

It solves the day-night problem properly — full colour accuracy by day, full infrared sensitivity at night — at the cost of a moving part, which is a reliability and vibration consideration in industrial use.

Fixed dual-band is the alternative and is the usual choice where a moving mechanism is unwelcome, which in most industrial mountings it is.

Wavelength matching

Where a dual-band filter is used with an illuminator, the two must agree.

A filter passing a band around one wavelength and an illuminator emitting at another gives a system that sees nothing at night, and the mismatch is easy to create when the two are sourced separately.

Check both figures against each other before ordering either, and confirm the illuminator's actual peak rather than its nominal description.

Testing whether a filter is present

Point a remote control at the camera and press a button.

Most remote controls emit near infrared, invisible to the eye. A camera with no filter shows a bright flashing point; one with a cut filter shows nothing or almost nothing.

Thirty seconds, and it confirms which variant is actually in front of you — which matters when several are on a bench and the labelling has come off.

A dual-band filter shows the emitter only if the remote's wavelength falls inside the passband, which makes the test informative in a different way.

What the filter does not fix

Illumination that is wrong in the visible range.

A filter removes infrared contamination and nothing else — it does not correct a colour cast from a light source, and it does not improve contrast between two similarly coloured things.

Both of those are illumination questions, and reaching for a different filter to solve them is a common and unproductive move.

In one line

Match the filter to the illumination, and test under the illumination rather than under whatever is above the bench.

The short version

• Silicon sensors respond well into the near infrared, and unfiltered that contaminates every colour channel
• Without a filter: colours desaturate, black fabrics photograph pale, vegetation goes bright, and focus can shift
• Three options: infrared cut for visible colour work, dual band for systems using an infrared illuminator at night, none for deliberate infrared imaging
• The three are physically different parts, chosen at order time, which is why they appear as a position in a variant code
• Match the choice to the illumination rather than to a default
• Fluorescent and white LED light contain little infrared, so bench testing under them hides the problem entirely