Time-of-flight cameras build a depth map by timing a 940 nm VCSEL pulse against the return, and they do it in sunlight that dwarfs the signal. The narrowband infrared filter in front of the sensor is what lets the weak return through while rejecting the ambient flood, so its selection is a set of trade-offs between centre-wavelength stability, bandwidth, angle shift and blocking. This guide works through those trade-offs for a ToF receiver.
IR Filters in Time-of-Flight Sensing
A ToF receiver images the scene through an infrared bandpass centred on the VCSEL wavelength, usually 940 nm. The sensor sees the modulated return plus a large sunlight background, and depth accuracy depends on the filter passing the return while cutting the background across the visible and near-infrared. The filter is therefore specified by its pass band and by its out-of-band blocking at least as much as by its peak transmission, because in daylight the blocking decides the signal-to-noise floor.
How it compares with the alternatives
Against a broadband IR window, a narrowband filter dramatically improves sunlight rejection at the cost of peak transmission and of tighter wavelength control. Against a software-gated receiver alone, an optical filter removes the ambient light before it saturates the sensor, protecting dynamic range that software cannot recover. A dielectric stack gives the sharpest edge and deepest blocking but is angle sensitive; a less sharp filter is more forgiving off axis but lets more background through. The choice is a balance, not a maximum on any one axis.
Coating and deposition considerations
The filter is a dielectric stack whose layer periods set the pass at 940 nm; the same stack defines the blocking shoulders, so coating uniformity across the aperture is what keeps the pass centred everywhere on the part. A hard oxide top layer resists cleaning and humidity, which matters for a sensor exposed to the environment. Because the VCSEL and the filter are made separately, the coating must be centred on the VCSEL's actual wavelength including its temperature drift, or the system loses transmission exactly when it heats up.
What drives cost and lead time
Cost rises with how deep the blocking must go and how tight the bandwidth is, because both need more layers and a cleaner run. A filter blocked to optical density four is cheaper than one blocked to six, and a 10 nm band is cheaper than a 5 nm band. Lead time is set by the coating run and by any custom aperture or pattern, not by the substrate. The practical lever is specifying the blocking depth the scene actually needs rather than the deepest the shop can make, and letting one coating serve the receiver across its temperature range.
Substrate and material selection
The substrate is usually a near-IR-transmitting glass or a thin wafer, chosen for flatness and for low autofluorescence rather than for colour. For a close, wide-cone receiver the substrate and any cement layer should add minimal angle-dependent shift, and a thinner part reduces that shift. Where the filter also must survive outdoor humidity and cleaning, a hard-coated glass substrate beats a fragile thin-film on plastic. Pick the substrate for the environment and the cone, then let the coating do the spectral work.
Requirements specific to Time-of-Flight Sensing
ToF adds a sunlight-rejection axis that indoor IR sensing ignores. The filter must hold its pass at 940 nm across the VCSEL's temperature drift and across the receiver cone, while blocking ambient light deeply enough that the return is not drowned. Specify the centre-wavelength shift with temperature and angle, the blocking depth the daylight scene needs, and a hard coat for outdoor life, and centre the coating on the actual VCSEL wavelength rather than a nominal one.
- Pass centred on the VCSEL wavelength including its temperature drift
- Blocking depth set by the daylight scene, not a maximum
- Centre-wavelength shift with angle stated across the receiver cone
- Hard coat for outdoor humidity and cleaning
Framework references: filter specifications per ISO 10110 where applicable; the VCSEL channel near 940 nm follows the illumination design. Confirm the filter centre against the VCSEL's measured wavelength across its operating temperature rather than a nominal value.
Selection data at a glance
| Parameter | Typical value | Notes |
|---|---|---|
| Pass band | 940 nm, 8-15 nm | Matches VCSEL |
| Peak transmission | ≥ 90 % typical | On band centre |
| Blocking | OD 4-6 out of band | Sunlight floor |
| Angle shift | stated across cone | Not on axis only |
| Temp shift | centred over range | VCSEL drift |
| Substrate | NIR glass or wafer | Flat, low fluoresce |
Frequently asked questions
Can a broadband IR window replace a narrowband ToF filter?
Only indoors. In sunlight a broadband window lets the ambient flood swamp the weak VCSEL return, so depth accuracy collapses. The narrowband filter's job is to cut the background before it reaches the sensor, which a wide window cannot do.
Why centre the filter on the VCSEL temperature drift?
Because the VCSEL wavelength shifts as the device warms, and a filter centred only on the cold wavelength loses transmission exactly when the system is hot and the signal is already weak. The coating must be centred on the VCSEL's actual wavelength across its operating range.
Is deeper blocking always better?
Not cost-wise. Blocking to optical density six needs more layers and a cleaner run than blocking to four, and the scene may only need four. Specify the depth the daylight environment demands rather than the deepest the shop can make.
Need a quotation for ir filters?
Send your drawing, target specification or coating requirement and an optical engineer will reply within 24 hours with a price and a lead time.
Request a Quote Email an engineer