1550 nm Narrow Bandpass Filter for LiDAR: How to Specify the Receiver Filter
A 1550 nm narrow bandpass filter for LiDAR (Light Detection and Ranging) sits directly in front of the InGaAs (indium gallium arsenide) detector, and it has exactly one job: discard solar photons while passing the return pulse. What decides whether it works is not the center wavelength printed on the datasheet. It is whether the passband still covers the laser line after you add coating tolerance, the angular cone of your receiver, and thermal drift in both the laser and the filter.
Get that budget wrong and the filter attenuates your signal instead of the background.
Why a 1550 nm LiDAR receiver needs a narrow filter
The reason systems move from 905 nm to 1550 nm is eye safety. Above roughly 1400 nm, the cornea, lens, and aqueous and vitreous humors absorb the radiation so strongly that essentially nothing reaches the retina, so the damage threshold is set by corneal and lens injury rather than retinal injury. That raises the permissible emission substantially and lets a 1550 nm system range farther than a 905 nm system, though it does not make 1550 nm safe at arbitrary power.
Higher transmit power does not automatically give you range, because the detector also sees the sun. In a direct time-of-flight (ToF) receiver, background power scales as:
Pb = Lλ · Δλ · AR · Ω · T
where Lλ is the spectral radiance of the scene, Δλ the optical bandwidth, AR the receive aperture area, Ω the instantaneous field of view solid angle, and T the transmission of the receive path. Aperture and field of view are fixed by your range equation and scan architecture. Bandwidth is the one term the optical designer controls freely.
1550 nm helps here too, but less than people assume. In the ASTM G173 AM1.5 reference spectrum, global tilt spectral irradiance is about 0.27 W·m-2·nm-1 at 1550 nm, against about 0.74 W·m-2·nm-1 at 1000 nm. Roughly a third the background per nanometer, not zero. Narrowing the passband from 30 nm to 3 nm cuts background power by a factor of ten and background shot noise by about a factor of three, with no effect on the signal as long as the laser line stays inside the band.
This applies to direct-detection ToF receivers using an InGaAs APD (avalanche photodiode) or SPAD (single-photon avalanche diode). Coherent FMCW (frequency-modulated continuous wave) receivers reject incoherent background through the mixing process, so their bandwidth requirement is looser.
The physics: why the passband moves when light arrives off-axis
A narrow bandpass interference filter is a thin-film Fabry-Perot cavity. Transmission peaks where the round trip through the spacer layer is an integer number of wavelengths:
mλ = 2 n d cosθ’
Tilt the filter and θ’ grows, cosθ’ shrinks, and the resonance moves to shorter wavelengths. The shift is always toward the blue, never the red. The industry-standard expression uses an effective index neff that lumps the multilayer stack into a single number:
λθ = λ0 √(1 − (sinθ / neff)²) ≈ λ0 − λ0 sin²θ / (2 neff²)
This form is quoted by Semrock, Omega Filters, and other filter houses, with the 50% transmission point as the reference feature; a higher effective index means less angle shift. Because the shift goes as the inverse square of neff, the cavity is normally built from the high-index material, and raising the indices of both the high- and low-index layers reduces the sensitivity. Typical values for oxide stacks fall somewhere around 1.7 to 2.1, and the number is design-specific: ask your supplier for the neff of the actual design rather than assuming one.
Angle shift at 1550 nm
| Angle from normal | Shift at neff = 1.8 | Shift at neff = 2.05 |
|---|---|---|
| 1° | −0.07 nm | −0.06 nm |
| 2° | −0.29 nm | −0.22 nm |
| 5° | −1.8 nm | −1.4 nm |
| 10° | −7.2 nm | −5.6 nm |
| 15° | −16 nm | −12 nm |
Values computed from the expression above for a 1550 nm design. The real neff of your coating decides which column applies.
Two consequences follow. First, an f/2 focusing beam carries a marginal half-angle near 14 degrees, which would blue-shift the edge rays by more than 10 nm and smear the effective passband across the cone. A narrow filter belongs in collimated space, or the design must be centered for the specified cone. Second, s- and p-polarized components see different effective indices at oblique incidence. Oblique or focused illumination always blue-shifts the band, and with randomly polarized light the band also broadens — peak transmission falls and the edges soften, so a filter measured collimated at normal incidence will not reproduce that curve in your system.
What belongs on the drawing
| Parameter | What to state | Why it matters |
|---|---|---|
| CWL (center wavelength) and tolerance | Nominal plus tolerance, referenced to your AOI and cone, not 0° collimated | Coating tolerance is usually the largest single term in the budget |
| FWHM (full width at half maximum) | A maximum, not a target value | Narrower is not automatically better; see the budget below |
| Transmission | Minimum absolute transmission at 1550 nm, in your geometry | Peak transmission at an unspecified wavelength tells you nothing about the laser line |
| OD (optical density) and blocking range | Blocking level plus the wavelength span, e.g. OD ≥ 4 across the detector’s full response | An unblocked window inside the detector band leaks background straight through |
| AOI (angle of incidence) and cone half-angle | Both, explicitly | The coating is centered for these; omitting them means the supplier guesses |
| Thermal coefficient | pm/°C plus the operating range | Automotive ranges are wide; the shift is small but not negligible |
| Surface quality | Scratch-dig per MIL-PRF-13830B, e.g. 60-40 or 40-20 | Scatter into a single-element detector raises the noise floor |
| Transmitted wavefront / surface figure | P-V (peak-to-valley) in waves at a stated test wavelength | Matters when the filter sits in a converging beam feeding a small active area |
| Clear aperture and CWL uniformity | Both, for apertures beyond a few millimeters | Coating thickness varies across a large part, so CWL varies with it |
| Substrate and environmental spec | Material, thickness, and the durability standard you require | Hard sputtered oxide coatings survive humidity and thermal cycling that soft coatings do not |
ISO 10110 is the usual drawing convention for surface and material notation; MIL-PRF-13830B remains the common scratch-dig reference. If the part ships into the EU, state the RoHS requirement explicitly rather than assuming it.
One specification habit is worth adopting. Rather than dictating CWL and FWHM tolerances separately, state an absolute transmission at the wavelength you care about together with a maximum bandwidth — for example, greater than 92% transmission at the laser line and FWHM below a defined value — because additional shape tolerances drive cost quickly. It tells the coating designer what the system actually needs and leaves them room to hit it.
The same discipline applies to the rest of the receive path. Fold mirrors and beam-steering optics carry their own reflectance and wavefront budgets, and a high-reflectivity 1550 nm optical mirror from a supplier such as GIAI Photonics is specified for the same AOI as the filter it feeds, for the same reason.
Building the wavelength budget
This is the calculation that determines your minimum bandwidth. Work out where the laser line can be, where the filter passband can be, and make sure they always overlap. Consider a receiver with the filter in collimated space, residual half-angle 2 degrees, mounting tilt within 1 degree, operating from −40 °C to +85 °C, with an EDFA (erbium-doped fiber amplifier) source seeded by a temperature-stabilized DFB (distributed feedback) diode.
| Contribution | Direction | Example magnitude | Source of the number |
|---|---|---|---|
| Laser line placement over life and temperature | Both | ±0.3 nm | Source datasheet; uncooled diodes are far worse |
| Filter CWL manufacturing tolerance | Both | ±2 nm | Common catalog-grade tolerance; tighter is available at cost |
| Filter thermal shift | Red when hot, blue when cold | ±0.2 nm | 3 pm/°C over a 60 °C excursion from ambient |
| Cone half-angle 2° | Blue only | −0.3 nm | Angle shift equation, neff = 1.8 |
| Mounting tilt 1° | Blue only | −0.07 nm | Angle shift equation |
Worst case, the laser sits at 1550.3 nm while the filter passband has drifted to 1547.4 nm. The separation is about 2.9 nm, so the half-maximum points must be at least that far from the nominal center, implying an FWHM near 6 nm just to keep the line inside the passband at all. To keep it inside the flat top, where transmission is above 90%, you need more than that.
The dominant term is the coating tolerance, not the physics. That is why production 1550 nm LiDAR receive filters are commonly several nanometers wide even though sub-nanometer coatings exist. You have two levers: buy a tighter CWL tolerance, or accept a wider passband and more background. Which is cheaper depends on volume.
Hard-coated filters typically shift somewhere between 2 and 5 pm/°C over normal instrument temperature ranges, depending on substrate material and design. Over an automotive range this is well under a nanometer, so the filter is rarely the thermal problem. The laser usually is. A DBR-stabilized 1550 nm diode can hold about 0.1 nm/°C, roughly five times better than a broad-area device — and 0.1 nm/°C across 125 °C is 12.5 nm, which no narrow filter will tolerate. Either the source is stabilized, or the filter must be wide enough to be useless.
The trade-offs
- Narrower FWHM costs transmission and money. Steeper edges need more cavities and more layers, meaning longer coating runs, tighter thickness control, and lower yield.
- Narrower FWHM demands tighter everything else. A 1 nm filter needs sub-nanometer CWL control, sub-degree mounting, and a stabilized source. The filter is rarely the expensive part of that chain.
- Higher neff reduces angle shift but constrains materials. Raising the effective index means higher-index layers, with their own absorption and film-stress consequences.
- Deeper blocking costs transmission. Going from OD 4 to OD 6 over a wide range usually means more layers or a second blocking component, both of which cut in-band transmission.
- Larger clear aperture makes uniformity a real spec. Coating thickness varies across the substrate, so CWL varies with it, adding a bandwidth term of its own.
Common specification mistakes
Specifying the filter at normal incidence and then using it in a converging beam. The most frequent and most expensive error. The datasheet curve was measured collimated at 0 degrees; your f/2 cone blue-shifts and broadens it. Either move the filter into collimated space or tell the supplier your AOI and cone half-angle so the design is centered for them.
Blocking that does not cover the detector’s response. InGaAs photodiodes respond from roughly 900 nm to 1700 nm. Blocking only 1400 to 1700 nm leaves the 900 to 1400 nm region wide open, where solar irradiance is two to three times higher than at 1550 nm. Specify the blocking range against the detector’s actual responsivity curve.
Treating “average OD” as “minimum OD”. An average blocking number can hide a narrow leak. If your system sees a specific interferer — another LiDAR, a headlamp, an EDFA’s amplified spontaneous emission — call out minimum OD at that wavelength separately.
Centering the filter exactly on the laser wavelength. Because angle tuning can only shift the band toward the blue, specifying a small red offset in the CWL lets the filter be tilted into position during alignment. A filter centered exactly on nominal, delivered on the blue side of tolerance, cannot be recovered.
Ignoring transmitted wavefront in a converging beam. A filter with poor figure sitting near focus in front of a 200 µm SPAD spreads the spot and costs you photons. If the filter cannot go in collimated space, the wavefront spec is not optional.
Frequently asked questions
What FWHM should I specify for a 1550 nm LiDAR receiver?
Run the budget rather than picking a number. Sum the laser line uncertainty, the coating CWL tolerance, the thermal shift, and the angle shift from your cone and mounting tolerance, then choose an FWHM that keeps the laser inside the flat top under worst case. With catalog-grade CWL tolerance this usually lands in the several-nanometer range, not below one nanometer.
Can I put the narrow bandpass filter directly in front of the detector?
Only if the beam there is slow enough. Near focus the cone half-angle is large, and the blue shift scales with the square of the angle, so a fast beam smears the passband badly. If the mechanical layout forces that position, specify the cone half-angle on the drawing so the coating is centered for the resulting average shift, and expect lower effective peak transmission.
How much does temperature move the passband?
For hard-coated interference filters the shift is a few picometers per degree Celsius, so even a 125 degree automotive range moves the band well under a nanometer, toward the red as it warms. The larger thermal term is almost always the laser. An unstabilized diode can move a nanometer or more per ten degrees, which dominates the entire budget.
Does polarization matter if my source is polarized?
At normal incidence, no. At oblique incidence, s- and p-components see different effective indices and the passband splits, which broadens the band and lowers peak transmission for unpolarized light. If your return signal is depolarized by the target, which it usually is, you get the broadened response regardless of what the transmitter emits.
Do FMCW receivers need the same filter?
Usually not as narrow. Coherent detection rejects incoherent background through the mixing process, so background shot noise is far less of a driver. A filter is still worth having to keep broadband light off the detector and prevent saturation, but the bandwidth can typically be relaxed, which recovers transmission and eases the alignment tolerance.
Specifying with a supplier
Send the coating house your AOI, cone half-angle, operating temperature range, laser line uncertainty, and detector response curve alongside the wavelength and bandwidth. GIAI Photonics builds narrow bandpass filters, bandpass filters, and infrared filters to drawing, and those five inputs are what let any supplier center the design on your system rather than on a nominal number.