Automotive Night Vision Camera Design: Matching Filters to the Sensor and Light Source

Both emitters and filters can look matched at the parts list when both are marked “850 nm”. Some of this light is still lost in the camera. The emitter output scatters both sides of 850 nm, the detector response is uneven. Confirm the curves, ratify the match.

Overlay the emitter, filter and detector curves on top of each other at the same wavelength scale. Next place the filter into a camera with the production cover, and check for flare or loss of signal in image.

Start by describing the camera and the failure to prevent

Quick decision: Start with the measured band where emitter output and detector response overlap. Check the filter’s blocking outside that band, then look at the actual camera image. Two parts with the same stated center wavelength need not behave alike in the module.

Key takeaways

• Use one wavelength scale for the source, filter, detector and cover traces, including the tails of each curve.
• Put passband, blocking, angle, and physical limits on the drawing.
• Approve the optical sample and assembled camera as separate decisions.

Filter approval should include three separate stages: drawing review, measured sample inspection, and camera-level acceptance. Each stage requires its own acceptance record to prevent a compliant spectral measurement from being mistaken for verified camera performance.

“Night vision” can mean active near-infrared imaging or thermal imaging. Active NIR systems illuminate a scene and record reflected light. Thermal systems measure emitted radiation. For active near-infrared (NIR) camera systems, the filter specification should clearly identify the imaging architecture, illumination source, and detector requirements to prevent incorrect spectral matching.

Where the camera is and why it doesn’t work. A unit facing the front might view a dark road then sunlight on a sign; a cabin unit works at shorter distances with different angles and lighting. Tell the supplier whether the concern is weak return, lost contrast, flare or a known interfering source. “Automotive grade” does not describe a test.

The filter serves two functions: transmit sufficient light from the camera, and block that which does not. Its drawing sets part-level limits; the image trial checks the whole optical path. Include the source, detector, lens, cover and installed angle in that trial.

Collect the detector response, the emitter spectrum at the planned drive setting, a drawing of the filter position and any lens or cover data. Mark what is missing. An estimate is enough to choose a sample to try, but not enough to set a pass limit.

Put the curves on one wavelength axis

Where the emitter, filter and sensor curves overlap, the camera has a chance to collect useful signal. The cover or adhesive can reduce that band further. Reject candidates with poor overlap on paper, then check the survivors in the module.

Read the source and sensor together

Ask for the emitter curve at wherever your camera drive them. In case curves shift due to whatever reason, such as when output shifts across operating temperature, obtain these characteristics and include them in the relevant sample plan or mark the range of values currently still untested. The shoulders will often be right up against the filter edge and it obscures a peak count. A peak number hides the shoulders, which may sit right against the filter edge.

Compare detector response across the proposed passband with the emitter and filter traces. If cover data is missing, check its effect in the module test.

Specify more than center wavelength and FWHM

Write down center wavelength, full width at half maximum (FWHM), both passband edges, tolerances and the least transmission allowed in band. Check the whole measured trace against the drawing. The shoulders can differ even when center and FWHM match.

If transmission is not considering a measured loss, then ask for reflectance and absorbance at the same scan range Let this be the case for 4 conditions with all three curves. The report is almost meaningless if it does not list its scan range, angle and which sample was used. NIST’s spectrophotometry overview describes optical measurement methods; Camera spec needs to express limits that matter for the device.

On the drawing, distinguish measured supplier results from camera-team decisions. Ask the supplier for transmission, reflection and absorption traces and physical inspection results. The camera team sets allowed passband and blocking ranges, sample angle and the image condition for a pass. Otherwise a good optical report can be mistaken for proof that the assembled camera works.

If a filter passes its drawing but contrast falls near a reflective sign, repeat the test with that source on and off. Keep exposure and target position fixed. For a cabin camera, hold the target position steady while comparing the approved and candidate filters.

For example, we achieve the passband limit in a bench trace but clip useful signal at the field edge from an angled camera installation. Use the trace at a specific angle to explore which part determines that, and with respect of what this loss matters for your application based on the camera image. Utilize the angle-specific trace to explore that component, and utilize the camera image to find out if that loss matters for that application.

If a trace doesn’t look like what the renderer results, compare transmission with reflection and absorption before changing camera limit from Transmission. These measures do not substitute the image trial, but rather these measurements indicate where optical energy is being disbursed. Preserve the raw curves with the sample identifier so that a later lot could be compared on relative basis.

Check angles at the installed filter position

An interference filter measured straight on may shift when mounted at an angle. Show the supplier its position and the chief-ray angle or cone, then agree on measurement angles before the sample is scanned.

Keep a normal-incidence scan as a reference, but measure at the installed angle too. Record sample orientation and fixture. If either changes, note it so the difference is not blamed on the coating.

Define blocking around the unwanted light

“High blocking” needs a wavelength range and a limit. Identify what must be rejected—daylight, a display, another IR emitter or an internal reflection—and put the relevant interval on the drawing. Suppliers can then quote against the same requirement.

An OD number needs its test window. Give the minimum required optical density, wavelength interval, scan range, angle and measurement conditions. If transmission is near the instrument floor, the reported value has that limit too. Compare readings only on the same basis.

Optical density (OD) describes blockage on a log scale; a higher OD means less light transmitted. Wavelength interval and measurement floor are needed along with the string number to give it any meaning. NIST’s spectrophotometry work provides context for optical measurement; the camera team still has to choose the blocking limit from its own stray-light test.

The Optica Publishing Group paper on incidence angle and temperature examines how those conditions affect interference-filter spectra. For camera drawing: treat angle as test condition, send geometry for supplier and – keep same angle given to all candidates.

Recreate the suspected interference in the camera with target and exposure fixed. Switch the source on and off, then check flare, contrast, saturation, and exposure. Meeting an OD limit does not by itself establish image performance.

Approve the part and the camera as separate decisions

The spectral scan checks the drawing, not fit or beam clearance. Before assembly, check sample ID, thickness, outline, effective optical aperture, and specified surface condition. Keep these checks with the measured trace.

For the module comparison, change only the filter. Keep illuminator drive, exposure, target position, and warm-up fixed. Test weak-return and bright backgrounds where relevant, and save frames with settings for repeat review.

Use the camera team’s image or detection limit. A sample can pass the drawing and fail in the camera, or the reverse. Tie both results to the sample ID and spectral report.

Carry the approved sample into production inspection

The incoming plan should reproduce the approved measurement. Record passband and blocking limits, scan range, angle, dimensions, lot, and method. Decide how to handle borderline readings in advance.

If the team retains a reference part, store its label, spectral curve, and test conditions together. Ask the supplier to report a change in substrate, coating, production location, or measurement method. A change may call for a new module check even when the part number stays the same.

The mechanical record matters too. An acceptable curve cannot compensate for an aperture smaller than the beam or an outline that does not seat. Check dimensions, surface condition, and effective aperture against the drawing, beside the spectral trace.

Before approval, compare source, filter, sensor, and cover data; test interference; and record dimensions and camera results. Transfer the limits to incoming inspection and link each lot report to the baseline.

Discuss an automotive filter specification with BoDian Optical

The infrared filter range covers narrowband, broadband, long-pass, and short-pass filters. Choose from the camera’s signal and rejection needs. BoDian Optical can customize filter shape and size to a drawing; include the outline and clear aperture in the request.

For sample review, request the measured transmission curve and compare reflection or absorption if the loss remains unclear. Check dimensions, surface condition, and effective optical aperture against the drawing, then confirm the camera result in the assembled system.

Send BoDian Optical the illuminator spectrum, sensor response, filter angles, cover data, and the camera test that will decide pass or fail. Read About us for supplier background or use the contact page to send the drawing and test conditions. Keeping the requested data together helps the sample review start with the actual camera constraints.