Short-wave infrared, or SWIR, LED technology describes semiconductor light sources emitting between roughly 970nm and 1700nm. The band is invisible to the human eye, yet materials interact with these wavelengths in distinctive and repeatable ways. That interaction is what turns SWIR illumination into a working tool for machine vision, where visible light often cannot separate the objects, defects or moisture a system must detect.
What the SWIR Band Covers
SWIR sits just beyond the red end of the visible spectrum. Human eyes stop responding near 780nm, and standard silicon camera sensors lose sensitivity at roughly 1100nm, so most SWIR imaging relies on indium gallium arsenide (InGaAs) detectors rated for the 900-1700nm region. This pairing — an LED source and an InGaAs camera that share the same spectral window — is the foundation of every SWIR inspection setup.
Because SWIR light is invisible, photometric units such as lumens do not apply. Output is specified radiometrically, in watts. Every term in a SWIR datasheet answers a radiometric question: how much optical power, at which wavelength, spread over what bandwidth, under which operating conditions.
The Core Terminology
| Term | Definition | Why It Matters |
|---|---|---|
| Peak wavelength | The wavelength at the center of the emission spectrum | Must match the material signature and the camera and filter window |
| Spectral half width | Emission width measured at half of peak intensity | A narrow band keeps contrast high when paired with filters |
| Radiant power | Optical output expressed in watts at a stated current | The honest measure of output when lumens are meaningless |
| Operating temperature | Junction temperature range over which data is valid | Parameters shift with heat; validity ranges prevent misuse |
| Radiometric units | Watts and watts per steradian, not lumens | The correct language for invisible radiation |
These parameters are not independent. Peak wavelength, spectral shape and radiant power all shift with drive current and junction temperature, so a datasheet value is only meaningful when stated together with its test current and test temperature.
Why Materials Behave Differently at SWIR Wavelengths
Water absorbs strongly in several SWIR regions — around 1450nm, for example — which makes moisture gradients visible to a system even when a surface looks uniform in visible color. Certain plastics show distinctive spectral signatures in the band, so automated separators can distinguish polymer types that human eyes cannot tell apart. Semiconductor surfaces reflect predictably according to doping and layer structure, and silicon wafers with thin films return measurable, repeatable signals.
These properties explain the main application areas. In food sorting, SWIR illumination reveals moisture gradients, foreign material and internal defects that appear identical under visible light. In plastic recycling, different polymer families absorb differently, so sorting by spectral response becomes reliable at line speed. In semiconductor wafer inspection, the known reflectance of silicon and thin-film layers exposes subsurface defects and thickness variations. In each case the inspection decision depends on spectral contrast, which is why wavelength selection comes before power selection.
Specifying the Whole Chain
SWIR is a system question rather than a single-component question. A working station combines the LED source with an InGaAs camera, spectral bandpass filters that reject unwanted wavelengths, and calibrated optics that deliver uniform illumination across the field of view. Source, camera, filter and optics are specified as one chain, and a mismatch at any link destroys the inspection result.
Thermal management deserves particular attention. A source that drifts in wavelength as its junction heats pushes the emission outside the filter passband, and contrast degrades even though the LED still lights up. Verified peak wavelength, spectral bandwidth, radiant power and temperature behavior — stated in radiometric units and measured under real operating conditions — are the baseline data any integrator needs before committing to a design.
Frequently Asked Questions
Why are SWIR outputs rated in watts instead of lumens?
Lumens are weighted by the eye’s sensitivity curve, which falls to zero beyond the visible range. SWIR radiation is invisible, so radiant flux in watts is the only meaningful output measure.
Can a standard silicon camera image SWIR illumination?
Only marginally. Silicon sensitivity ends near 1100nm, so serious SWIR work uses InGaAs detectors matched to the source’s spectral band.
Why does junction temperature matter so much in SWIR designs?
Peak wavelength shifts as the junction heats. A small drift can move the emission outside the filter passband and reduce the contrast the camera depends on.
