SWIR Cameras: Scientific Guide to Short-Wave Infrared Imaging

Short‑wave infrared (SWIR) is a region of the spectrum from roughly 0.9–1.7 µm for standard SWIR and up to about 2.2–2.5 µm for extended SWIR systems. SWIR cameras, including InGaAs SWIR cameras and extended SWIR cameras, detect reflected and emitted photons in this band to reveal contrast from moisture, silicon transparency, and chemical composition that standard machine vision cameras cannot see.

Standard SWIR cameras utilize InGaAs sensors for the 0.9–1.7 µm range. For extended SWIR (eSWIR) up to 2.5 µm, systems like the ZephIR 2.5e utilize advanced T2SL (Type-II Superlattice) detectors, providing higher sensitivity and lower noise at longer wavelengths compared to traditional modified InGaAs. 

Together, these systems underpin demanding applications in semiconductor inspection, additive manufacturing, security, food and agriculture, recycling, and free‑space optics, where being able to see “beyond” visible light is essential for improving yield, throughput, and safety.

Why SWIR is necessary

Short‑wave infrared sits between the near‑infrared (NIR) band used by silicon sensors and the mid‑wave infrared (MWIR) band dominated by thermal emission. In this region, scattering often decreases with wavelength while absorption from water, lipids, and other chromophores increases, creating wavelength windows where penetration is high but contrast is driven by subtle compositional changes.

This physical behavior allows SWIR cameras to see through certain obscurants and highlight moisture or chemical gradients that standard and NIR cameras miss. Application engineers have found that SWIR makes composition, structure, and process conditions directly visible to automation and analytics.

Pembroke’s SWIR solutions, specifically the WiDy and SenS series, feature sensors with Wide Dynamic Range (up to 120dB). This allows for imaging scenes with extreme contrast—such as industrial welding or laser profiling—without pixel saturation, a capability standard linear SWIR cameras often lack.

SWIR vs. NIR cameras

Near‑infrared cameras typically cover roughly 0.7–1.0 µm and are often based on silicon sensors. SWIR cameras extend sensitivity beyond 0.9 µm into bands where water, polymers, and silicon behave very differently. 

NIR cameras are well‑suited for low‑light imaging and simple contrast enhancement, whereas SWIR cameras enable material discrimination, silicon inspection, and moisture‑sensitive imaging that NIR alone cannot provide.

Many facilities use NIR and SWIR cameras together. NIR is used for basic quality checks, while SWIR is ideal for higher value, composition-driven decisions.

SWIR vs thermal (LWIR) cameras

SWIR cameras form images primarily from reflected and weakly emitted light around 0.9–2.5 µm, so their contrast is strongly tied to material composition, surface properties, and illumination. Long-wave infrared (LWIR) thermal cameras, by contrast, operate around 8–14 µm and are dominated by blackbody emission from object temperature. This makes them ideal for thermography but less effective for applications like through‑silicon inspection or moisture‑based sorting, where SWIR is the better choice.

In practice, SWIR imaging behaves more like “visible‑like” imaging with enhanced material contrast, while LWIR is the tool of choice when absolute or relative temperature is the primary signal of interest. Many facilities use LWIR for thermal process monitoring alongside SWIR for composition and structure.

Spectral bands vs. materials and applications

Positioning SWIR alongside broader optical and infrared bands offers a useful look at its engineering applications:

Short‑wave infrared: bands, photons, and materials

Standard SWIR cameras are optimized for the 0.9–1.7 µm window, which aligns with the native bandgap of lattice‑matched InGaAs on InP substrates. Extended SWIR designs shift detector composition and optics to cover roughly 1.7–2.5 µm, capturing additional absorption bands from water, hydrocarbons, and industrial materials through higher dark current and more demanding optics.

Material contrast in SWIR often arises from a balance of absorption peaks and reduced scattering. Water exhibits strong absorption near 970, 1190, 1450, and 1940 nm, while scattering in tissues and other turbid media drops with wavelength, improving contrast at depths where visible imaging is dominated by blur. Silicon becomes semi‑transparent in the SWIR band used by standard InGaAs sensors, enabling inspection of wafers, solar cells, and packaged devices that appear opaque in visible images.

Common industrial materials also behave differently in SWIR compared with VIS or NIR. Many plastics and organics exhibit characteristic absorption “fingerprints” in the 1.0–2.0 µm range, enabling spectral discrimination of polymer types, food constituents, and contaminants for sorting and quality control. Metals typically remain reflective across the VIS to SWIR range, but coatings, oxides, and surface films introduce spectral structure that SWIR cameras can examine for process monitoring.

InGaAs and other SWIR sensor technologies

The bandgap in naturally supports photon detection up to ~1.7 µm when grown on InP. The material is deposited epitaxially, with the indium fraction tuned to set the bandgap and thus the long‑wavelength cutoff of the device.

Extended InGaAs increases the indium content to reduce the bandgap, pushing sensitivity out toward 2.2–2.5 µm. Advances in material growth, passivation, and device design have significantly improved extended InGaAs performance, narrowing the gap with traditional SWIR HgCdTe detectors at comparable cutoffs and operating temperatures.

HgCdTe (MCT) offers excellent tunability from SWIR through LWIR but typically requires deep cooling and complex fabrication. Compared with HgCdTe, InGaAs generally provides lower dark current and higher shunt resistance at ~1.7 µm cutoffs and can operate at or near room temperature with modest TEC cooling, which reduces system cost and complexity for many SWIR applications.

SWIR camera architectures: area scan, line scan, and hyperspectral

Area scan SWIR cameras use two‑dimensional focal plane arrays to capture full frames, making them suitable for general‑purpose imaging, microscopy, surveillance, and many inspection tasks where objects are static or move slowly. In these sensors each pixel converts incident SWIR photons into charge, and the camera reads out frames at video or high‑speed rates through standard machine‑vision interfaces.

Line scan SWIR cameras use a single row of pixels to image a narrow line across the field of view, relying on the motion of either the object or the camera to build a two‑dimensional image over time. This architecture excels in continuous processes like web inspection, conveyor‑based sorting, and reel‑to‑reel manufacturing, where every product passes through a fixed line for 100% inspection at high speed.

Hyperspectral SWIR cameras add a spectral dimension to each spatial pixel, typically using dispersive optics, tunable filters, or on‑sensor filter arrays to form a three‑dimensional data cube. Push broom (line‑scan) hyperspectral systems combine a spatial line with spectral dispersion so that each frame encodes space × wavelength, whereas snapshot systems use specialized filter or interferometer designs to capture multiple bands simultaneously on an area sensor.

Optical design for SWIR: lenses, filters, and coatings

SWIR lenses must use glasses and crystals with high transmission in the 0.9–1.7 µm (or 2.5 µm) region, which often excludes standard visible‑only crown/flint combinations that absorb or scatter strongly in this band. Designs can combine IR‑grade glasses, fused silica, chalcogenides, or other specialty materials, balancing chromatic aberration, thermal behavior, and manufacturability.

Anti‑reflection coatings optimized for SWIR wavelengths are critical because visible‑optimized coatings can exhibit high reflectivity and ghosting when used beyond their design band. Multi‑layer SWIR AR stacks reduce Fresnel losses at sensor windows, lenses, and protective optics, improving throughput and contrast in low‑signal applications.

Spectral filtering is a key part of SWIR system engineering. Longpass filters are often used to define the lower edge of the SWIR band, suppressing residual visible/NIR light, while bandpass filters isolate specific absorption features—such as water peaks around 1.45 µm—for targeted inspection tasks. Protective windows and housings must likewise be chosen for SWIR transmission and environmental robustness, ensuring that front‑end optics do not introduce spectral artifacts.

Noise, cooling, and dynamic range in SWIR cameras

Noise in SWIR cameras arises from dark current, read noise, and fixed‑pattern components such as pixel‑to‑pixel offset and gain variation. Dark current grows rapidly with temperature and is especially critical in extended InGaAs and HgCdTe detectors, which is why thermoelectric cooling is commonly used to support long integration times and improve signal‑to‑noise ratio.

Read noise is determined by the sensor and readout integrated circuit architecture and defines the floor for short‑exposure or low‑flux imaging. Careful design of clocking, amplifiers, and analog‑to‑digital conversion can keep read noise low enough that photon shot noise and dark current dominate in typical exposures.

Dynamic range depends on the ratio between full‑well capacity and total noise and is also influenced by bit depth. Many SWIR cameras offer 8‑, 10‑, or 12‑bit output modes, where higher bit depth allows more precise digitization of subtle intensity differences. Calibration steps such as DSNU (dark signal non‑uniformity) and PRNU (photo‑response non‑uniformity) correction, often combined with on‑board lookup tables and gamma processing, reduce fixed‑pattern noise and produce cleaner images for downstream analytics.

Illumination strategies for SWIR imaging

SWIR imaging can be passive—relying on ambient sources like starlight, nightglow, or solar reflection—or active, using dedicated LEDs and laser diodes in SWIR bands. Passive modes are well-suited for defense, astronomy, and low‑light surveillance because SWIR cameras can see through haze and smoke better than VIS sensors under certain conditions.

Active SWIR illumination enables precise control over wavelength and intensity, which is crucial for machine vision, spectral imaging, and process monitoring. Arrays of SWIR LEDs or diode lasers can be matched to specific absorption bands—such as water content in food or coatings on metals—to maximize contrast at modest power levels.

Laser‑based SWIR illumination introduces coherence and speckle, which can create granular interference patterns that complicate quantitative imaging. System designers often mitigate speckle through beam shaping, wavelength multiplexing, or mechanical scrambling, while also respecting eye‑safety standards that vary with wavelength and exposure geometry.

Engineered applications: from silicon wafers to food sorting

In semiconductor manufacturing, SWIR cameras take advantage of silicon’s transparency in the 1.0–1.6 µm range to image through wafers and certain coatings to detect voids or cracks. Extended SWIR further opens bands that reveal water or polymer absorption, enabling inspection of battery components, solar cells, and composites.

In additive manufacturing and welding, SWIR cameras monitor melt pools and solidification dynamics, providing temperature‑sensitive contrast and surface information that complements thermal imaging for process control. Recycling systems use SWIR spectral signatures to distinguish between plastics, paper, and organics, enabling automated separation of polymer types and improved resource recovery. 

In surveillance, SWIR enables covert imaging with eye‑safe illumination and improved visibility through certain obscurants. SWIR cameras support free‑space optical communication links that operate through atmospheric windows. In art restoration, SWIR imaging can reveal underdrawings, over‑paint, and restoration layers in paintings thanks to differential absorption and scattering in pigments and binders. 

In food and agriculture, SWIR imaging identifies defective products by highlighting differences in moisture content, sugar or fat distribution, and foreign materials not visible to VIS cameras. High‑speed line‑scan SWIR systems on conveyors can classify fruits, nuts, and grains by internal defects or contamination at industrial throughputs.

Broadly, compact SWaP SWIR cameras are common for embedded or portable uses, cooled scientific cameras serve R&D and spectral work, and line-scan cameras are suited for high‑speed webs and conveyors where every unit must pass under a single inspection line.

Selecting the right SWIR camera

Selecting a SWIR camera starts with defining the spectral band: standard 0.9–1.7 µm covers many reflection‑based machine vision tasks, while extended 1.7–2.5 µm is needed for certain water, hydrocarbon, and specialty industrial bands. Resolution, pixel size, and frame rate then follow from application requirements such as field of view, minimum defect size, and conveyor speed.

Interface and system integration constraints guide the choice between GigE Vision, Camera Link, USB3, CoaXPress, or embedded interfaces like CSI‑2, balancing bandwidth, cable length, and determinism. Environmental conditions—including temperature, vibration, and exposure to dust or chemicals—inform decisions about cooling, housings, and IP ratings.

From a buyer’s perspective, most SWIR cameras fall into a few practical families with distinct trade‑offs in size, cooling, and performance.

SWIR camera families vs specs and use cases

These categories match how commercial SWIR portfolios are usually segmented (area scan vs. line scan, compact vs. cooled) and make it easier to match a camera to a specific application and integration model.

Integrating SWIR cameras into machine vision and embedded systems

SWIR cameras are delivered with standard machine‑vision interfaces like GigE Vision, USB3, Camera Link, and CoaXPress for industrial PCs, or CSI‑2 and similar buses for embedded platforms. Synchronization and triggering—via hardware I/O, encoder inputs, or network‑based triggers—align image capture with motion systems or illumination to ensure consistent measurement across a line or scene.

On the processing side, SWIR data can feed into FPGA, GPU, or edge‑AI pipelines for real‑time analytics and classification. Hyperspectral or multi‑band SWIR systems often require dimensionality reduction and feature extraction before inference to keep bandwidth and compute demands manageable.

Modern SDKs and driver stacks expose low‑level control over exposure, gain, cooling, and calibration, as well as access to on‑camera corrections for DSNU, PRNU, and nonlinearity. This allows engineers to build robust, production‑ready systems where SWIR images are treated as another high‑value modality alongside visible cameras, 3D scanners, and thermal imagers, rather than a niche specialty tool.

Contact our experts today to find the right SWIR camera solution for your application.