Introduction
Infrared (IR) imaging is hitting new highs with small-pitch sensors like InfiRay's 8μm LWIR detector, shrinking pixel sizes to pack more detail into compact arrays. But how do these designs stack up against the diffraction limit, especially in Long-Wave Infrared (LWIR, 8-14μm)? Let's unpack the tech and its real-world impact for engineers chasing cutting-edge performance.
The 8μm LWIR Breakthrough
The 8μm uncooled microbolometer unleashed-a vanadium oxide (VOx)-based array-delivers a stunning 1920x1080 resolution. Each pixel absorbs LWIR radiation, shifting resistance to map heat signatures, with a custom ASIC chip refining the signal for clarity (NETD < 50mK). This small pitch crams over 2 million pixels into a lightweight package, ideal for drones or handheld gear, but it's the interplay with optics that defines its limits.
Diffraction Limit Reality
For LWIR at 8μm wavelength with F/1.2 optics, the diffraction-limited spot (Airy disk) is ~23.42μm, the so called Airy Disk Size. It specifies that the diffraction-limited spot size is:
- d = 2.44 * λ * F#, by Using λ = 8μm (0.000008m) and F/1.2 optics,
- d = 2.44 * 0.000008 * 1.2 ≈ 23.42μm
This is the diameter of the smallest spot the optics can focus onto the detector plane at the 8μm wavelength (i.e., the shortest bound of the LWIR spectrum).
Nyquist sampling calls for ~12μm pixels (To resolve this spot without aliasing, the pixel size should be at least half the Airy disk diameter) to match this, resolving the smallest optical detail. InfiRay's 8μm pixels oversample this spot (~3 pixels across), falling below the Nyquist threshold. Does it break the diffraction limit? No-it's a physical ceiling tied to wavelength and aperture-but it bends the rules for practical gain.
(The Airy Disk - from Wikipedia)
Why Oversampling Wins
Smaller pixels (Oversampling's advantages) are manifold:
- Enhanced Spatial Resolution: By sampling sub-diffraction intensity gradients, 8μm pixels improve edge definition and detail perception, critical for distinguishing small features (e.g., a distant target's silhouette).
- Improved Modulation Transfer Function (MTF): Smaller pixels extend the MTF curve, preserving contrast at higher spatial frequencies, even if optical resolution caps at ~11.71μm
- Reduced Aliasing: Oversampling mitigates high-frequency artifacts, ensuring cleaner images for automated analysis in EO/IR systems.
- Support for Digital Enhancement: Dense sampling enables deconvolution and super-resolution algorithms, reconstructing sharper images by modeling the optical point spread function (PSF).
SNR and The Sensitivity Trade-Off
There's a catch: an 8μm pixel (64μm²) grabs ~44% the energy of a 12μm one (144μm²), risking SNR (Singal to Noise Ratio) unless offset by low-noise VOx and fast lenses. Research (e.g., SPIE) pegs 5-6μm as the practical minimum with F/1.2 optics under high-SNR conditions, where sensitivity holds up. 8μm sits comfortably above this, striking a sweet spot.
*************************************************************************
TECHNICAL DETAILS ABOUT SNR
Smaller pixels reduce the IR energy collected, impacting SNR, which is critical for LWIR microbolometers detecting subtle temperature differences (measured as NETD).
- Photon Collection: Pixel area dictates signal strength:
6μm pixel: Area = 6μm * 6μm = 36μm²
5μm pixel: Area = 5μm * 5μm = 25μm²
Compared to 12μm (144μm²), 5-6μm pixels collect 17-25% of the energy, reducing signal.
- Noise Floor: Noise sources (thermal, readout, 1/f) don't scale linearly with area. As pixel size drops, SNR falls unless noise is minimized through high-sensitivity materials (e.g., VOx) or fast optics (F/1.2, which increases light collection by ~44% over F/2).
- High SNR Conditions: With F/1.2 optics and modern uncooled detectors, 5-6μm pixels achieve acceptable NETD (e.g., <50mK) in high-SNR scenarios (e.g., high-contrast targets or long integration times). Below 5μm, SNR degrades too much for practical use without extraordinary advancements, as the signal drops faster than noise can be suppressed.
*************************************************************************
Practical Implications and Industry Context
Oversampling's benefits-sharper details, compact high-resolution arrays-shine in applications like surveillance and targeting. We have witnessed the 1920x1080 sensor, launched amid 2025's IR innovation wave, outperforms traditional 12-17μm designs in detail without cryogenic cooling. Comparatively, RTX's MWIR sensors prioritize sensitivity for warm targets, while Anduril's "Iris" series (2024) embeds AI at the pixel level for multi-target tracking. Also InfiRay's uncooled LWIR niche lies in its SWaP efficiency and resolution, leveraging engineering ingenuity within diffraction's bounds.
Optically, the 8μm pitch doesn't resolve sub-11.71μm features-diffraction forbids it-but practical resolution exceeds Nyquist expectations through oversampling and processing. This aligns with industry trends toward integrating small-pitch sensors with AI, where dense data fuels machine learning for threat classification in EO/IR .
Future Directions
Pushing pitches below 8μm-say, to 5-6μm-promises further resolution gains but demands innovations in noise reduction (e.g., advanced VOx alloys) and optics (e.g., F/1 or adaptive lenses). Super-resolution techniques, combining multi-frame data or sub-pixel shifts, could amplify these sensors' effective resolution, a frontier for AI-driven systems. Fabrication challenges, like pixel uniformity and fill factor, also loom larger at smaller scales, necessitating precision manufacturing advances.
Other Impacts on the IR system
Small-pitch IR sensors transform system architecture by boosting thermal response speed, reducing Size, Weight, Power, and Cost (SWaP-C), and enabling multi-spectral integration. Smaller pixels enhance detection rates for dynamic tracking, while compact arrays shrink optical and power subsystems, optimizing SWaP-C for platforms like wearables or micro-drones. Dense grids also support hybrid spectral designs, merging wavelength bands into versatile, unified modules. Yet, achieving array uniformity demands precision manufacturing, challenging cost and scalability. This drives IR architectures to balance performance gains with SWaP-C constraints.
Conclusion
Small-pitch LWIR sensors like 8μm design exemplify a pragmatic stretch beyond the diffraction limit's theoretical edge. By oversampling the optical spot, they deliver enhanced resolution and flexibility without defying physics, balanced against sensitivity trade-offs mitigated by fast optics and smart design. For engineers, these systems offer a blueprint: optimize within constraints, harness digital tools, and pave the way for IR imaging's next leap-pixel by pixel.