Thermal Imaging Technology Guide — How It Works & Key Specifications

How Thermal Imaging Works

All objects above absolute zero (-273.15°C) emit infrared radiation. Thermal cameras detect this radiation using a focal plane array (FPA) of detector pixels, each converting incoming IR photons into an electrical signal. The resulting 2D array of temperature measurements is rendered as a visible thermogram — warmer objects appear brighter. Unlike visible-light cameras, thermal cameras require no illumination and can see in total darkness, through smoke, fog, and most obscurants.

VOx Microbolometer — The Industry Standard

Vanadium Oxide (VOx) microbolometers are the dominant uncooled thermal detector technology. Each pixel consists of a suspended VOx thin-film membrane that absorbs IR radiation, heats up proportionally, and changes electrical resistance. This resistance change is read out by a ROIC (Readout Integrated Circuit) beneath the detector array.

Key VOx Advantages

• Room-temperature operation — no cryogenic cooling needed
• High TCR (Temperature Coefficient of Resistance): ~2-3%/K
• NETD typically <40mK at f/1.0 — competitive with early cooled detectors
• Long lifetime — hermetically sealed vacuum package, no consumable parts
• Wide operating temperature: -40°C to +85°C
• Low power consumption: <1W for typical QVGA/VGA modules

ZanVision uses proprietary VOx thin-film deposition processes to achieve industry-leading uniformity and sensitivity across the entire FPA.

LWIR vs MWIR — Spectral Band Comparison

For 95% of commercial applications, uncooled LWIR provides the best balance of performance, cost, size, and reliability.

NETD — Thermal Sensitivity

Noise Equivalent Temperature Difference (NETD) is the primary sensitivity metric for thermal cameras. It represents the smallest temperature difference the camera can detect above its inherent noise floor, expressed in millikelvins (mK). Lower is better.

• NETD <20mK — Cooled MWIR detector, scientific/long-range military grade
• NETD 20-35mK — Premium uncooled VOx, professional surveillance/inspection
• NETD 35-50mK — Standard uncooled VOx, general security/industrial
• NETD >50mK — Entry-level or consumer-grade thermal cameras

NETD is typically measured at f/1.0 aperture. Real-world sensitivity degrades with slower optics — a camera with 50mK NETD behind an f/2.0 lens will have approximately 100mK effective NETD.

DRI Ranges — Detection, Recognition, Identification

DRI follows Johnson's Criteria, which defines the number of pixel pairs needed across a target for each discrimination level:

• Detection (D): ~1.5 pixels across target — "something is there"
• Recognition (R): ~6 pixels across target — "it's a person, not a vehicle"
• Identification (I): ~12 pixels across target — "it's a specific person/vehicle type"

DRI range depends on: detector resolution (more pixels = longer range), lens focal length (longer = more magnification), pixel pitch (smaller = better spatial resolution at same focal length), target size, and atmospheric conditions.

Example: 640×512, 12μm, 100mm lens

• Human detection (1.8m × 0.5m): ~3.8 km
• Human recognition: ~1.0 km
• Vehicle detection (4.5m × 2.3m): ~9.5 km
• Vehicle recognition: ~2.4 km

Pixel Pitch Evolution — From 17μm to 12μm

Pixel pitch — the physical size of each detector element — has steadily decreased over the past two decades. This miniaturization drives three key benefits:

• More pixels per wafer: 12μm pitch allows 4× more detectors per wafer than 25μm, reducing cost
• Higher resolution, same optics: 1280×1024 12μm FPA fits in the same optical format as 640×512 17μm
• Smaller optics, same resolution: A 12μm pixel with a 75mm lens achieves the same IFOV as a 17μm pixel with 100mm — reducing lens size, weight, and cost by ~30%

ZanVision's current production uses 12μm pixel pitch across the full product line from 256×192 to 1280×1024.