Thermal Image Sensor: Microbolometer Tech Guide 2026
Beneath the false-color image any thermal camera produces sits a piece of microelectronics that does the actual work — the thermal image sensor itself. Spec sheets list these sensors by resolution, material, and sensitivity, and the differences between them drive most of the price spread in the thermal camera market. A buyer reading “320 by 240 vanadium oxide microbolometer with 50 mK NETD” is reading a specific description of capability. This guide walks through what the sensor actually is, how the major detector materials compare, why some sensors need cooling and others do not, and which specifications matter for which kinds of work.
What a thermal image sensor is
A thermal image sensor is a flat array of tiny detectors arranged in a precise grid — typically rectangular, with row and column counts giving the sensor’s resolution. Each detector measures the heat radiation arriving at its small portion of the focal plane and reports a value. The camera’s electronics read all the detectors many times per second and assemble the readings into a frame, then another, then another, producing a video stream of thermal images.
Two detector technologies dominate the consumer and building-inspection market. Uncooled microbolometers use either vanadium oxide (VOx) or amorphous silicon (a-Si) as the active material. Both work at room temperature, both detect in the long-wave infrared band (8 to 14 micrometers), and both have come down in price enough that they appear in $250 phone dongles as well as $2,500 inspector-grade cameras. Cooled detectors using indium antimonide (InSb), mercury cadmium telluride (HgCdTe or MCT), and other exotic materials remain the standard for scientific, military, and astronomy applications where the highest possible sensitivity is required.
How a microbolometer detector works
Each individual microbolometer is a thin film of detector material — VOx or a-Si — suspended on a tiny membrane just a few microns thick, with electrical contacts at the edges. The membrane sits over a small cavity, thermally isolated from the substrate as much as the physics allows. When long-wave infrared radiation hits the detector film, the film absorbs the energy and warms up. The warming is microscopic, fractions of a millikelvin, but the film’s electrical resistance changes measurably as it warms.
The camera’s readout integrated circuit (ROIC) sits behind the detector array and samples the resistance of every pixel in turn, many times per second. The samples are digitized, calibrated against a reference, and converted to temperature values. Refresh rates are typically 9 Hz for export-controlled consumer cameras and 30 Hz or 60 Hz for commercial and industrial units. The 9 Hz limit is a regulatory artifact — under U.S. ITAR rules, faster frame rates trigger export restrictions, so most consumer cameras are deliberately limited to stay under that threshold. The home inspection tools overview covers how this affects scanning workflow.
Vanadium oxide versus amorphous silicon
The two dominant uncooled detector materials each have advantages, and most manufacturers have settled on one or the other for their consumer product lines.
Vanadium oxide (VOx)
VOx microbolometers were developed at Honeywell in the late 1980s and now dominate the inspector-grade and industrial segments. Their main advantage is high thermal coefficient of resistance (TCR), which translates directly to lower NETD numbers — typically 30 to 60 mK in modern designs. They are slightly more expensive to manufacture and require process steps that overlap with semiconductor fabrication but are not identical to it. FLIR, Fluke, and many premium brands use VOx detectors in their professional product lines.
Amorphous silicon (a-Si)
a-Si microbolometers came later and offered a path to lower manufacturing cost by sharing more process steps with standard CMOS fabrication. They typically run NETD figures between 50 and 100 mK — adequate for residential and commercial building work, weaker for fine industrial diagnostics. ULIS (now part of Lynred) is the dominant supplier of a-Si detector arrays, and several consumer brands use them in lower-cost product lines.
From a buyer’s perspective, the material itself matters less than the resulting NETD specification on the data sheet. A well-made a-Si sensor at 50 mK can outperform a poorly-binned VOx sensor at 80 mK in practice. Comparing the listed NETD is more useful than comparing detector material names.
Sensor resolutions in the consumer and inspection markets
Resolution drives price. Each step up roughly doubles the number of detectors that must be individually manufactured, tested, and binned, and the manufacturer’s yields decrease with array size because any single dead pixel can disqualify a chip from premium grading.
Common resolutions in 2026: 80 by 60 (entry-level dongles, fading from the market), 160 by 120 (standard for $250 to $500 consumer units), 240 by 180 (mid-tier consumer and entry inspector), 320 by 240 (standard inspector-grade), 384 by 288 (some inspector and commercial units), and 640 by 480 (industrial and high-end inspection). Above 640 by 480 the market shifts toward research and military equipment.
A 320 by 240 sensor delivers four times as many pixels as a 160 by 120 sensor — a meaningful difference when scanning a wall from across a room. Doubling sensor pixel count is the most reliable way to improve thermal image quality because it directly increases spatial resolution. Increasing NETD sensitivity helps subtle anomalies appear; increasing resolution helps spatial detail appear. Both matter, and the HD thermal imager guide covers the marketing gymnastics that often surround them.
Cooled detectors and why they exist
For scientific, military, and astronomical work where the absolute best sensitivity is required, detector arrays are cooled to cryogenic temperatures — often around 77 Kelvin (the boiling point of liquid nitrogen) using a Stirling cycle cooler integrated into the camera body. At those temperatures, thermal noise inside the detector drops dramatically, allowing the camera to detect temperature differences below 10 mK, far better than any uncooled sensor.
The materials used in cooled detectors are exotic. Indium antimonide (InSb) is common for mid-wave infrared (3 to 5 micrometers). Mercury cadmium telluride (HgCdTe, often called MCT) covers both mid-wave and long-wave bands depending on alloy composition. Quantum well infrared photodetectors (QWIP) appear in specialized applications. Each material has narrow tradeoffs in sensitivity, bandwidth, and noise that drive the choice for a given research or military application.
Cooled cameras cost from $25,000 to several hundred thousand dollars, weigh significantly more than uncooled equivalents because of the cooler and dewar assembly, and consume noticeably more power. They are completely outside the residential and standard commercial inspection markets, which is why home inspector training programs from InterNACHI and ASHI focus entirely on uncooled microbolometer-based equipment.
Specifications that matter for inspection-grade sensors
When comparing two thermal cameras at similar price points, six sensor specifications drive the practical difference.
Native sensor resolution. The pixel count of the actual detector array, not the display resolution. Higher is better for spatial detail.
NETD (thermal sensitivity). Lower is better. Below 80 mK is good; below 50 mK is excellent.
Spectral range. Standard for building work is 7.5 to 14 micrometers. Narrower or different bands signal a specialized product.
Frame rate. 9 Hz is consumer/export-controlled; 30 Hz or 60 Hz is commercial. Faster rates make handheld scanning smoother.
Pixel pitch. The center-to-center distance between detectors, in micrometers. Smaller pitch (12 micrometers, 17 micrometers) is newer and allows smaller cameras at the same resolution.
Operating temperature range. The temperatures the sensor itself can tolerate. Most uncooled sensors work between -10 and 50 degrees Celsius. Industrial cameras for hot-process work read higher target temperatures up to several hundred degrees, but the sensor itself still operates at room temperature.
What sensor quality cannot fix
A better sensor does not compensate for poor focus, dirty lens optics, or operator interpretation errors. The most common cause of misleading thermal images is not sensor quality — it is reflections from shiny surfaces, low-emissivity readings on bare metal, or insufficient temperature contrast in the scene. A $4,000 sensor in the hands of an untrained user still produces ambiguous data when the operator misses these traps. The physics-in-plain-language explainer walks through the interpretation issues that no sensor upgrade can solve.
Sensor specifications matter for buyers in the same way camera specifications matter for photographers — they set the upper limit of what is possible. The actual result depends on the user, the conditions, the scene, and the interpretation. Buyers comparing two cameras should pick the one whose sensor specifications match the work they will actually do, then invest in the training that turns the sensor’s data into a useful answer.
Sensor history and how the price has fallen
Thermal image sensors were originally developed for military applications in the 1960s and 1970s, with the earliest practical microbolometer arrays emerging from Honeywell research in the late 1980s. Through the 1990s a usable thermal camera cost $20,000 to $50,000 and was sold almost exclusively to industrial, military, and scientific buyers. The technology was export-controlled, the production volumes were tiny, and the per-unit manufacturing cost was correspondingly high.
Two trends drove costs down through the 2000s and 2010s. First, manufacturing volume grew as commercial uses expanded — building diagnostics, electrical maintenance, predictive industrial monitoring, and eventually consumer applications. Higher volume amortized fixed manufacturing investment across more units. Second, the introduction of amorphous silicon detectors allowed manufacturers to share fabrication infrastructure with standard CMOS chip production, dramatically reducing the marginal cost per sensor.
By 2020, basic phone-attachment thermal cameras were available under $300. By 2026, the entry price for a real thermal camera has held steady around $250 to $300 even as image quality at that tier has improved. The trend will likely continue as new sensor designs (smaller pixel pitches, larger arrays, better noise floors) move through the manufacturing supply chain. The infrared camera price tier breakdown covers the current market pricing in detail.
Pixel pitch and what it implies for camera size
Pixel pitch is the center-to-center distance between adjacent detectors on the sensor array, measured in micrometers. Smaller pitch values mean the same total resolution fits in a physically smaller chip, which allows smaller cameras at the same resolution. Pitch has decreased over the years from around 50 micrometers in early designs to 25 micrometers, then 17 micrometers, and current designs at 12 micrometers and below.
The practical implication is that a 320-by-240 sensor with 12-micron pitch is dramatically smaller than the same resolution sensor with 25-micron pitch. This is why phone-attachment thermal cameras with usable image quality have become possible — modern small-pitch sensors fit in a dongle that plugs into a phone, where the same resolution at older pitches would have required a full-size camera body. Smaller pitch can also reduce manufacturing cost if the rest of the process supports it, since more sensors fit on each silicon wafer.
Smaller pitch is not universally better, though. Smaller detectors collect less infrared energy per pixel, which can hurt sensitivity (raise NETD). Sensor designers balance pitch reduction against signal-to-noise performance for each generation. The best modern small-pitch sensors achieve excellent NETD numbers, but the engineering trade-off is real.
How sensor specifications appear in the field
A working inspector comparing two candidate cameras might see one spec sheet listing “320 by 240 vanadium oxide microbolometer, 12 micron pitch, less than 50 mK NETD, 7.5 to 14 micrometers spectral range, 9 Hz refresh.” A second sheet might list “320 by 240 amorphous silicon, 17 micron pitch, less than 80 mK NETD, 8 to 14 micrometers spectral range, 9 Hz refresh.” Both cameras have the same resolution and both will perform reasonably for residential inspection. The first camera reads slightly more subtle anomalies due to better sensitivity; the second is likely less expensive due to lower-cost sensor manufacturing.
For everyday building diagnostics, both cameras work well. For finding the subtlest moisture patterns or detecting marginal insulation gaps, the first camera’s lower NETD is meaningfully better. Buyers should match the spec to the actual work rather than chasing the best sensor specifications regardless of need. InterNACHI’s training programs include practical guidance on matching equipment to anticipated inspection scope.
References
- DOE thermographic equipment guidance — U.S. Department of Energy
- InterNACHI infrared certified inspector standards — InterNACHI
- ASHI thermal imaging in inspection guidance — American Society of Home Inspectors
- ASHRAE measurement and instrumentation publications — American Society of Heating, Refrigerating and Air-Conditioning Engineers
Thermal imaging cameras
Infrared cameras reveal hidden moisture, missing insulation, and air leaks. Phone-attachment models are the budget entry point; standalone units have higher resolution.
| Product | Why | Buy |
|---|---|---|
 FLIR ONE Pro (phone) | Plugs into iPhone/Android; inspector favorite. | Amazon — $349.00 |
 Topdon TC001 | High-res phone module at a low price. | Amazon — $199.99 |
 FLIR C5 Compact | Standalone pocket camera with Wi-Fi. | Amazon — $449.00 |