Nearly every drone thermal camera — including all UAVThermal payloads — is built around an uncooled vanadium oxide (VOx) microbolometer. It is a grid of microscopic thermometers, not a photon detector, and that single fact explains most thermal camera behaviour: why sensitivity is quoted in millikelvin, why the image briefly freezes in flight, why fast movement smears, and why these sensors fly at 110 g when cooled imagers need kilograms.
Key takeaways
- A microbolometer measures heating, not photons. Each pixel is a suspended VOx bridge whose resistance changes as absorbed infrared warms it.
- Because nothing needs cryogenic cooling, the whole detector fits in a 69–130 g gimbal and runs from a few watts — this is what made sub-250 g thermal payloads possible.
- The physics costs you response speed and stability: motion blur at high angular rates, and thermal drift that must be re-zeroed periodically by the FFC shutter.
- Two numbers describe core quality: NETD (sensitivity, in mK) and pixel pitch (12 µm across the current UAVThermal range).
On this page
A grid of thermometers, not photodiodes
Each pixel is a microscopic bridge of vanadium oxide, suspended over the readout circuit on legs thin enough to isolate it thermally from everything around it. Incoming long-wave infrared warms the bridge by a tiny fraction of a degree; the resistance of VOx changes measurably with that temperature; the readout circuit measures the resistance of every bridge in the array, many times per second, and converts the pattern into an image.
Across UAVThermal payloads the array is 256×192 or 640×512 bridges at 12 µm spacing, sensing the 8–14 µm band. Nothing in that chain requires cooling to cryogenic temperatures, because the detector is not trying to count individual photons — it is measuring the heat those photons deposit. That is the entire reason an uncooled core can be instant-on, sealed, and light enough to hang under a small multirotor.
What the physics gives and takes
| Property | What uncooled VOx gives you | What it costs |
|---|---|---|
| Mass and power | 69–130 g micro payloads, a few watts | Lower sensitivity than a cooled photon detector |
| Startup | Instant-on, no cooldown | — |
| Response speed | Adequate for survey and most tracking | Thermal time constant causes smear at high angular rates |
| Stability | No consumable cryocooler to wear out | Array drifts with its own temperature; needs periodic FFC |
| Sealing | Fully sealed, no moving cooler | Sun exposure can permanently damage pixels |
| Cost | An order of magnitude below cooled | Range limited compared with cooled mid-wave systems |
The response-speed limitation is the one that surprises operators. A microbolometer pixel must physically warm and cool, so a target crossing the frame quickly — or a fast gimbal slew — produces smear that no processing removes. It is not motion blur from exposure time; it is the detector’s thermal time constant. Slower, deliberate slews are not a stylistic preference, they are a requirement of the sensor.
NETD and pixel pitch across the range
| Payload | Thermal resolution | Pixel pitch | NETD | Weight |
|---|---|---|---|---|
| MV-2M | 256×192 | 12 µm | <50 mK @ 25 °C | 110 g |
| MV-2P | 640×512 | 12 µm | <40 mK @ 25 °C | 130 g |
| OP-90A | 640×512 | 12 µm | <50 mK @ 25 °C | 608 g pod |
| OP-125A | 640×512 | 12 µm | <50 mK @ f/1.0, 25 °C | 1055 g |
| LX-9B | 640×512 | 12 µm | <50 mK @ f/1.0, 25 °C | 1158 g |
Read the qualifier, not the number. An NETD of <50 mK quoted at f/1.0 and one quoted with no aperture stated are not comparable figures. The aperture changes how much energy reaches the detector, and a slower lens will always measure worse. This is the single most common padding in thermal spec sheets — see how to read a thermal spec sheet.
Why the image freezes: flat-field correction
Every bridge in the array responds slightly differently, and each one drifts as the sensor body warms up in flight. Factory non-uniformity correction handles the manufacturing spread; it cannot handle drift that happens after takeoff. Flat-field correction closes a shutter across the array for a fraction of a second, giving every pixel an identical scene, and re-zeroes the offsets against that reference.
The visible symptom is a brief freeze and a faint click. It is normal, it happens more often in the first minutes after power-on while the core is still stabilising, and skipping it produces fixed-pattern noise — faint grids or streaks that stay put while the scene moves beneath them. The full chain is covered in the thermal image processing pipeline and the calibration chain.
Uncooled versus cooled, in one rule
Cooled photon detectors win on extreme range and on fast optics, because they genuinely count photons and can be paired with apertures an uncooled core cannot exploit. They also cost, weigh and consume roughly an order of magnitude more, and they contain a cryocooler with a finite service life.
For UAV payload work under 2 kg, uncooled VOx is the practical universe. The rule is simple: if the aircraft has to carry it and the mission has to be routine, the detector is uncooled. The full comparison of uncooled and cooled cores covers the cases where the answer changes.
What this means when you specify a payload
- State NETD with its aperture and sensor temperature, or the figure is unenforceable.
- Decide resolution from the smallest thing you must resolve at your working distance, not from the larger number.
- Budget for FFC interruptions if the mission involves continuous tracking — ask how often the shutter fires and whether it can be deferred.
- Plan slew rates around the detector time constant; a payload that smears on fast pans is behaving normally.
- Write sun-exposure handling into the operating procedure. It is the one failure mode that permanently damages the array.
Never point the payload at the sun. The detector has no consumable element and no cooler to wear out, so its practical lifetime is set by the electronics and mechanics around it — and by whether anyone ever parks the field of view on the sun. Sun-burn protection helps; it is not a licence to try.
Related reading
- Thermal Sensitivity (NETD) Explained for UAV Payload Buyers
- Thermal Pixel Pitch: Why 12 µm Replaced 17 µm on Drone Payloads
- Uncooled vs Cooled Thermal Cores: Why Drones Chose Microbolometers
- Why Thermal Cameras Live at 8–14 µm: The LWIR Window
- Sub-250 g Drones, Real Thermal: What 69–130 g Micro Payloads Can Do
- The Thermal Image Processing Pipeline: NUC, FFC, AGC and Detail Enhancement
- From NETD to DRI: How Thermal Detector Performance Is Actually Calculated
- Thermal Infrared Optics: Germanium Lenses, f/1.0 Apertures and Athermalization
FAQ
Why do thermal cameras pause briefly during flight?
Flat-field correction. A shutter closes across the array to give every pixel a uniform reference, re-zeroing the drift that accumulates as the sensor warms. It is normal, takes a fraction of a second, and happens most often in the first minutes after power-on.
Do microbolometers wear out?
The detector itself has no consumable element and no cryocooler, so lifetime is set by the electronics and mechanics around it. The one thing that does destroy pixels permanently is pointing the payload at the sun.
Why does my thermal image smear when I pan quickly?
Each pixel has to physically warm and cool, so the detector has a thermal time constant. Fast slews or fast-crossing targets produce smear that no image processing removes. Slow the slew rate rather than looking for a software fix.
Questions about the technology? Talk to our engineers — we reply within 2 business days.