Imaging components. Procurement support.

Reading a thermal camera core datasheet: what each number means and the conditions behind it

Thermal camera core datasheets look much alike: 640 × 512, 12 μm, 40 mK, ±2°C. The numbers can only be compared when you know the conditions behind them, and those are usually in the footnotes. Here is how to read each line, using the maker’s figures for MINI 640-2 Radiometric as the example.

Detector: type, resolution, pixel pitch, band

Uncooled VOx microbolometer, 640 × 512, 12 μm pitch, 8–14 μm. Resolution times pitch gives the active area, 7.68 × 6.14 mm here. Pitch divided by focal length gives the IFOV, the angle one pixel sees: 1.3 mrad with a 9.1 mm lens. Two cores with the same resolution but different pitches need different lenses for the same field of view, so compare them lens and all.

NETD: always with its conditions

NETD is the temperature difference that gives a signal equal to the pixel’s temporal noise. It is a noise figure, not a promise of what difference you will see in the processed image. MINI 640-2 Radiometric lists ≤40 mK at 25°C, F1.0, 25 Hz; 25 Hz is the test condition, not this core’s 30 Hz output.

The f-number matters most. The power reaching each pixel scales roughly as 1/N² (more exactly 1/(1 + 4N²)), so an F1.1 lens instead of F1.0 raises NETD by about 17–21%, before transmission losses in the lens (calculated, first-order). The standard 9.1 mm lens on this core is F1.1, so the listed figure is not at the f-number you will use. Compare cores only at the same f-number, frame rate and temperature. And NETD is about noise, not accuracy: a 40 mK core can still read 2°C off.

Frame rate and thermal time constant

MINI 640-2 Radiometric runs at 30 Hz; the MINI 640-2 imaging core runs at 60 Hz. Frame rate matters for moving scenes and video, less for monitoring that reads a few values per second. The thermal time constant (under 12 ms here) is how fast a pixel follows a change: about 63% of a sudden step after one time constant, about 95% after three. It sets how much a moving target smears, and a higher frame rate does not remove that.

Measurement ranges, accuracy and ambient

Two ranges, −20°C to 150°C and 100°C to 650°C, with accuracy of ±2°C or ±2%, typical. Three things to note. “Typical” is not a guaranteed limit. “Or” gives a fixed figure and a proportional one, and the datasheet does not say which applies where; at 300°C they differ threefold (2°C against 6°C), so ask how it is defined. And measurement has its own ambient range, −10°C to +70°C here, narrower than the −40°C to +80°C for imaging: a core that images in the cold may not measure there. On a real target, emissivity, reflections, target size in pixels and focus usually matter more than the accuracy line.

Non-uniformity correction

The datasheet line usually says only “shutter correction”, “shutterless” or both; MINI 640-2 Radiometric supports both. What it does not say is what you need for your product: how often correction runs in the mode you will use, how long the image pauses with a shutter, and whether temperature output is held or flagged during it. Ask those three questions.

Power, size and weight: “typical” and “without lens”

Under 0.7 W, typical, with USB output at 25°C, from a single 5 V supply: about 140 mA typical (calculated). Allow headroom for start-up and shutter current peaks, which the datasheet does not list; interface, ambient temperature and operating mode also change the figure. The size, 21 × 21 × 10.3 mm, and the weight, under 8.6 g, are without the lens. The 9.1 mm lens with its flange adds 10.5 g, so about 19 g for core and lens without the interface board (calculated), and it adds length.

Lens lines: field of view, sharp range, detection distances

  • Field of view: a listed figure can differ from the focal-length formula, mostly because of lens distortion. For the 9.1 mm lens on this core the maker lists 48.7° horizontally; 2 × atan(3.84 / 9.1) gives about 45.8°, about 6% narrower (calculated). Use the listed figure for coverage, and the IFOV for pixels on a target at the image centre.
  • Sharp range: “1.2–5.5 m” for the 9.1 mm lens on this core is the depth of field at the fixed factory focus. Closer or farther, the image softens. It is not a measurement range.
  • Detection, recognition and identification distances (680, 230 and 110 m for a standing person with this lens): these are pixel-count criteria for a trained observer, at about 50% probability, with good contrast and clear air. At 680 m a standing person is only about 2 pixels tall (calculated). They are not distances at which temperature can be measured.

Environment lines

Vibration 6.06 g random on all axes; shock 80 g, 4 ms, terminal-peak sawtooth, three axes in both directions; humidity 5–95% non-condensing; storage −50°C to +85°C; all measured under laboratory conditions, with the test report as the final word. A vibration figure without its spectrum and duration cannot be compared with another supplier’s, so ask for the test profile if your product sees serious vibration. “Non-condensing” means your housing has to keep moisture off the core.

Interfaces and signal levels

Check which outputs carry what. MINI 640-2 Radiometric comes in USB 2.0, MIPI (2-lane) or DVP output versions, each carrying temperature data; control runs over UART, I²C or USB. Check signal levels too: DVP and the I²C slave port default to 1.8 V (switchable to 3.3 V); UART and the I²C master port run at 3.3 V. The connector is a 50-pin, 0.4 mm board-to-board type.

Comparing two cores

  1. Resolution and pitch, or compare lens and all
  2. NETD at the same f-number, frame rate and temperature, and the f-number of the lens you will actually use
  3. Frame rate and thermal time constant for your motion
  4. Measurement ranges, how the accuracy figure is defined, and the measuring ambient
  5. Correction interval, pause and how temperature output behaves during it
  6. Power with its conditions and peaks; size and weight with the lens
  7. Field of view as listed, sharp range, and what the detection distances assume
  8. Vibration test profile, signal levels, and which interface carries temperature

The figures above are the maker’s for MINI 640-2 Radiometric, measured under laboratory conditions.

View MINI 640-2 Radiometric

Related guides

Comparing thermal cores?

Send the datasheet lines that matter for your design and how you will use the core. We reply within 24 hours on working days.

Discuss your requirements