Skip to content
The Thermogram

Published September 14, 2026 · 6 min read

What datasheets don't tell you

Seven documented discrepancies, one geometrically impossible field of view, and nearly half of devices too poorly documented to be honestly scored.

Building a technical database forces you to cross-check sources. While surveying 28 thermal cameras from manufacturer documentation, we didn't just collect numbers: we found contradictions. Here they are, all verifiable.

InfiRay contradicts itself on NETD

The official P2 Pro datasheet, published by IRay Technology, lists a NETD of 50 mK or better. The brand's own store, Amazon, and almost every retailer advertise under 40 mK.

Between 40 and 50 mK, that's a 25% difference in sensitivity, on the criterion informed buyers check first. We used the datasheet's value.

Bosch advertises worse figures than its own retailers

On its own product page, Bosch lists an infrared accuracy of ±3.0°C for the GTC 400 C. Several French retailers list ±2°C.

It's the most unusual discrepancy in our survey: here, the manufacturer is more conservative than its own distribution chain.

Worse, several retailers describe the GTC 400 C with a -20 to 600°C range and a 256 × 192 sensor — actually the specs of the GTC 600 C. A buyer could order a 160 × 120 sensor believing they're getting a 256 × 192.

FLIR publishes an impossible field of view

The TG165-X's datasheet advertises a field of view of 51° × 66°. A vertical field wider than the horizontal one, on an 80 × 60 sensor that is nonetheless in landscape format. This is geometrically inconsistent: the two values are clearly swapped.

No one seems to have flagged it since 2022. Other retailers instead list 50° × 38.6°, the figures for the older TG165.

On the C5, FLIR advertises an IFOV of 6.3 mrad, while recalculating it from its own 54° field of view and 160 columns gives 5.89 mrad. Both values appear on the same datasheet.

HIKMICRO doesn't agree with itself on its own sensor

The official Gx0 series page advertises a 640 × 512 sensor, i.e. 327,680 pixels. Distributors and product listings state 640 × 480, i.e. 307,200 pixels. Twenty thousand pixels of difference depending on the source, at the same manufacturer.

UNI-T contradicts itself in its own manual

The UTi260B is advertised at 50 mK by specialist distributors and 60 mK on the manufacturer's Amazon listing. Its low end of range is given as -20°C on the UNI-T page and -15°C in the user manual.

Hidden functions, and missing ones

The reverse phenomenon exists too. The datasheets for HIKMICRO's B series and B20S mention neither the emissivity setting nor reflected-temperature correction. Yet both functions, which determine measurement accuracy far more than pixel count, are actually present: they appear in the user manual.

A manufacturer highlights a heavily interpolated resolution while staying silent about the functions that make its measurements reliable.

Conversely, some information only appears on an accessory's own datasheet. The HIKMICRO macro lens degrades accuracy from ±2°C to ±3°C or ±3% and caps the range at 150°C. None of that appears on the camera's own datasheet.

Share of specs published by the manufacturerTOPDON TC004 Mini26%FLIR ONE Pro LT29%HIKMICRO G2035%FLIR ONE Edge Pro36%TOPDON TS00138%FLIR ONE Pro38%Seek Thermal Reveal XR41%Seek Thermal Shot43%TOPDON TC00343%
Of the criteria we track, the share the manufacturer actually documents.

Silence, more common than error

The most widespread problem isn't contradiction, it's absence. Out of 101 devices, a quarter don't publish enough specs to receive any score at all for the most common use case, building diagnostics, and more than a third beyond that only get a partial estimate rather than a full score.

  • TOPDON publishes neither field of view, nor focal length, nor aperture for the TC001: the IFOV cannot be calculated.
  • HIKMICRO publishes neither weight, nor dimensions, nor IP rating, nor drop resistance for its G series, despite it being the best-performing range in our database.
  • Seek Thermal publishes neither measurement accuracy nor its NETD conditions.
  • Bosch publishes no frame rate at all for the GTC 400 C.
  • FOTRIC publishes neither weight, nor dimensions, nor an IP rating.

Our rule

When a manufacturer doesn't publish a spec, we write that down. No value is ever estimated, no field is filled in by inference. The corresponding criterion drops out of the calculation and lowers the confidence index shown next to the score.

Below 70% confidence, no score is published, and the device carries the note "insufficient data". That's not a judgment on the product: it's a statement about its documentation.

There is one exception, though. On criteria that specifically concern the manufacturer's candor, such as the actually measured resolution, not publishing counts as zero rather than being neutralized. Without this, silence would pay off more than the truth.

Read the full methodology.


All values cited come from manufacturer datasheets and the sources listed on each product page. No data is estimated.