If you fly an ArduPilot or PX4 aircraft, most of the thermal payload comparisons published online are about products you cannot mount. The Zenmuse and Vue TZ20-R families connect through DJI’s proprietary interface and fit DJI airframes only. Your decision space is independent payloads that speak open interfaces — and the criteria are different, because you are the integrator. This page works through what to check, in the order that eliminates the most options fastest.
- Control interfacesMAVLink / UART / S.BUS / CRSF / PWM / Ethernet
- Flight stacksArduPilot, PX4
- Gimbal mass range17 g – 125 g
- Payload mass range69 g – 1,158 g
- Thermal resolutionsUp to 640×512
- Airframe typesMultirotor, fixed-wing, VTOL, FPV
Key takeaways
- Confirm the interface first. It eliminates more candidates than any other criterion, and it is not negotiable after purchase.
- MAVLink is the capable option — pointing, mode and telemetry over the existing link. S.BUS, CRSF and PWM work but are one-way control.
- Size the payload to the airframe you already fly. Mass is deducted from endurance on every flight, permanently.
- Radiometric or not is a mission requirement, not an upgrade. Settle it before comparing resolutions.
- Buy the resolution your working distance needs, not the highest number available.
On this page
Why the usual comparisons do not apply
The published comparison landscape for UAV thermal payloads is dominated by DJI ecosystem products, because that is where the volume is. Those articles assume a Matrice underneath, and the specifications they compare — SkyPort compatibility, DJI Pilot integration, which Matrice generation is supported — are meaningless on a custom build.
On your side of that boundary the questions change. Nobody has certified this combination for you, so the specifications that matter are the ones that determine whether the payload will talk to your flight controller, fit your weight budget, and produce the data your deliverable requires. Ecosystem maturity, which is a genuine DJI advantage, is replaced by your own integration work — which is the real cost of this route and worth acknowledging plainly. The trade is set out in the two-route comparison.
Interfaces: the first thing to confirm
Interface compatibility eliminates more candidates than anything else, and unlike almost every other specification it cannot be worked around after purchase. Check it first.
| Interface | What it gives you | Typical use |
|---|---|---|
| MAVLink | Bidirectional: pointing commands, mode control, status telemetry | ArduPilot / PX4 builds with a companion computer or GCS control |
| UART serial | Direct command link, vendor protocol | Simple integrations, companion-computer control |
| S.BUS | Channel-based control from the RC link | Operator-controlled pointing without GCS involvement |
| CRSF | Control plus telemetry back over the RC link | Long-range and FPV builds running ExpressLRS or Crossfire |
| PWM | Basic proportional control per channel | Minimal builds, simple tilt control |
| Ethernet | High-bandwidth video and control | Larger platforms, IP video pipelines, multi-payload aircraft |
For most ArduPilot and PX4 builds MAVLink is the right answer, because pointing control and status come back through the link you already have rather than needing a parallel channel. But confirm which MAVLink gimbal messages a specific payload implements — support is not uniform, and “MAVLink compatible” on a datasheet can mean anything from full mount protocol coverage to a minimal subset. Interface behaviour is covered in more depth in payload integration protocols.
If you are on an FPV or long-range build: CRSF is usually the pragmatic choice. It carries control and telemetry over the ExpressLRS or Crossfire link you are already flying, which avoids adding a companion computer purely to point a camera.
Sizing to the airframe you already have
On a DJI airframe the payload is fixed and the aircraft is designed around it. On your build the opposite is true: the aircraft exists, and the payload has to fit it. That inverts the sizing question in a useful way, because you already know your available mass.
Independent payload mass spans a wide range — gimbals alone from 17 g to 125 g, and complete payloads from 69 g to 1,158 g. That range is the main structural advantage of this route, and it is worth using deliberately rather than defaulting to the most capable unit you can lift. Payload mass is deducted from endurance on every flight for the life of the aircraft, and the relationship is worse than proportional because carrying more requires more lift to carry it. We work through the numbers in payload weight versus flight time.
A practical approach: set your endurance target first, work back to the payload mass that allows it, and treat that as a hard ceiling when comparing sensors. It is a more disciplined constraint than “what will the airframe lift”, which almost always leads to buying too much payload.
Sensor specifications that actually decide it
With interface and mass settled, only a few sensor specifications genuinely change the outcome.
Radiometric or not. If your deliverable is a temperature — an inspection report with thresholds, solar panel fault classification, electrical survey — you need radiometric output, and a non-radiometric sensor cannot be made to substitute regardless of its resolution. If you need to see heat rather than measure it, thermal video is sufficient, lighter and cheaper. Settle this before anything else.
Resolution against working distance. 640×512 extends detection and identification range over 384×288, but it costs money and often mass. If your targets are close or large, the lower tier may finish the mission with more flight time. Match it to the distance you actually work at, using DRI ranges and the 256 / 384 / 640 comparison.
Field of view. Determined by lens focal length, and it sets your area coverage rate and your pixels-on-target simultaneously, in opposite directions. This is frequently the specification that decides mission feasibility, and it is frequently the one buyers check last.
NETD and frame rate matter at the margins — low-contrast scenes and moving platforms respectively. Vendors report both inconsistently, which is one reason we wrote how to read a thermal spec sheet.
The order to work through it
Working through it in this order eliminates the most options at each step and avoids falling in love with a sensor you cannot mount:
- Interface. Confirm the payload speaks something your flight controller exposes, and confirm the specific message set if it is MAVLink.
- Radiometric requirement. Decide whether you are measuring temperature or observing heat. This is binary and it halves the field.
- Mass ceiling. Set your endurance target, derive the maximum payload mass, treat it as fixed.
- Field of view. Match to working distance and area coverage rate — this is where mission feasibility is won or lost.
- Resolution. Choose the tier your working distance requires, within the mass ceiling you already set.
- Mechanical and power. Confirm mounting geometry, vibration isolation and supply voltage against your airframe.
Browse configurations under EO/IR payload cameras and stabilized UAV gimbals. If you are still deciding between this route and buying an integrated aircraft, the H20T comparison covers the same-resolution case directly.
Related reading
- UAV payload integration protocols
- Zenmuse H30T alternatives: DJI airframe or independent payload
- DJI H20T vs an independent 640×512 thermal payload
- How payload weight trades against flight time
- 256, 384 or 640: choosing thermal resolution
- DRI explained: detection, recognition, identification ranges
- How to read a thermal camera spec sheet
- Stabilized UAV gimbals
FAQ
What thermal camera works with ArduPilot?
Any payload that speaks an interface ArduPilot exposes — MAVLink, UART serial, S.BUS, CRSF or PWM, with Ethernet on larger builds. MAVLink gimbal control is the most capable option because it carries mount commands and telemetry in both directions. What will not work is a DJI payload; those use the SkyPort/PSDK interface and are locked to DJI airframes.
Can I control a thermal gimbal over MAVLink from Mission Planner or QGroundControl?
Yes, where the payload implements MAVLink gimbal messages. That is the route that gives you pointing control, mode switching and status telemetry through the same link as the rest of the aircraft, without a separate control channel. Confirm the specific message set the payload supports before purchase rather than assuming full coverage.
How much can a thermal payload weigh on a custom multirotor?
It depends on your airframe, but the rule of thumb is to treat payload mass as a direct deduction from endurance rather than a spare capacity figure. Independent thermal payloads span roughly 69 g to 1,158 g and gimbals alone from 17 g to 125 g, so on most builds the constraint is your endurance target rather than lift capability.
Do I need a radiometric thermal payload?
Only if your deliverable is a temperature value — inspection reports with thresholds, solar panel analysis, electrical surveys. If you need to see heat rather than measure it, thermal video output is sufficient and usually lighter and cheaper. This is the single specification most worth settling before you compare anything else.
Is 640×512 always better than 384×288 for a custom build?
Not always. Higher resolution extends detection and identification range, but it costs money and often mass, and on a weight-limited airframe that mass comes out of flight time. If your targets are close or large, a 384×288 sensor may complete the mission with more endurance. Match the sensor to your working distance rather than buying the highest number.