A gimbal cannot hold a line of sight it cannot measure. Every stabilisation figure on a datasheet is really a statement about the inertial measurement chain — how well the payload knows its own attitude, and how fast it knows it. Everything mechanical downstream is just the actuator.
Key takeaways
- The IMU is fighting bias drift and temperature sensitivity, not noise — noise averages out, drift does not.
- Attitude error propagates directly into geolocation error, and the effect grows with slant range.
- Dual IMU and temperature control appear on datasheets because they attack the two dominant error terms; both cost weight and power.
- Specified accuracy is achievable only with correct integration — mounting rigidity and boresight alignment determine whether you actually get it.
On this page
What the IMU is fighting
A MEMS gyro measures angular rate, and integrating rate gives angle. The problem is that any bias in the rate measurement integrates into an angle that grows without bound. A gyro with 10 °/hour bias accumulates 0.17° of error in one minute of pure integration — far worse than the 0.01° the payload is supposed to hold.
Accelerometers provide an absolute reference for pitch and roll, because gravity always points down. But accelerometers also measure the aircraft’s own acceleration, so during a turn or a gust they lie. The fusion problem is deciding, moment by moment, how much to trust each.
Yaw has no gravity reference at all. It is held by magnetometer, by GNSS course, or by carrier heading transferred from the aircraft — all of which are slower and noisier than the gyro. This is why yaw drift is the most common stabilisation complaint in the field, and why drift troubleshooting almost always starts there.
The error budget
| Error source | Typical magnitude | How it is managed | Residual after management |
|---|---|---|---|
| Gyro bias instability | 5–15 °/h | Complementary filter with accelerometer | < 0.01° |
| Gyro scale-factor error | 0.1–1 % | Factory calibration over temperature | < 0.005° |
| Temperature-induced bias drift | Large, non-linear | IMU temperature control / oven | Near-eliminated |
| Accelerometer noise | 0.5–2 mg | Low-pass, trust-weighting during manoeuvre | < 0.01° |
| Vibration rectification | Highly variable | Isolation mounts, notch filters | Design-dependent |
| Magnetometer disturbance | Up to several degrees | Carrier AHRS fusion, GNSS heading | 0.05–0.5° yaw |
| Structural flexure | 0.005–0.05° | Stiffness, boresight calibration | Mostly residual |
Why dual IMU and temperature control appear on datasheets
Two IMUs with complementary characteristics let the fusion filter cross-check and reject a sensor that has been disturbed — by a hard landing, a vibration resonance or a thermal transient. It is redundancy in the statistical sense rather than the safety sense.
IMU temperature control matters more than most integrators expect. MEMS gyro bias is strongly temperature dependent and the dependency is non-linear, so calibrating it out across the full operating range is difficult. Holding the IMU at a constant elevated temperature turns a hard calibration problem into an easy one. This is why payloads with temperature-stabilised IMUs hold accuracy through a climb from 30 °C ground to −10 °C at altitude while others drift.
Carrier AHRS fusion closes the yaw problem. Instead of relying on a magnetometer sitting inside a metal pod near motor currents, the payload consumes the aircraft’s own attitude solution and uses it as a slow reference for its fast gyro.
What this means when you specify a payload
- Ask whether the quoted accuracy is pitch/roll or all three axes. Yaw figures are almost always worse and are sometimes quietly omitted.
- Ask whether the figure is static or in flight. Static bench accuracy tells you about the encoder; in-flight accuracy tells you about the fusion.
- Ask whether carrier attitude input is supported and over what interface. Without it, yaw stability depends on a magnetometer in a hostile magnetic environment.
- Confirm the vibration environment the figure assumes. A number quoted on a bench and a number quoted behind spinning propellers are different numbers — see payload mechanical integration.
How this shows up in our payloads
Our stabilised products use dual-IMU complementary algorithms with IMU temperature control and carrier AHRS fusion, which is the combination that produces the ±0.01° figure quoted on AX-20T and on multi-sensor pods such as OP-90D. The three techniques are not independent marketing features; each one closes a specific term in the error budget above.
Related reading
Technology: gimbal stabilization technology and payload mechanical integration.
Field practice: what 0.01° means at 1,000 m, 1-axis vs 2-axis vs 3-axis and gimbal troubleshooting.
Integration decisions that determine whether you get the specified accuracy
The stabilisation figure on a datasheet is achievable, but not automatically. It assumes the payload is receiving what it needs from the aircraft and is not being fed a vibration environment its filters were not designed for.
Three integration decisions carry most of the outcome. Whether carrier attitude is fed to the payload determines yaw stability more than any property of the gimbal itself. Whether the isolation system places its resonance clear of both the blade-pass frequency and the gimbal control bandwidth determines whether the fusion filter sees clean data or rectified noise. And whether the payload is mounted rigidly enough that structural flexure does not add its own angular error determines the floor below which no amount of control effort helps.
This is why two integrators using the same payload on similar airframes routinely report different stabilisation quality. The payload is the same; the three decisions above were not.
- Feed carrier attitude to the payload over MAVLink or the vendor protocol — this is the single highest-impact yaw improvement.
- Size isolators to the actual suspended mass so the resonance lands where it was designed to.
- Keep the mounting structure stiff; flexure between IMU and optics is an error the control loop cannot see.
- Calibrate the magnetometer in the installed configuration, not on the bench, or bypass it with dual-antenna heading.
- Verify accuracy in flight at the throttle settings you will actually use, not only in hover.
FAQ
Why does my gimbal drift in yaw but not in pitch and roll?
Gravity gives the accelerometer an absolute reference for pitch and roll, so gyro bias in those axes is continuously corrected. Yaw has no equivalent reference and depends on a magnetometer or on carrier heading, both of which are slower and noisier. Yaw drift is expected behaviour, not a fault, unless it exceeds the specified figure.
What does IMU temperature control actually buy?
It removes the largest and least predictable term in gyro bias. Holding the IMU at a constant temperature turns a non-linear calibration problem into a simple one, which is why temperature-stabilised payloads keep their accuracy through large ambient swings.
Is 0.01° stabilisation meaningful at long range?
Very. At 1,000 m, 0.01° is about 17 cm of image movement. At 40x zoom that is the difference between a readable target and an unusable smear — stabilisation error is magnified by exactly the same factor as the image.