What's the difference between compass variance and compass motor interference?

Variance is the EKF's observation that compass readings disagree with predictions (XKF4.SM rising). Motor interference is one specific cause — current-induced magnetic bias that varies with throttle. The fix for interference is CompassMot calibration plus physical separation; the fix for other variance causes (stale calibration, external structures) is different.

When should I use COMPASS_MOTCT = 1 versus 2?

Use COMPASS_MOTCT = 2 (current-based) if you have a current sensor in your power module — almost every modern build does. It tracks reality more closely during throttle transients. COMPASS_MOTCT = 1 (throttle-based) only when no current sensor is fitted; it's a less accurate fallback.

What's a healthy magnitude for COMPASS_MOT_X/Y/Z after calibration?

Under 60 mGauss total vector magnitude is excellent. 60–100 mGauss is acceptable — the calibration helps but doesn't eliminate the issue. Above 100 mGauss means physical separation between the compass and the power leads is needed before any calibration can fully compensate.

Why do I need to remove propellers for CompassMot calibration?

The autopilot will spin motors at high throttle during the calibration sweep. With props attached, the drone would attempt to take off or thrust against its restraints, potentially injuring you or damaging the airframe. Props off, drone secured, motors free to spin — that's the safe configuration for the calibration.

Can I run CompassMot indoors?

Yes — the calibration is about modelling current-induced bias, not about reading the Earth's magnetic field. Indoors works as long as you're away from large metallic structures (steel beams, reinforced concrete) that would add a different bias source confusing the calibration result.

How often should I recalibrate?

After any change to the power harness routing, after any motor or ESC replacement, and after any structural change that could shift the compass relative to the power leads. A clean build doesn't need periodic recalibration if nothing has changed, but checking the COMPASS_MOT magnitude after major maintenance is good practice.

Detecting and Calibrating Compass Motor Interference in ArduPilot

TL;DR: Compass motor interference is current-induced magnetic bias from ESCs and power leads that distorts the magnetometer reading as a function of throttle. ArduPilot’s COMPASSMOT calibration learns the relationship and stores it in COMPASS_MOT_X/Y/Z (mGauss per amp at full throttle) so the EKF can subtract the expected interference. COMPASS_MOTCT selects whether the compensation tracks throttle (1) or current (2, more accurate). After calibration, XKF4.SM versus RCOU output should no longer correlate — if it still does, physical separation between compass and power leads is the next step.

What current does to a magnetometer

A current-carrying wire generates a magnetic field proportional to the current. The field falls off with distance, so a compass mounted close to high-current power leads or ESC outputs sees a bias that changes with throttle. For high-current rigs (heavy multirotors, large agricultural drones) the bias can be tens of mGauss — comparable to the Earth’s magnetic field itself.

The bias is repeatable for a given current draw, so the autopilot can model and subtract it. COMPASSMOT is the calibration that learns the model.

The parameters CompassMot writes

COMPASS_MOTCT — compensation type. 0 = disabled, 1 = throttle-based, 2 = current-based. Current-based is more accurate but requires a current sensor.
COMPASS_MOT_X, COMPASS_MOT_Y, COMPASS_MOT_Z — bias coefficients in mGauss per amp at full throttle (current mode) or mGauss at full throttle (throttle mode), per axis. The autopilot multiplies by the current and adds to the raw compass reading.

For a healthy build, the magnitude of the COMPASS_MOT vector after calibration is below 60 mGauss; above 100 mGauss means physical separation between the compass and the power leads is needed before any calibration can help.

The CompassMot calibration procedure

Remove propellers. The autopilot will spin motors during calibration. Props would cause real motion.
Secure the drone to a heavy table or sandbag. Motors at high throttle without props still produce torque.
Disable arming checks for the calibration only. The compass will read wildly during the run — if you don’t disable, the autopilot will refuse to enter the calibration mode.
Connect to Mission Planner: Setup → Mandatory Hardware → Compass → CompassMot. Choose throttle-based or current-based at the dialog.
Spool throttle slowly from minimum to maximum over 10 seconds and back. The autopilot samples compass reading against current/throttle at every point and fits a linear coefficient per axis.
Confirm the saved coefficients. The dialog reports the magnitude. If it’s below 60 mGauss, you’re done. If it’s between 60 and 100, you have residual interference but the calibration will partially compensate. If it’s above 100, physical re-routing of wires or relocation of the compass is required.
Re-enable arming checks. Forgetting this step is the most common post-calibration mistake.

Reading the result in the log

Plot XKF4.SM (the EKF’s magnetometer test ratio) versus RCOU.C1–C4 sum during a normal flight:

Before calibration, XKF4.SM rises during high-throttle segments.
After successful calibration, XKF4.SM is uncorrelated with throttle — throughout the throttle envelope it stays near 0.
If XKF4.SM still tracks throttle after calibration, the calibration is insufficient and physical re-routing is required.

Confirming it in MAVExplorer

MAV> graph XKF4.SM
MAV> graph RCOU.C1 RCOU.C2 RCOU.C3 RCOU.C4
MAV> param show COMPASS_MOTCT COMPASS_MOT_X COMPASS_MOT_Y COMPASS_MOT_Z

The parameter values give you the calibration that was active at flight time. If the values are zero but throttle-correlated interference is visible in XKF4.SM, calibration never happened (or was reset).

Throttle versus current mode — which to use

COMPASS_MOTCT = 1 (throttle-based) is the older method. It assumes total current is proportional to total throttle, which is approximately true in hover but breaks down in aggressive flight. Use this only if you don’t have a current sensor.

COMPASS_MOTCT = 2 (current-based) is the better method when a current sensor is available. It compensates against actual measured current, which tracks reality more closely during transients. Almost every modern build has a current sensor in the power module — use current-based.

Per-motor compensation for difficult builds

On rigs where one motor’s power lead routing is particularly close to the compass, a global compensation model misses the asymmetry. ArduPilot supports per-motor compensation via the COMPASS_PMOT_* parameter family, calibrated similarly but with per-motor coefficients. This is advanced setup; the global model is sufficient for most builds.

When CompassMot isn’t enough

If after a successful calibration XKF4.SM still climbs with throttle, the cause is one of:

Compass too close to power leads. The simplest fix is mechanical: a 50–100 mm mast separating the compass from the main power harness usually solves this.
External compass not selected as primary. Check COMPASS_PRIMARY — an internal compass mounted directly above the FC almost always picks up FC-generated interference. Move to an external GPS+compass module on a mast.
Heavy current-carrying loop near the compass. Twisted-pair power wiring radiates much less than parallel runs. Re-route the main +/− pair to be tightly twisted.
Soft-iron or hard-iron contamination of the airframe. Steel hardware near the compass can magnetise. Use brass or stainless fasteners within 50 mm of the compass.

When this isn’t actually motor interference

Stale compass calibration. Compass offsets stored in COMPASS_OFS_X/Y/Z are independent of the motor compensation. A wrong offset produces variance that doesn’t correlate with throttle — that’s a calibration problem, not interference. See our compass variance post.
External magnetic structures. Flying over rebar, power lines, or vehicles produces position-correlated variance, not throttle-correlated variance.
Failing magnetometer chip. Random spikes uncorrelated with anything point at hardware degradation.

When LogHat helps — and when it doesn’t

LogHat overlays XKF4.SM against summed RCOU output and reports the correlation coefficient — high correlation means motor interference is still significant after whatever calibration was active. The parameter snapshot shows the COMPASS_MOT_X/Y/Z values in force at takeoff. What we can’t do is run the calibration for you — that’s a bench-top process with the airframe physically secured.