6. Motors and mixing

The mixer

The pilot or autopilot asks for four things: throttle (up/down), roll, pitch and yaw. The mixer turns these into one command per motor.

For each motor at arm position \((x, y)\) (right, forward) with spin \(s\) (+1 for CCW, −1 for CW), the command is roughly

\[ u_i = \text{throttle} \;-\; x_i\,\text{roll} \;+\; y_i\,\text{pitch} \;+\; s_i\,\text{yaw}. \]

Each command must stay between MOT_SPIN_MIN and MOT_SPIN_MAX. If the sum doesn't fit, ArduPilot (and AnaTele's simulator) keeps roll and pitch, because they keep the drone upright, and gives up yaw first.

Try it: push the roll slider until one motor hits a limit.

Healthy motors

In the nominal flight all four commands move together around a mean of about 1383 µs while airborne. They spread apart briefly in turns and accelerations.

A failing motor

In the motor-failure scenario, the simulator makes motor 3 lose thrust. Its health falls as \(h_3 = \max(0.3,\; 1 - 0.06\,(t - t_f))\) from \(t_f = 40.85\) s. The autopilot doesn't know this. It only sees the drone tilting the wrong way, so it commands motor 3 harder and the opposite motor (4) softer.

Read this the right way round

The motor with the highest command is often the weak one. The controller is trying to make up for missing thrust. A log records commands, not thrust, so AnaTele says "command anomaly", never "motor 3 failed".

Saturation

A motor is at ceiling if its PWM is within 2 µs of the ceiling, and at floor if within 2 µs of the floor. The limits come from the log's own parameters:

\[ \text{floor} = \text{PWM}_\text{min} + \text{SPIN}_\text{min}\,(\text{PWM}_\text{max} - \text{PWM}_\text{min}) \]

\[ \text{ceiling} = \text{PWM}_\text{min} + \text{SPIN}_\text{max}\,(\text{PWM}_\text{max} - \text{PWM}_\text{min}) \]

With 1000–2000 µs and spin limits 0.15/0.95, that's 1150 µs and 1950 µs.

Before the event, motor 3 spends 31.6 % of the airborne samples at its ceiling. Its longest unbroken stretch is 2.2 s. The motor-ceiling-saturation rule needs at least 1 s, so it's triggered. No motor stays at the floor for 2 s, so motor-floor-saturation-pre-event is not triggered.

Imbalance

AnaTele also averages each motor's command over the pre-event window (as a fraction of the PWM range). In the failure flight the means are M1 0.47, M2 0.47, M3 0.65, M4 0.34: a spread of 0.31, above the 0.12 screening limit. motor-output-imbalance-pre-event is triggered. Wind, payload offset and yaw demand can also cause imbalance, and the report says so.