AFPM Motor Controller Tuning

Use phase-neutral equivalent Rs and differential Ld/Lq for the operating region. Record winding connection, temperature, test frequency, rotor angle and bias current. A line-line reading needs the correct winding conversion; do not paste it as a phase value.

Continuous phase current is RMS. The peak limit is phase peak and must be at least √2 times continuous RMS. The amplitude-invariant dq calculations use peak current; a controller with another transform convention needs conversion.

Only after checking scaling and algorithm conventions. Kp is V/A and Ki is V/(A·s). Per-unit current/voltage bases, speed units, sample time, anti-windup and vendor register scaling are separate inputs to the implementation.

The displayed Ki/fPWM is for an Euler accumulator with one current update per carrier period. A trapezoidal integrator uses Ki·Ts/2 times the sum of consecutive errors. Incremental PI forms combine proportional and integral coefficients differently.

For the linear decoupled plant 1/(Ls+Rs), the PI zero cancels the electrical pole, giving a first-order ideal closed current loop. Sampling delay, saturation, parameter drift and imperfect decoupling limit the real bandwidth.

No. It is a target. For an ideal cancelled plant with a pure delay of 1.5 PWM periods, phase margin is approximately 90° − 540° times the bandwidth/PWM ratio. At 0.10 this is 36° before other dynamics. Measure actual timing and response.

The half-duty estimate ΔIpp = Vdc/(4·Lq·fPWM) is divided by √2 times continuous phase RMS current. It is peak-to-peak ripple as a percentage of continuous phase peak, not RMS ripple or a magnet thermal limit.

First check inverter frequency limits, switching loss and thermal headroom. The frequency is calculated for an illustrative 12% ripple target. A designed series choke, a lower bus voltage or another inverter may be more practical; include their losses and integration cost.

It is the estimated fundamental peak phase-voltage error magnitude (4/π)·tdead·fPWM·Vdc for an idealized current-polarity model. It is not sixth-harmonic voltage amplitude. The duty percentage is time lost relative to one PWM period, not electrical phase lag.

For the dq equations used here, add vd_ff = −ωe·Lq·iq and vq_ff = +ωe·(Ld·id+ψf) to PI voltage outputs. A different Park transform or summing convention changes signs; verify the firmware equations first.

Electrical angle error is pole pairs times mechanical angle error. At 20 pole pairs, 0.1° mechanical gives 2° electrical. Acceptable error depends on the torque and unintended-Id budget rather than a universal tolerance.

No. The sign of unintended Id depends on offset direction and transform convention. The table reports its absolute magnitude. Its cosine torque approximation excludes reluctance torque, saturation and harmonic flux effects.

Follow the motor/inverter alignment procedure using a known stationary stator vector, verified phase sequence, an unloaded guarded shaft and conservative phase-current limits. Validate polarity and sensor direction at low speed before loading. Back-EMF reference angle depends on the transform convention.

The model uses Kp = J/(a·TΣ·Kt), Ti = a²·TΣ and Ki = Kp/Ti. TΣ is assumed to be 1/ωcc + 2Tpwm, including inner current-loop response and illustrative transport/feedback delay. Actual speed task periods, filtering and resonances can require much slower tuning.

Back-EMF provides no position signal at zero speed. High-frequency injection needs detectable saliency and a validated estimator; near-unity Lq/Ld makes it harder. Assess loaded starts and fault recovery, or use position feedback where required.

The selected rated point at Id = 0 requires more fundamental phase voltage than Vdc/√3 in linear SVPWM. Reduce speed or load, review a hardware-rated bus change, or solve an approved field-weakening operating point. Feedforward cannot create extra bus voltage.

No. Ich is the current for complete d-axis flux cancellation in a linear model. It does not determine safe negative Id, loaded speed range or irreversible demagnetization. Partial weakening can still be possible when Ich is above the current limit.

It is unknown without the magnet grade, temperature-dependent demagnetization curves, geometry and reverse-field operating point. Stator coolant temperature is not rotor magnet temperature. Obtain the manufacturer’s Id-versus-temperature limits and validate rotor heating.

No. ADC selection needs current range, shunt/gain, noise, effective resolution and sampling windows. Bearing-current mitigation needs the actual common-mode circuit, cable and shaft measurements. Inductance alone cannot select either component.

Use the static method and commissioning checklist below, collect phase parameters and inverter timing, and request a manual review. Default results are illustrative only. With JavaScript enabled, invalid fields pause results; reset defaults to recover, then copy or download a spec for review.