sw/
← projects
Systemantics·2014–15

Field-oriented control of BLDC motors

Research internship at Systemantics, Bengaluru: comparing trapezoidal commutation with field-oriented control for small-drone BLDC motors on a Texas Instruments evaluation kit.

Problem

Small drones live on brushless DC motors, and how those motors are commutated decides how smooth, quiet and efficient they are. Trapezoidal commutation is simple but causes torque ripple at low speed; field-oriented control (FOC) promises smooth torque across the speed range at the cost of more computation. The internship was to understand both and compare them on real hardware.

What I did

At Systemantics in Bengaluru, over December 2014 and January 2015 under Dr Jagannath Raju, I studied the BLDC motor and its mathematical models and ran both schemes on a TI DRV8312 evaluation kit with the InstaSPIN GUI.

For trapezoidal commutation I swept the flux threshold that triggers each switch and showed commutating too early (0.16 pu), at the right time (0.30 pu) and too late (0.66 pu). For FOC I worked through the chain: Clarke and Park transforms, PI control of the d and q currents, the inverse transforms and space vector modulation. I also characterised the motor: from a 12 V supply, 4,000 rpm and 1.3 A at no load and 0.39 Ω winding resistance, I derived its constants and a damping friction coefficient of 0.024 N·m·s, and measured its no-load efficiency.

Result

Both schemes ran on the kit. Trapezoidal commutation was simple and efficient at high speed but caused torque ripple at low speed; FOC gave smooth motion at low speed, efficient operation at high speed and the most torque per watt, at the cost of higher algorithm complexity. Suggested next steps were higher-speed runs, dynamic loading, other motors, and measuring torque ripple on an oscilloscope.

diagrams

scroll sideways on small screens

Field-oriented control of a BLDC motorThe speed reference omega r goes to a speed regulator, giving the q-axis current reference; with the d-axis reference, current PI regulators output Vd and Vq. The inverse Park transform gives V alpha and V beta, and SVM or PWM produces phase voltages Va, Vb, Vc for the BLDC motor. Phase currents Ia and Ib go through the Clarke transform (a-b to alpha-beta) and the Park transform (alpha-beta to d-q), feeding Id and Iq back to the current regulators. A speed estimator uses I alpha, I beta, V alpha and V beta to estimate the rotor angle theta, used by the Park transforms, and the speed omega, fed back to the speed regulator.SpeedRegulatorCurrent PIRegulatorsInverse ParkTransformSVM/PWMBLDCMOTORα - βa - bd - qα - βSpeed EstimatorωrIqrIdrVdVqVαVβVaVbVcIaIbIαIβIdIqθVαVβω
controlField-oriented control: Clarke and Park transforms turn the phase currents into d and q components, two PI loops regulate them, and the inverse Park transform and space vector PWM drive the motor. A speed estimator supplies the rotor angle and speed.Source: the Systemantics internship slides, 2015
Comparison of BLDC commutation methods
Commutation methodSpeed controlTorque controlAlgorithm complexity
Low speedHigh speed
TrapezoidalExcellentTorque RippleEfficientLow
SinusoidalExcellentExcellentInefficientMedium
FOCExcellentExcellentExcellentHigh
comparisonHow the commutation methods compare on speed control, torque at low and high speed, and algorithm complexity.Source: the Systemantics internship slides, 2015

figures

click to enlarge

fig. 01Trapezoidal commutation at the right flux threshold (0.30 pu); switching at 0.16 pu was too early and at 0.66 pu too late.Systemantics internship slides, 2015
fig. 02The TI DRV8312 evaluation kit used to run both commutation schemes.Systemantics internship slides, 2015
esc