Key Takeaways
- BLDC control is a timing problem first, because useful torque only appears while the inverter feeds the two phases whose back EMF sits on its flat top.
- Hall sensors resolve position into 6 discrete states and sensorless schemes read the floating phase instead, and each approach fails in ways a model makes visible.
- Torque ripple concentrates at the commutation points, so timing and current shaping remove more of it than any mechanical change to the machine.
Good BLDC control is a timing problem before it is a control problem, and the timing is set by where the rotor actually sits.
A brushless DC machine makes steady torque only while the inverter energizes the two phases whose back electromotive force sits on its flat top, so everything depends on knowing rotor position well enough to switch at the right instant. Electric motor systems were responsible for 53% of global electricity consumption in 2023, and most of them are commutated electronically. That makes commutation logic one of the highest leverage pieces of code in the drive.
Equations get you part of the way and waveforms get you the rest. Once Hall signals, phase currents, back EMF and torque share one time axis, the reason torque dips at every commutation becomes something you can point at rather than an abstraction. That’s the difference between knowing the 6 step sequence and understanding it.
Trapezoidal back EMF shapes every choice in BLDC control
A BLDC machine is wound so each phase back EMF rises to a flat plateau and holds it for roughly 120 electrical degrees. Feed constant current into a phase during that plateau and its torque stays constant. Every other decision in BLDC control follows from that shape.
Picture a 4 pole machine at 3,000 rpm. One mechanical revolution covers 2 electrical cycles, so an electrical cycle lasts 10 ms and each of the 6 commutation sectors lasts about 1.67 ms. Two phases carry current in series through a sector while the third floats.
The plateau is why the control law stays simple. You don’t need a continuous position signal, only 6 boundaries, because torque per amp is close to flat between them. It’s also why the scheme suffers on a machine wound for sinusoidal back EMF, where square current blocks make torque wobble at 6 times the electrical frequency. Confirming that shape on your own machine is the first thing worth doing.
How 6 step commutation energizes 2 phases at once
Six step commutation walks the 3 phase inverter through 6 switching states per electrical cycle. In each state one phase ties to the positive rail, one ties to the negative rail, and one is left open. Current flows through 2 windings in series, and the open phase is where sensorless schemes listen.
Tracing one revolution by hand makes the pattern stick. With Hall outputs labelled A, B and C, the code 101 puts current into phase A and out of phase C while B floats, and 60 electrical degrees later the code becomes 100 with C carrying the return. Torque magnitude is set separately. The commutation table decides which transistors conduct, while PWM on the high side devices sets the average voltage across the conducting pair. Keeping those jobs apart pays off, since you can change the carrier without touching the sequencing and inject a timing error without disturbing the current regulator.
Hall sensors report rotor position in 6 discrete states
Three Hall effect sensors spaced 120 electrical degrees apart produce a 3 bit code that updates 6 times per electrical cycle. Each code maps straight onto a commutation state, so the position decoder is a lookup table. The codes 000 and 111 are invalid and point to a failed sensor or broken wiring.
Sensor placement error shows itself immediately in a model. Shift one sensor 5 mechanical degrees on a 4 pole machine and you have moved it 10 electrical degrees, delaying 2 of the 6 commutations and leaving the other 4 alone. Phase current goes asymmetric and the torque trace picks up a once per revolution beat on top of the normal 6 per cycle ripple. That fault hides well on hardware and stands out in a model, which makes it a good teaching case. The error arrives as structure rather than noise, and the structure names the sensor.
Sensorless commutation reads the floating phase back EMF
Sensorless 6 step control drops the Hall sensors and watches the terminal voltage of the floating phase. That phase back EMF crosses zero exactly halfway through the sector, so detecting the crossing and waiting 30 electrical degrees gives you the next commutation instant.
The weakness appears as soon as you model a start. Below a few hundred rpm the back EMF is too small to pick out against PWM switching noise, so there is nothing to trigger on. Practical drives run open loop through a forced alignment and ramp, then hand over once amplitude is sufficient.
Watching that handover succeed at one ramp rate and fail at another teaches more than a state diagram. SPS lets you probe the floating phase voltage, the freewheeling current and shaft torque in one run.
Filtering is a tradeoff you can’t avoid. A heavier low pass filter cleans up crossing detection but adds phase lag that grows with speed, so a filter tuned at 1,000 rpm commutates late at 5,000 rpm. A speed sweep exposes the lag.
Why torque ripple peaks at every commutation instant
Torque ripple in a 6 step drive concentrates at the 6 commutation points, where current has to leave one winding and build in another. Winding inductance stops both from happening instantly, so briefly the machine carries the wrong current pattern and shaft torque dips.
Five contributors show up in a careful model, and they stack rather than cancel.
- Winding inductance limits how fast current falls in one phase and rises in the next.
- The back EMF flat top on a physical machine is shorter than a clean 120 degrees.
- PWM carrier ripple rides on phase current and adds a high frequency component to torque.
- Hall placement error moves individual commutations early or late and skews the pattern.
- Cogging torque from slot and magnet interaction adds a position dependent term independent of current.
The notch grows with speed, because the voltage left across the winding sets the current slew rate and back EMF eats into it as speed rises. A drive that looks clean at 500 rpm shows a much deeper notch near base speed.
“Torque ripple in a 6 step drive concentrates at the 6 commutation points, where current has to leave one winding and build in another.”
Practical ways to cut torque ripple without new hardware
Most torque ripple reduction on an existing machine comes from timing and current shaping rather than mechanical change. Advancing the commutation point, raising the PWM carrier and tightening the current regulator each remove part of the notch.
The stakes reach past shaft smoothness. Cost effective efficiency measures across electric motor systems could save around 25% of their electricity use, cutting total global electricity use by about 10%, and commutation losses sit inside that total.
| Ripple reduction move | What changes in the waveform | What it costs you |
| Commutation advance angle | Switching earlier lets current build before the plateau begins, filling part of the notch. | The right advance moves with speed and load, so a fixed value underperforms. |
| Higher PWM carrier frequency | Current ripple between switching edges shrinks, so high frequency torque ripple drops. | Switching losses rise and thermal headroom gets tighter. |
| Faster current regulator | The regulator drags the outgoing phase down and the incoming up harder. | Bus voltage caps the slew rate, so extra gain stops helping. |
| Shaped current overlap | A deliberate overlap between outgoing and incoming phases smooths the handover. | The shaping table must be tuned per machine. |
| Sinusoidal control on a suitable machine | Continuous current and position feedback remove the 6 step notch almost entirely. | You need sinusoidal back EMF, a position sensor or observer, and more processing. |
Ordering matters more than the list. Raising the carrier is cheap, so it goes first, while advance angle costs nothing in hardware and gives the largest single gain. Shaped overlap takes the remaining effort, and it’s the one that genuinely needs a model.
What an editable BLDC model actually teaches you
A BLDC model earns its place when you can open the commutation logic, the inverter and the machine, change one parameter and follow the consequence through all three. Understanding arrives through the link between a timing decision and a torque trace.
Put a student in front of a working 6 step model and ask them to break it. Move a Hall sensor 10 electrical degrees, halve the carrier frequency, double the winding inductance, then look at the torque. SPS SOFTWARE suits that work because the component models are open and editable, the parameters are visible rather than buried, and everything sits inside MATLAB and Simulink, so a model built for a lab session extends into a research study.
“Understanding arrives through the link between a timing decision and a torque trace.”
None of this removes the need for careful measurement on the bench. What it does is shorten the distance between a hypothesis and an answer, and it stops the 6 step sequence from being a table students memorize. Engineers who have watched the notch move as they trimmed the advance angle carry that judgment into every drive they build.
