Sensorless Control: When You Can Skip the Encoder

The question arrives as a drive question: does this application need an encoder, or will sensorless do. It is a system question, and by the time the motor and drive are on site most of the available options have already been decided.

What sensorless control is actually estimating

A vector-controlled drive needs to know where the rotor is in order to place the stator current correctly. An encoder tells it. Sensorless control infers it from the electrical quantities the drive already measures, which are the terminal voltages and the phase currents.

There are two distinct ways to do that inference, and they work in opposite halves of the speed range.

Method one: back-EMF, and why it fades

A turning rotor induces a voltage in the stator. That back-EMF carries the rotor angle, and a wide family of methods extracts it, from simple flux integration through to sliding-mode and extended back-EMF observers.

The weakness is structural. Back-EMF is proportional to speed. Halve the speed and the signal you are measuring halves with it, while the noise, the inverter’s dead-time distortion and the uncertainty in your stator resistance estimate all stay where they were. Keep going and the signal disappears into them. At standstill the rotor induces nothing at all.

No amount of observer design fixes this, because the information is not in the measurement. This is the reason a sensorless drive that behaves impeccably at 40 Hz can be unusable at 2 Hz.

Method two: injection, and what it needs from the rotor

The second family sidesteps the problem. Rather than waiting for the rotor to generate a signal, the drive injects one: a high-frequency voltage superimposed on the fundamental, with the rotor position read out of the current response.

This can work at standstill, where the first method has no back-EMF signal. It still depends on the motor presenting position-dependent saliency that the drive can excite and measure reliably.

The motor-drive combination needs usable saliency. The method reads position from an angle-dependent current response. Geometric rotor saliency is the clearest source, while saturation can create additional saliency in some designs. If the response is too small or too distorted for the drive to separate from noise and inverter nonlinearity, the theoretical signal is not enough.

Which is why this is a rotor decision

Rare-earth magnet material has a relative permeability close to air. On a surface-magnet rotor the magnets are effectively part of the air gap, so the rotor looks uniform to the stator field and the machine is non-salient. On an interior-magnet rotor there is iron between adjacent poles, so the magnetic path differs between the axes and the machine is salient. That difference is covered in more detail in interior versus surface permanent magnet rotors.

So the sentence “this drive supports sensorless operation” describes an available control function, not guaranteed installation performance. The motor, drive, parameter set and load have to be assessed together.

The handover is where the problems live

A drive covering the full range runs injection at low speed, a back-EMF observer at high speed, and switches between them somewhere in the middle.

That crossover deserves attention during commissioning. Both methods are estimating the same quantity by different physics, and they will not agree perfectly. A load that sits near the handover speed, or crosses it repeatedly, will exercise the transition far more than the factory test did. If an application spends its life near the boundary, say so at the enquiry stage rather than discovering the behaviour on site.

When the encoder stays

Sensorless has genuinely displaced encoders in a large share of general industrial drives. The cases where it should not:

• Full torque held at standstill, indefinitely. Injection can produce torque at zero speed, but holding a suspended load on estimated position is a different risk category from turning a pump.
• Position control, or speed accuracy tighter than the estimator’s error. Estimation error is a specification; ask for it rather than assuming.
• Overhauling loads that can drive the machine backwards through zero speed, where the estimate has to survive a sign change under load.
• Safety functions that require a verified speed or standstill signal, where an estimate is not an acceptable source.
• Hoists and anything where losing position means dropping something.

For a fan, a centrifugal pump, a compressor or a conveyor that starts light and runs above a few hertz, an encoder is usually cable, a coupling, and one more thing to fail.

What to ask, and in what order

1. What is the lowest speed the application must produce useful torque at, and must it hold torque at zero?
2. Is the load overhauling at any point in its cycle?
3. Given those, is a salient rotor required, and does the series being quoted have one?
4. What is the drive’s estimation error over the speed range that matters, and where is the handover?

Questions one and two are yours to answer, and they decide the other two.

On the motor side, rotor construction varies by series rather than by family. The GTYP high-speed permanent magnet series publishes an interior permanent-magnet rotor with a reluctance-torque component. That makes it a candidate for saliency-based estimation; it does not by itself establish zero-speed sensorless performance. Confirm the specific motor, drive, firmware and required load point before committing to the architecture. Apply the same check to every series in the permanent magnet range.

Related: Interior vs surface permanent magnet rotors · Why a VFD needs EMC design · Retrofitting an old VFD project: induction or PM