Enabling high-performance motor control at low speeds

Sensorless motor control has primarily been implemented in applications where the majority of the operating time is at higher electrical frequencies (mechanical speeds). This is due primarily to the fact that most sensorless techniques require a back-EMF (Bemf) signal that is generated by the rotor's rotation at a minimum frequency. Being able to continuously estimate the rotor flux angle at zero and very low speeds and stably move between low-speed and high-speed estimators can improve the effectiveness of sensorless start-up under load.

Texas Instruments' FAST software observer is used in the InstaSPIN-FOC software. FAST's minimum frequency of operation is much lower than that of other observers, sometimes below 1Hz. But it still requires a minimum frequency.

With sensorless techniques such as FAST, the initial rotor flux angle is unknown and, until enough Bemf is measured so that the algorithm can start estimating correctly, the estimates are unpredictable. But this estimated angle – even though incorrect – will be feeding the control system that will be applied to the motor and that may induce rotor movement. With just a small amount of rotor movement, though, enough Bemf voltage is produced so that the algorithm can converge on a reasonable angle estimate, allowing controlled high-torque drive into an area of excellent operation. So if enough torque is generated for rotor movement, this method can be used to start the motor, but it may not be consistent in start-up performance.

Generating enough torque
As the starting load is increased, the torque you can generate will be based on the current and the alignment of the fields (determined by the accuracy of the angle estimate). To insure you can generate enough current, it is imperative that the speed controller's maximum (positive and negative) output be larger than the rated current that is required to generate the rated torque. In the example in Figure 2, note the below waveform captured when starting a motor under full load. Producing the torque required to move this rated load requires 4A of current. In this case, the speed controller's maximum output was set to (6.0), and you can see that this 6A current was reached in the first electrical cycle to move the rotor. In this example, FAST was able to provide a valid angle, which allowed the control system to regulate the current usage immediately to only the required 4A.

Even though you are generating a stable feedback angle, that angle is not necessarily aligned properly to generate maximum torque. You are basically just sweeping a stator field and waiting for the rotor field to lock on and synchronise. When the stator field is not properly oriented, you will not produce enough torque or, in the worst case, produce torque in the opposite direction required. Improving this situation requires a better starting angle for the control system. But how do we do this when most sensorless control algorithms, including FAST, can't provide a valid angle at zero speed?

Alignment
One way to do initial alignment in a field-oriented control (FOC) system is to inject a DC current into the Id portion of the control system (none into Iq). This is the D-axis, which is defined as the orientation of the rotor flux.