Simplifying sensorless motor control

In my previous article, we tackled a passive sensorless control algorithm that calculates flux angle from back-EMF signals inside the motor. The algorithm allows us to eliminate the shaft sensor in a field-oriented motor control system. But back-EMF observers suffer from several maladies which are common to sensorless techniques:
 • The angle accuracy from a back-EMF observer (or any observer for that matter) is only as good as your estimate of the motor parameters. Even if you are lucky enough to find a datasheet for your motor, the parameters you need are often difficult to find, in wrong units, or missing altogether.
 • Even if you find the motor parameters, they may change due to motor heating.
 • The amplitude of the back-EMF signals is proportional to motor speed. So at zero speed, there aren't any back-EMF signals. Since you need back-EMF signals to spin the motor, but the motor needs to be spinning before you have back-EMF signals, how do you get the whole process started??
 • Once the motor is running, low-speed performance is usually very poor because the back-EMF signals are so small. ADC resolution and measurement tolerances become critical at low speeds.

Fortunately, tools are available to address these issues. Texas Instruments, for instance, developed a suite of tools called InstaSPIN-FOC that is designed to get your motor spinning instantly; sometimes in a matter of minutes, and supports sensorless designs. InstaSPIN-FOC has already been used to control motors ranging from 3W to 3 Megawatts, as well as speeds from 0.4Hz to over 4000Hz.

Let's go through a typical example with TI's tools to show how you would use this technology. The first tool you will probably use is the MotorID tool. This interrogates the motor to automatically extract all of the parametric data required by InstaSPIN to do sensorless control.

Considering the wide range of applications, you will probably need to provide a little guidance in the "Identification Settings" section when using MotorID. But once you start the process, it only takes a minute or two in most cases for the tool to learn the motor parameters. Assuming you have reasonable interrogation parameters and don't have a difficult motor (like a motor with extremely low inductance), the motor parameter estimates are usually very accurate. These parameters are not only supplied to InstaSPIN-FOC's sensorless observer, but also are used to automatically set the P and I values for InstaSPIN's PI current controllers, as described in my blog series on PI Tuning.

Once you have the parameters you can do sensorless FOC with your motor. However, in order to regulate motor speed, you still need to tune the speed loop. The most critical piece of information you need is system inertia, which is often difficult to obtain. One solution is to use InstaSPIN-MOTION to find inertia. This is a separate algorithm developed by LineStream Technologies that is bundled with InstaSPIN-FOC on many C2000 microcontrollers (MCUs). Alternatively, you can simply start with low values for P and I, and gradually increase them until you are satisfied with the speed transient response.