Velocity estimation accuracy increased 25%.

Improved velocity estimation algorithm based on digital halls increased the accuracy at all velocities with about 25%.

neptuneethercat-1.jpg

October 2015

In general digital hall sensors are used to estimate the velocity of brushless actuators. Typically the resolution of these digital hall sensors are the limiting the accuracy of the whole system when used in position control or velocity control. 

A common solution to improve the accuracy of the velocity measurement is to combine digital halls with a (digital) encoder, which is a costly practice. We decided to face the real cause of the accuracy problem: Improving the accuracy of the velocity estimation based on the digital halls. 

Our original algorithm to estimate the velocity is based on a moving average filter. Especially at low speeds, the accuracy of this algorithm is known to be limited. By changing the length of this filter, we are able to increase the accuracy of the estimation. However, increasing this length introduces a latency (for example 200ms). At fast changing target velocity, this latency is limiting the maximum acceleration and deceleration of the actuator.

With our newly developed algorithm, we combined this moving average filtering technique with an algorithm focusing on estimating the position at low speeds. With this combination we have successfully increased the accuracy at all velocities with about 25%, while minimizing the latency (just 65 ​ms). At low velocities the estimation is significantly more improved, making low velocity control applications based on digital halls possible.

The algorithm has been tested for 2 different applications. Once fully verified and tested, we will include it in a new firmware release, so everybody can benefit if this improvement.