.

# Overview of Feedforward Tuning

Most torque systems can be modeled as a double integral system with coefficients for inertia (Kaff), viscous (Kvff) and static (Kff) friction. Although this model is not perfect, it is an effective way of reducing the following error. The feedforwards on the XMP were based on this model. The challenge lies in finding these three coefficients.

There are two major methods of feedforward tuning covered in this section: shape-based and measurement-based.

## Shape-based Feedforward Tuning

Shape Based Feedforward Tuning is based upon the following concepts:

 Inertial force will result in a position error roughly proportional to the commanded acceleration magnitude. (i.e. The position error and commanded acceleration magnitude plots have similarly shaped traces.) Velocity based friction will result in a position error roughly proportional to commanded velocity magnitude. (i.e. The position error and commanded velocity magnitude plots have similarly shaped traces.) Static friction will result in a position error roughly proportional to the sign of commanded velocity. (i.e. The position error and commanded velocity plots have similarly shaped traces.)

This means that if you command a trajectory, of which you know the acceleration and velocity (easy on the XMP), you can determine the required feedforwards by observing the magnitude of the commanded acceleration and velocity compared to the position error. By changing the feedforward coefficients and checking the change in position error you can deduce the ideal feedforward coefficients.

## Measurement-based Feedforward Tuning

The relationships described above also make it possible to use a similar technique—Measurement-based Feedforward Tuning. Since computers allow you to quickly quantify relationships between two variables such as, Command Acceleration and DAC Output, you don't have to iterate through an entire process to find the answer. You can simply calculate the correct feedforward value in one iteration.

 Legal Notice  |  Tech Email  |  Feedback        Copyright © 2001-2021 Motion Engineering