.

Shape-based Feedforward Tuning

Shape-based Tuning has been broken up into four main sections:

 

 

Before You Start

You will need to tune the closed loop parameters (PID or PIV or etc.) before tuning the feedforwards. The closed loop tuning does not need to be perfect, but you will need to make several aggressive moves to optimize your feedforward coefficients.

Before attempting to turn the servo with Shape-based Feedforward Tuning, you will need to properly set up Motion Scope so that you can evaluate the motion with the appropriate data. You can easily designate which traces you wish to include on the motion plots by clicking the "Traces" button and selecting the following traces:

 
  •  Position Error (Ax n Pos Err)
  •  Commanded Velocity (Ax n TC.Velocity)
  •  Commanded Acceleration (Ax n TC.Accel)

It is also important to read the changes in position error from one trial to the next. In Motion Console, turn off auto scaling on the position error trace by clearing the "During Acquire" and "During Display" options under Traces Edit.

A Trust TA 9000 will be used for the following examples, unless otherwise noted (www.trustautomation.com).

What is a good starting move?
Make a move that is representative of a large, quick move in your application. As long as you are not clipping the DAC output or your drive, the faster the move, the easier it will be to interpret the resulting plot.

Which feedforward term do I start with?
It doesn't matter which feedforward term you tune first. You will often find that once you tune all the feedforward terms, you will want to readjust one or two of the terms; this is normal. However, it can get complicated with oscillations and systems that do not have a single dominant feedforward term.
Here are the general rules:
     - Add velocity feedforward if the position error plot looks like the velocity plot.
     - Add acceleration feedforward if it looks like the acceleration plot.
     - Add friction feedforward if it looks like the sign of the velocity plot.

For all the shape-based tuning examples, the integral will be set to zero (Ki = 0).
Ki = 0 because the integral will distort the shapes that we are looking for. This does not mean that feedforwards will not work when the integral is ON. It is just difficult to see the patterns when the integral is working effectively, since it will try to reduce the position errors with time. After you are done setting feedforwards, revert back to the original value of Ki.

For more information about the different types of shape based tuning, please see the following sections:

 

Previous | Next

       Legal Notice  |  Tech Email  |  Feedback
      
Copyright ©
2001-2010 Motion Engineering