2
\$\begingroup\$

This is a control theory question, but because there is no control theory stack exchange, I thought this was the closest appropriate location:

I've been studying feedforward control system designs, specifically motion stages that use acceleration feedforward control. In this control system, a control signal is generated based on changes in the command signal (as opposed to relying solely on feedback control).

My question, what does feedforward typically do to the system response bandwidth? If done correctly, I would assume it increases the bandwidth? Say for example the feedforward controller is designed as the inverse of the plant, is there a proof that illustrates that bandwidth is increased?

Furthermore, most feedforward control setups combine feedforward control with feedback control. In this scenario, how does the bandwidth between a feedback system and feedback-feedfoward system compare? Assume feedforward control is used in the sense of creating a command signal that attempts to compensate for changes in the reference command.

\$\endgroup\$
  • \$\begingroup\$ Not sure, but I think robotics.stackexchange.com would also be a good place to ask. I personally agree that Control theory questions be asked here, however. \$\endgroup\$ – rrz0 Jan 3 '18 at 17:01
1
\$\begingroup\$

Control theory has a lot of terms that are used loosely by many people, myself included, so I'll start off with some definitions. Please correct me if I'm wrong or just talking about something different. The feedforward component of a system uses prior knowledge of the control system plus plant (output which is being controlled) to change the setpoint into the control system to get where you're going faster. So, if I have a poorly controlled oven, and I want it to hit 350 F quickly, I might set it for 400 F until I see that it is around 300, then change the setpoint to 350.

Depending on your system layout, the transfer function of the feedforward would be factored into the whole system transfer function, and therefore change the bandwidth. This can be a bit dangerous, since it can mean you are putting something with resonance in front of your component with filtration, so in real systems you need to make sure the feedforward is not going outside allowed parameters (eg I can't set my oven for 1000 F). But the theory is sound.

| improve this answer | |
\$\endgroup\$
1
\$\begingroup\$

Positive acceleration feedback is used to reduce the effective inertia in the system, thereby increasing \$\omega_n\$ and increasing bandwidth. Check this out with a simple 2nd order system, where positive acceleration feedback will reduce the \$s^2\$ coefficient in the TF denominator.

Feedforward control applies the plant’s disturbance signal earlier in the forward path (ie nearer the system input/output summing junction), thereby counteracting the effects. But requires that the disturbance is measurable or can be estimated, which is not always possible, particularly if the disturbance affects a part of the plant that is inaccessible to measurement.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ I'm unfamiliar with positive acceleration feedback, however, that seems to be different than what I refer to as acceleration-based feedforward control. In my example, the desired position will act as the disturbance. That is, when I change the commanded position, it will see this as a disturbance and attempt to eliminate the "soon-to-be" position error without waiting for a feedback signal. In this scenario, how would it affect the bandwidth? \$\endgroup\$ – Izzo Jan 3 '18 at 17:47
  • \$\begingroup\$ Can you show a block diagram? \$\endgroup\$ – Chu Jan 3 '18 at 18:28
  • \$\begingroup\$ I'm primarily going off of "Figure 8. Basic Feedforward and P.I.V. Control Topology" from compumotor.com/whitepages/ServoFundamentals.pdf \$\endgroup\$ – Izzo Jan 3 '18 at 22:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.