Filters in control systems

[sundin]

Hello all,
I have a question that may seem simple to some of you, but one that I have been struggeling to find the answer to.

Its basic control systems. How are filters used in a controller? I know that the difference between the feedback and reference signal is the input to the controller, but I cannot understand how filters (using gain and phase) are used to regulate the input to the "plant" based on this difference.

I tried to use the example of a simple dc motor as a plant to understand the concept, but I am still at a loss.

Can someone please explain.....
thank you.

 

[djs]

Hello;
One way that filters are used is to reduce noise in the input signals. For example filtering out 60 Hz noise generated from AC lines.

 

[nbucska]

You want as narrow band as possible to reduce the noise
but the LP filter causes phase lag, reducing the phase
margin -- so as usual in engineering, this is a compromise
too.

<nbucska@pcperipherals DOT com> subj: eng-tips
read FAQ240-1032

 

[jrs50125]

First of all, filters (any combination or resistors and capacitors) are not simply used to remove noise. More importantly in controler design, they're to provide a desirable output from the plant (obviously). However, you seem to be struggling with how phase and gain margin are used in controller design. To get a good grasp of this concept you first need to be familiar with Bode Plots, since this is where these concepts come. But simply stating it, by adjusting the phase and gain margin (modifing the values of R & C) you can directly change properties of the output such as:
rise time (time it takes to reach desired output),settling time (time it takes for the plant to stop oscilating), %overshoot(how far output is above or below desired value, and various other values related to the time domain output.

 

[laundry]

Hi sundin
It depends on what filters we are talking about? high pass and band pass filters are all to clean up signals in varied frequencies thats it.othere tipes are inductors these avoid resonance amplification of the harmonic currents generated by invertors,and to protect pf capacitors from overloading.also the reactance of series reactors is normaly chosen so that the tuned frequency of the circuit falls below 5th harmonic which is the lowest harmonic generated by ordinary 6 pulse convertors.

 

[johnwiss]

The filter in conjunction with the feedback makes a transfer function that has the form
Y=G(s)H(s)/{1+G(s)H(s)}
where we are forming a control system with a plant H and a "filter" or "loop filter" G. The expression is the Laplace transform of the (usually) differential equations that describe the output of the loop that is attempting to generate an output that when subtracted to the input to the loop produces an error signal that the loop attempts to force to zero for a class of input characteristics (like a step, ramp, parabola) relationship. For example if we are talking a phaselock loop the input would be the phase error of a received signal to the desired constellation point in PSK digital communications.

G might be as simple as a gain factor constant, or it may have a differential equation associated with it. We want to choose an H to produce a desired response (Y) to a class of inputs we need to design the loop to "track"

You need to understand feedback control systems and do a little work and search on the web or god forbid look in a textbook. It took me about 15 seconds to find this link:

http://homepage.mac.com/sami_ashhab/courses/control/lectures/lecture_22/Lecture_22.html

that explains the basics for a single-input, single-output linear loop that allows simple analysis.

The math is Laplace transforms, linear differential equations, and some calculus to invert Y(s) in my example when excited by a step, a ramp, or a parabola (U(s), 1/s, 1/s^2) using the product of excitiation and transfer function Y and then typically breaking the new TF into partial fractions and using inversion rules (table of transforms) or maybe Bode, or root-locus approaches to understand how to select the loop filter G to make the control loop track changes in the parameter to be controlled by motion, oscillator errors (in the case of PLLS) in a desired fashion (like settle to an error of so many % in a specified number of seconds, or symbols (in a comm. system)


It seems like people don't even bother using Google or technical search engines (like citeseer) using "tutorial" or "introduction" to get a basic understanding of the topics in this newsgroup.

Sundin--get a controls ebook off the internet or go buy or borrow one and look at simple examples and understand them-- then you can ask guys who do this sort of thing as part of our job regularly (like me) more specific questions that will help your understanding.