DSPRelated.com
Forums

PI implementation using IIR

Started by maanas August 4, 2009
hello everyone
i am new to this DSP world and stuck in a problem. could you please guide
me in understanding about PI controller and IIR filter implementations
similarity. i.e whether i can implement IIR for a PI controller.

my project : i have to maintain constant current through a load by
constantly monitoring it and adjusting the PWM signal to the switch. i was
thinking of implementing discrete PI  controller. somewhere i read about
IIR filters implementing as PID controller. please guide me.


On Tue, 04 Aug 2009 10:55:08 -0500, maanas wrote:

> hello everyone > i am new to this DSP world and stuck in a problem. could you please > guide me in understanding about PI controller and IIR filter > implementations similarity. i.e whether i can implement IIR for a PI > controller. > > my project : i have to maintain constant current through a load by > constantly monitoring it and adjusting the PWM signal to the switch. i > was thinking of implementing discrete PI controller. somewhere i read > about IIR filters implementing as PID controller. please guide me.
In it's purest sense a PI controller is an IIR filter, but a generic IIR filter implementation (such as a biquad) is rarely close to the best way to implement a PI controller, for a variety of reasons. Something about the way that you're using your terminology suggests that you are conflating "IIR filter" with something more specific, such as a biquad. An IIR filter is any signal processing element, generally single- input, single-output, whose response to a non-zero input is of infinite duration. It is generally understood to be a linear filter whose impulse response has components composed of decaying exponentials. A biquad is a specific implementation of a sampled-time IIR filter. A decent PI controller will include provisions to limit integrator windup, it will most likely be written so that the action of the code is easy to understand, and it may well have provisions for different proportional gains in the forward and feedback paths, to tune the response of the thing to commands vs. disturbances. None of these provisions exist in a biquad implementation. Go here to find an article about PID controllers: http:// www.wescottdesign.com/articles/Sampling/pidwophd.html. If you really want to learn this stuff I have a book: http:// www.wescottdesign.com/actfes/actfes.html. I hope this helps. -- www.wescottdesign.com
maanas wrote:
> hello everyone > i am new to this DSP world and stuck in a problem. could you please guide > me in understanding about PI controller and IIR filter implementations > similarity. i.e whether i can implement IIR for a PI controller. > > my project : i have to maintain constant current through a load by > constantly monitoring it and adjusting the PWM signal to the switch. i was > thinking of implementing discrete PI controller. somewhere i read about > IIR filters implementing as PID controller. please guide me.
If there is no inductance in the circuit, the current response is virtually instantaneous. If the driving voltage is PWM, the current will be too. So I assume that your load current is filtered somehow, and that filter is necessarily analog. You may not need the I part of a PI controller. Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������
On Tue, 04 Aug 2009 15:27:15 -0400, Jerry Avins wrote:

> maanas wrote: >> hello everyone >> i am new to this DSP world and stuck in a problem. could you please >> guide me in understanding about PI controller and IIR filter >> implementations similarity. i.e whether i can implement IIR for a PI >> controller. >> >> my project : i have to maintain constant current through a load by >> constantly monitoring it and adjusting the PWM signal to the switch. i >> was thinking of implementing discrete PI controller. somewhere i read >> about IIR filters implementing as PID controller. please guide me. > > If there is no inductance in the circuit, the current response is > virtually instantaneous. If the driving voltage is PWM, the current will > be too. So I assume that your load current is filtered somehow, and that > filter is necessarily analog. You may not need the I part of a PI > controller. > > Jerry
Switching MOSFETs (which I _assume_ is what the guy means, but who knows?) have some effective non-zero (and often nonlinear) resistance, so having an integrator in one's controller is often a good idea. Which just means that you have to make sure you're measuring current with adequate precision in the face of all the switching transients. So, OP: What do you mean when you say "switch"? -- www.wescottdesign.com