1
$\begingroup$

How these processes are different from simple IIR 1 order filtering, FIR filters in terms of amplitude and phase characteristics?

$\endgroup$
4
$\begingroup$

Yes, integration and differentiation can be linear filters. You can start from laplace properties that say:

$ \int_{0}^{t} {x(t)dt} \longrightarrow \frac{X(s)}{s} \\ \frac{d}{dt}x(t) \longrightarrow sX(s) $

So you can find transfer function of integration and differentiation:

$ H_{INT}(s) = \frac{1}{s} \\ H_{DIFF}(s)=s $

You can convert these transfer function in digital IIR filters for example by bilinear transform or others digitization techniques. However you should notice that $ H_{DIFF}(s) $ is not causal, then you must add a pole to the transfer faction far from useful signal frequency, and it begin:

$ H_{DIFF_{causal}}(s)= \frac{s}{\alpha s + 1}$ where $\alpha$ is the time constant of the derivative filter a small real $>0$.

Using bilinear transform $H_{INT}(z)$ begin a trapezoidal integrator, using Euler transofrm $H_{INT}(z)$ begin a rectangular integrator, you can see this difference and the digitization in the MATLAB PID page.

I wrote a simple PID program in C that computes these operations using Euler transform, a PID in closed loop doesn't need to be accurate so Euler works well.

In literature there are a lot of ways to implement derivative filters (IIR and FIR) that workaround in elegant way anti-causal problem, however in a lot of situations you can simply digitize the analog transfer functions to make IIRs and FIRs.

$\endgroup$

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.