question for theorists: zeros & phase

[Magnus Danielson]

From: Martin Czech <martin.czech at intermetall.de>
Subject: RE: question for theorists: zeros & phase
Date: Mon, 27 Sep 1999 14:46:16 +0200 (MET DST)

> :::At work I have this picture stuck up in my room. The picture is from Bob Pease
> :::"Troubleshooting Analog Circuits". It's a Ziggy strip with a man visiting his
> :::doctor with a larchis animal chewing on his foot while the doctor concerned
> :::says "I'll have to run some tests, but I'm pretty sure that it's either an
> :::alligator or a crocodile." (the viewer can see the same). The point is, some
> :::things are not worth analysing further, since it doesn't help you in the
> :::problemsolving (in this case actually remove the alligator/crocodile from
> :::the poor mans leg). I think it applies as well to your phase trouble.
> 
> No, not quite. It's about testcases I wanted to use for my little program.
> Then the question arrises: what is the "real" phase response, and is
> the computed somwhat similar?
> 
> That's when the trouble started.
> 
> And, after all, FIR filters actually have some of their zeros planned
> to be on the unit circle.  All poles of a continous system get one zero
> on the unit circle at PI when doing bilinear transformation.

Right. But when do you pass PI (-1, 0)?
This is the Nyquist frequency to begin with. Also, if you want to know the
phase after you have passed PI (or whatever point you happend to have one
of these nasty things on) just take your analysis there. This is the neat
thing about the analysis tools given, you evaluate them for any given point
without caring about the results from any other analysis point.
So, you can go ahead and jump to some point beyound PI, which by the way
will sligthly above -PI, it is the same thing.

So if you only worry about how the phase changes as you pass these things,
don't bother to do it rigth at the singularity, just go ahead and then
invent some value to get a continous curve (if that is needed).

> They may get a bit off side due to parameter resolution.
> 
> So , when you try to design something you try to test it afterwards, and it would
> be nice to know how things should be (with theoretical values).

Certainly, but let's not forget the issue about whatever you are calculating,
it's meaning in real world and its applicability to your design problem.

So Martin, do you now know how to deal with this?

Cheers,
Magnus