[sdiy] standard filters from ladder topology?

[Magnus Danielson]

From: "JH." <jhaible at debitel.net>
Subject: Re: [sdiy] standard filters from ladder topology?
Date: Mon, 5 Sep 2005 02:01:29 +0200
Message-ID: <002c01c5b1ac$f8c071a0$0200a8c0 at jhsilent>

> > For that application I think you will enjoy using either a Butterworth or
> > Bessel filter. The Butterworths maximum flatness keeps the passband free,
> but
> > the group delay is not really optimum. The Bessel/Thompson is much better
> in
> > this respect with a fairly quite passband, but benefits in maximum
> flatness of
> > group delay, which would show up as a slightly added delay to that of the
> BBD,
> > so that is certainly usefull. If you care about phases, Chebyshev variants
> is
> > not for you.
> 
> I read about some intermediate filter types in that AD book on opamps from
> last year.
> Something that has better phase response than Bessel in the pass band, and
> a slope almost as good as Butterworth in the stop band. I don't remember the
> name, but the idea was not to design for the same goal over the whole range.
> Instead, optimise for phase up to some frequency, then allow for some
> gradual
> attenuation till the -3db point is reached, and finally make a steeper slope
> your main goal.

Well, then you have left the world of cookbook filters (Where Butterworth,
Bessel/Thompson, Gauss, Chebyshev, Cauer, Elliptic and Legendre are house-hold
names (at least in THIS house-hold)) and gone into filter-synthesis where maybe
Least-Square Error or Pade approximations is betterly used. Only maybe you
where refering to Cauer/Elliptic filter where one toss in a few zeroes at the
right places to make life a little more interesting.

> I have no idea if this is really useful either, but probably worth
> experimenting.

Filters are all about finding a usefull compromise between competing tasks
making the effort impossible at best, and even then a brain-buster to solve.
Finding a suitable approximation takes knowledge of tha actual problem and a
good knowledge of filters and filter approximations. Actually, they are all
approximations one way or the other, so you need to know what the approximation
attempts to acheive and choose accordingly.

Naturally, you don't want to mix up your first-degree Butterworth and Chebyshev
filters of the same cut-off frequency, that has caused major havoc at many
universities where such sloppiness have been recorded. ;O)

Cheers,
Magnus