All pass filters?

[jhaible]

> I think there's a difference between time delay and phase shifting. 
Phase
> shift is "time delay" *within a specific frequency band*.  Time delay as
> it's implemented in, say, a BBD delay line, is constant across all
> frequencies.  Phase shift will result in different time delays at
different
> frequencies.  

This is true, but ...

I think it's fatal to one's understanding to discuss time
> delay and phase shifting as though they're related.  Phase shifting is
> studied in terms of frequency domain analysis while time delay is not.

... but it's actually not that far from each other. We're speaking about
"Gruppenlaufzeit" (is that "Propagation time; probably not exactly) in
analogue filters, and while a pulse in time domain is certainly distorted
due to the different delay times for its spectral parts, you still notice
a certain "mean" delay. OTAH, using pure delay units together with basic
math
functions is the one and only way to create a filter function with digital
signal processing.
But I don't want to split hairs. The reason for posting this is a question:
Has someone from the list already built a "theda processor", i.e. combining
the benefits of all pass stages and BBD lines to get a more musically
distributed comb filter function than with pure delay? There was an EN
article about that, and it was a combination of many fixed all pass stages
and a variable BBD delay. This produces much more notches than a pure
phaser method, and doesn't have the "metallic" character of linear spaced
notches that you get from BBDs.

> Consider this: theory says you'll never be able to phase shift a fourier
> frequency more than 180 deg in either direction.

This I don't understand. Can you be more specific ?

JH.