Dual to AllPass.?

Wouldn't it be nice. No, in general that violates causality. (I'll be a bit less dogmatic about that than previously, since we talked about this several months back.)

A symmetric FIR filter centered around t=0 might qualify, if you don't need it to be real-time.

Cheers

Phil Hobbs

I was hoping you had a change of heart ;-)

(I've decided, therefore, that I've been fed some bad data from a real device :-( ...Jim Thompson

Couldn't you make a lowpass filter, then fix its phase shift with an all-pass?

You could make an approximation of a constant-delay filter (like Phil's symmetric FIR). But you can't keep the phase constant here in the real world.

You can flatten it out, or fix it in some restricted region (as in your example from last year) but you can't get rid of it everyplace. Group delay != real delay, but it's the leading order approximation.

Cheers

Phil Hobbs

Jim didn't specify exactly zero phase, over an infinite frequency span; neither request would be reasonable. But it should be practical to make a lowpass filter, followed by an allpass network, that has minimal phase shift up to, say,

I doubt that very much. If you can do that to an LC-type lowpass, you'll put Tim Wescott and Co. out of business.

Cheers

Phil Hobbs

I didn't say LC. I was thinking 1st order, RC, because that case is easy to think about.

But it could be done to a higher-order filter, just more work.

Point is, that the answer to Jim's question is that, yes, it can be done within reasonable practical limits.

First order RC filters don't have real delay--you can cancel out a lowpass with a highpass, over whatever bandwidth you like, if you can stand the loss.

Cheers

Phil Hobbs