Re: Xlating FIR filter delay

[Angilberto Muniz Sb]

Argh... so overloaded with information that I keep forgetting the basic stuff... Thank you for your considerations Marcus.

In the docs it says the xlating fir filter is a symmetric one, has linear phase shift and causes no phase distortion.

Lets say I have the following setup:

x(t) = cos(wt),

s1(t) = x(t) upshiffted to w1, s2(t) = x(t) upshiffted to w2

s(t) = s1(t) + s2(t).

Now, I apply two xlating filters to s(t):

r1(t) = s(t) filtered and downconverted to w

r2(t) = s(t) filtered and downconverted to w

I see a phase shift (or time delay) between x(t) and r1(t) or r2(t).

That´s the delay I´m trying de find in order to delay x(t) to align it to r1(t) or r2(t).

Angilberto.

On Thursday, December 19, 2019, 11:32:33 AM GMT-3, Müller, Marcus (CEL) <address@hidden> wrote:

Ah, no!

So, what you're seeing is no processing delay caused by our
implementation of the xlating FIR filter, but simply *group delay*,
which every linear system has.

I don't know your background, so it's a bit hard to recommend some
literature, but group delay would be covered in a "signals and systems"
text book.

> Guess I've found the answer:
> (len(filter_taps) - 1) / 2

That is true ONLY if your filter has linear phase. It has linear phase
iff it's symmetric to the center tap.

Best regards,
Marcus

On Thu, 2019-12-19 at 14:08 +0000, Angilberto Muniz Sb wrote:
> Yes Marcus, I'm interested in the "excess" delay, thats why I need to cancel the group delay.
>
> Guess I've found the answer:
>
> (len(filter_taps) - 1) / 2
>
> Will try it..
>
>
> Angilberto.
>
>
> Em quinta-feira, dezembro 19, 2019, 9:16 AM, Müller, Marcus (CEL) <address@hidden> escreveu:
>
> > Hi Angilberto,
> >
> > am I right to assume that you already are considering the group delay
> > of the filter that you specified and there's "excess" delay beyond
> > that?
> >
> > Best regards,
> > Marcus
> >
> > On Wed, 2019-12-18 at 19:59 +0000, Angilberto Muniz Sb wrote:
> > > Hi all,
> > >
> > > I have a dual tone signal that I down convert and split into two common frequency signals with the Xlating-fir filter. So far so good.
> > >
> > > However I noticed there is a phase shift between the signals generated (I assume this is due the filter delay). I have to compensate for that.
> > >
> > > The question: How to determine or estimate that delay?
> > >
> > > Thank you,
> > >
> > >
> > > Angilberto.