digital filter library?

[bill...@alum.mit.edu‏]

I went looking for a digital filtering library for Tcl and haven't
found one in ActiveTcl, by means of wiki searches, or just general
Googling. snack has a "filter" function, but in addition to some
specialized subcommands what it does is apply a filter whose
coefficients you already have computed. I was hoping for something
higher-level, e.g. "bandpass with center frequency f and
and bandwidth w" and that sort of thing. Is there something I have
missed? (And yes, I know that I could use swig etc. to wrap a C
library - what I want to know is whether there is something ready to
go.)

Bill

[David Gravereaux‏]

A generalized FFT library hooked into Tcl? I don't know of any. If you find one,
please post a link to your find as I could use one as I have an o'scope control
library and the next step for me is to add some analysis of the captured waveforms.

I looked at using VSIPL <http://www.vsipl.org/> but got a bit overwhelmed as I
should instead hit the book store on DSP topics instead :)

--
Sometimes I think the surest sign that intelligent life exists elsewhere in
the universe is that none of it has tried to contact us. -- Calvin

[bill...@alum.mit.edu‏]

Yes, except I don't actually need FFTs, so something like an FIR
design library would be sufficient.

[David Gravereaux‏]

bill...@alum.mit.edu wrote:
> Yes, except I don't actually need FFTs, so something like an FIR
> design library would be sufficient.
>

If you have some C code for it, I could whip-up an extension rather quickly.

--
I like to say "quark"! Quark, quark, quark, quark!
-- Calvin

[Kevin Kenny‏]

David Gravereaux wrote:
> A generalized FFT library hooked into Tcl? I don't know of any. If you find one,
> please post a link to your find as I could use one as I have an o'scope control
> library and the next step for me is to add some analysis of the captured waveforms.
>
> I looked at using VSIPL <http://www.vsipl.org/> but got a bit overwhelmed as I
> should instead hit the book store on DSP topics instead :)

If you're serious about that, and you're not shy of GPL, FFTW is the
only sensible choice. Damn fast, handles non-power-of-two sizes,
and pays serious attention to getting the DCT right. I use
it a lot for image processing.

If you *don't* want to go GPL, then I'd grab fftpack from
www.netlib.org, run it through f2c, and wrap the result. It's
not terribly fast (although for portable code, it's not
horrible), but at least it does non-power-of-2 sizes.

The trick to the whole thing will be an efficient representation
of the arrays. Howlett got a good start at that in BLT, but
his code didn't score high on the "plays nicely with others"
scale.

Now, if *you* run across a decent implementation of pseudo-log-polar
2DFFT, I need one. Do keep me in mind.

--
73 de ke9tv/2, Kevin