RE: [linux-audio-dev] Software filter engines for high end audio

New Message Reply About this list Date view Thread view Subject view Author view Other groups

Subject: RE: [linux-audio-dev] Software filter engines for high end audio
From: Anders Torger (torger_AT_ludd.luth.se)
Date: ke helmi  16 2000 - 08:08:54 EST


On Wed, 16 Feb 2000, Richard W.E. Furse wrote:

> What is Warped FIR? Does any of this get near the speed of fast
> convolution?

Warped FIR is a special kind of filter that instead of having constant
width of the frequency bands, it can be controlled by a 'warp factor'. An
ordinary FIR filter has equal bands over the whole area. For 44.1 kHz, a
3000 tap FIR filter has bands that are about 13 Hz wide. However in room
equalisation applications, one is interested of having small bands in the
bass range, and wide in the treble. This is where WFIR is usable, you can
have very narrow bands in the bass range, and let the bands be wider in
higher frequencies. This way a 300 tap WFIR filter can be more effective
than a 6000 tap FIR filter when it comes to correcting magnitude
problems. The drawback with WFIR is that it is not feasible to make phase
corrections, and it requires more processor power per tap than a FIR
filter.

> I'm interested in an implementation of a very fast FIR implementation for
> my MN audio library, even if only for its low latency, however I really
> would want float in->float out. Us Linux folk really do need to agree a
> decent plugin API ;)

How fast is very fast? I would like an example. If we are going to make
floating point calculations internally, we will be very slow. However if
the demand for dynamics is limited, it is possible to do fixed point
calculation (integer) internally, and just convert to and from floating
point at the edges. The FIR filter implementation I have mentioned here
(again, which is not complete) is not a very smooth one, since it
recompiles itself when coefficients are changed. But it is damn fast :-)

> Incidentally, I recently read a very interesting and straightforward paper
> by Angelo Farina on compensating room (and car) acoustics using Ambisonics.
> I think it was in the past couple of years' AES conference pre-prints.

What is Ambisonics? I will base my forthcoming room correction work on
papers produced concerning Roister's (www.roister.com) acoustic
compensator. They exist as AES pre-prints, forgot the numbers though.
 
/Anders Torger


New Message Reply About this list Date view Thread view Subject view Author view Other groups

This archive was generated by hypermail 2b28 : pe maalis 10 2000 - 07:23:27 EST