Re: kAudioUnitID_BandPassFilter
Re: kAudioUnitID_BandPassFilter
- Subject: Re: kAudioUnitID_BandPassFilter
- From: Brian Willoughby <email@hidden>
- Date: Mon, 14 Jan 2002 02:24:28 -0800
[ Depending on the implementation, there is usually a "roll off"
[ (for example, 6dB/per octave for a simple filter) outside the
[ passband.
At the risk of being overly pedantic (or perhaps even wrong ;-), I'd say that
there is *always* a "roll off." I don't think that even a digital filter can
have a slope of infinity.
Regarding the phase shift, I believe that is also par for the course.
Pointers to digital filtering algorithms for computer programming would be
helpful, although I realize that Lieven says he doesn't want to dig too deeply
into DSP.
The problem I see is that while a single filter can be processor intensive, a
whole bank of filters will probably bring the CPU to its knees. At some point,
it would be better to do a Fast Fourier Transform, picking the size of your
frequency bands to match your needs. Stacking up several stand-alone band-pass
filters would be much less efficient than a single FFT algorithm which
computes all the results at once - which has the advantage of using partial
results many times.
Brian Willoughby
Sound Consulting
On Friday, January 11, 2002, at 02:12 PM, Lieven Dekeyser wrote:
>
Hi!
>
>
Maybe this is a stupid question but I really don't know much about
>
audio processing :-)
>
What frequencies does a BandPassFilter AudioUnit with center frequency
>
5000 and bandwidth 600 filter?
>
Does it cut out exactly all frequencies below 4400hz and above 5600hz
>
and
>
leave those between 4400 and 5600 untouched like
>
this: (sorry about the ascii art ;) )
>
_____
>
| |
>
_____| |_____
>
^
>
5000hz
>
or is it more like this:
>
_____
>
/ \
>
___/ \___
>
^
>
5000hz