Re: Simple audio band-pass filters with sharp cut-offs



On 27 Nov 2005 12:48:23 -0800, mistawizard@xxxxxxxxx wrote:

>I want to split an audio signal into 5 bands by frequency range (top
>and bottom can be low-pass and high-pass) and send these into speakers
>with moderate power, say 10 watts. I can use amplified speakers if a
>5-way audio amp gets complicated.
>
>This is not for quality audio, but for educational demo purposes, so I
>don't need an even response and distortion is okay. I'm really looking
>for the sharpest cut-offs I can get, in as simple a box as I can build.
> Can anyone suggest how to accomplish this?
>
>I haven't made many coils before (rather embarassing for a ham op), but
>I'm fine with simpler circuits. I found a post about an IC that might
>do the trick, but I'm doubting it'll handle audio frequencies.
>

One good resource is Don Lancaster's "Active Filter Cookbook".
You will not need any inductors (one of the virtues of active
filters), but you will need some matched components. Look
at "equal component Sallen-Key" designs. These avoid the
need to have 2 different specific capacitors in each stage to
control Q and frequency. Instead, you use 2 equal-value
caps, and equal-value Rs, and you adjust Q with a separate
gain factor. Stability is alleged to be poorer than the unequal
Sallen-Key, but IMHO that's more than compensated by the
ease of correct tuning in the first place.

You will have to make some tradeoffs between complexity,
sharpness of slopes, and passband ripple.

Also, with only 5 bands to cover the audio range,
the bands will be too wide to use classical
bandpass fitler stages. You will need to create each band
from a separate high- and low-pass stage, or a chain of
simple bandpass filters summed together.

In fact, you may find it simpler to change your design to use more
than 5 bands, just to allow simpler bandpass stages. (They should
be well under an octave to use bandpass stages. Narrower
allows sharper tuning.)

You will come to appreciate why designers prefer to make
filters like this digitally. One A/D and DSP chip, and you can
generate any combination of filters and slopes with perfect
stability and no parts tolerances. You can use FFTs (a
whole bunch of simple bandpass filters, equally spaced)
or Finite Impulse Response (delay line with summed weighted
taps) or even the trickier Infinite Impulse Reponse types
(like analog filters) to create fitlers for this job.

Best regards,


Bob Masta
dqatechATdaqartaDOTcom

D A Q A R T A
Data AcQuisition And Real-Time Analysis
www.daqarta.com
Home of DaqGen, the FREEWARE signal generator
.



Relevant Pages

  • Re: Best Playback Quality (MS PEOPLE!)
    ... > It might depend on that too, also try start-dxdiag and see if the audio ... > think it's a cheap trick to lower the bitrate slightly. ... trashed the k-lite mega codec pack and moved to ace mega codec. ... in the course experimenting with graphedit i've stumbled across some filters ...
    (microsoft.public.windowsmedia)
  • Re: BBC Radio 4 Feedback (13:30 Friday) - Discussion about move to Digital at around 13:50.
    ... again there are filters introduced that start to operate at much ... If you look at the spectrum of an FM signal you'll find that 15 kHz is ... MP3 is a far better audio codec than MP2. ... you know nothing about the subject of digital radio, ...
    (alt.radio.digital)
  • Re: Dirty Digital [sic.]
    ... Dither which Martin Colloms had published in the August 1983 issue of HFN. ... So it is clear that many people in the consumer audio mags knew about this ... employed transverse digital filters. ... for the kit used and the noise floor of a dithered signal and a finite ...
    (uk.rec.audio)
  • Re: Icom R71
    ... Icom's manuals aren't the greatest in the world. ... > KIWA offers an audio upgrade kit for R-71. ... ICOM filters are going to ...
    (rec.radio.shortwave)
  • Re: Audio
    ... Richard Dobson wrote: ... > and further bands either side are at octaves or fractions/multiples of octaves ... > defined by Centre Freq, and the bandwidth is determined by the filter slope, ... The filters will still ...
    (comp.dsp)