I think you answered your own question. Normal arrays of 2 pole
bandpass filters would have a large amount of band overlap due to
their rather gentle slopes . If you use an array of high Q filters,
the bands would be so narrow that you'd have to use a zillion of them
to cover the spectrum. A good compromise is to use a 2 pole filter in
the center of the band, with a high Q filter at the upper and lower
edge of each band to "square the corners" and maximize the separation.
Moe
vocoders? (I would guess that for vocoder channels you'd want plenty
of discrimination between channels.)
<<<<<
bandpass filters would have a large amount of band overlap due to
their rather gentle slopes . If you use an array of high Q filters,
the bands would be so narrow that you'd have to use a zillion of them
to cover the spectrum. A good compromise is to use a 2 pole filter in
the center of the band, with a high Q filter at the upper and lower
edge of each band to "square the corners" and maximize the separation.
Moe
>>>>>Out of curiosity, what sort of response is this and why is it good for
vocoders? (I would guess that for vocoder channels you'd want plenty
of discrimination between channels.)
<<<<<
>
> A typical BPF (like the Emu UAF, Oberheim SEM filter in BPF mode,
> and many many similar filters) are two pole designs. For a good
> vocoder filter (a single vocoder BPF!) you'd connect 3 such filters
> in series, two of them with the same Q factor, the 3rd with a
> different Q, and center frequencies spread such that the "middle"
> filter is halfway between the "outer" filters on a log frequency
> scale. Two such blocks are needed for analysis and synthesis of one
> channel, and 20 channels were a good choice for the "classic"
> Sennheiser vocoder.