🚀 go-pugleaf

From article <2laeps$dik@news.cs.tu-berlin.de>, by derek@cs.tu-berlin.de (Dirk Kuhlmann):
> Hi,
> 
> I am new to the multimedia, audio and acoustics groups. Please don't flame
> me if this topic already was discussed or if I missed a FAQ.
> 
> The problem:
> 
> Audio data of 8kHz, 8 bit ulaw shall be converted to 44.1kHz 16 bit linear
> representattion. There is no DSP support. The conversion is part of a
> multi-channel audio mixing driver and therefore has to be as fast as
> possible. A FFT is not possible due to timing restrictions.

[...]

You haven't really stated the problem, just a difficulty with an attempt
at _one_ solution - note the last sentence below.

If the original sampling is done at 8KHz, then to reconstruct the
analogue sound (which you are effectively doing in order to sample at
44KHz) you must pass the time quantized samples through an appropriate
low pass filter with a cutoff of about 3.4KHz (would you believe a voice
circuit bandwidth? :-)).  FFT is one way, but a transversal filter may
be easier, as you can get away with a few bit shifts and adds instead of
a general purpose multiply when attenuating the samples before summing
them.  The shift register (delay line) need only be a circular buffer.

The biggest problem is aliasing if you must keep the 8Khz and 44.1KHz
fixed.
If not, 8 * 5 1/2 = 44.0 : you might find some trick in that, otherwise
you can run the 8KHz source 10% fast or slow for 5 or 6 times
oversampling.

If it has to be done in realtime, a few appropriate ICs is probably the
simplest way, rather than try and make a sluggish general purpose
computer emulate them!
Allan Duncan		ACSnet	 a.duncan@trl.oz
(+613) 253 6708		Internet a.duncan@trl.oz.au
Fax    253 6664		UUCP	 {uunet,hplabs,ukc}!munnari!trl.oz.au!a.duncan
    Telecom Research Labs, PO Box 249, Clayton, Victoria, 3168, Australia.