Re: Sega sound boards...

On Thu, 4 Dec 1997, mayday19 wrote:

> I dont know anything about all that digital logic stuff you guys would
> use to make it work..I'd like too... but in regular audio applications
> you basically set your sample rate by how high your highest frequency
> will be. The rule, or the 'Nyquest Thoerem' is that you must sample at at
> least 2 times the highest fyrequency recorded (to sound accurate).For
> example CD players sample at 44.1 KHz which is roughly twice the higest
> frequency humans can hear. Digital voice recorders usually sample @
> 32khz (I think a few first-generation DAT decks did too). If you do not
> follow this rule you will get false decending harmonics, which will sound
> bad. nobodys knows why but it just happens.

        Not true. We're getting off topic here, but the Nyquist criterion
is very well understood. Sampling below the Nyquist frequency will result
in aliasing of the signal like you alluded to. It makes more sense in the
frequency domain, and I'd suggest consulting and Digital Signal Processing
Textbook ("Digital Signal Processing" by Alan V. Oppenheim and Ronald W.
Schafer is still regarded as one of the "bibles" of DSP, although it's a
bit dated (1975) but does a good job of describing this, as do lots of
other books) or I'll be happy to explain it in email.

> I'm assuming you know this cause you said you doubt it is over 4khz. If
> this is the case, then a 8khz sampling rate will work. It is cutting it
> close, but shouldn't make a difference. It would be a good idea to add a
> filter to block anything above 4Khz if you want to sample at 8Khz.
> Sampling at 8Khz probably wont sound the best though. Of course the
> faster you sample, the better it will sound. There are HD recorders out
> now that can sample at 96Khz. Holy gigibytes Batman!

        96 kHz is, for some reason, catching on as the new audio standard.
Sampling above the Nyquist frequency generally does not make something
sound better. Equipment often boasts huge oversampling rates 128X, etc
because they use oversampled converters in them (i.e. Delta Sigma) which
have better performance, but for entirely different reasons (I can get
into this in email too, because it's what I do for a living :-) but it's
way off topic for the vectorlist.) Every part that is coming out of our
group (Digital Audio) these days supports up to 96 kHz sampling.

        Talking over the telephone is basically 8kHz so I see no reason
why it wouldn't work for storing the voice samples at that sampling
frequency. Since it looks like Al's already got the samples, all that
would need to be done is another sample-rate-conversion. Darn, too bad
this didn't happen in a few months, as we're about ready to release an
asynchronous sample rate converter chip. I love putting Crystal stuff to
work in important projects (i.e. Video Game stuff ;-) )

Joe
  
Received on Wed Dec 3 23:56:07 1997