Re: Sega sound boards...

From: Ray Ghanbari <ray_at_mayo.edu>
Date: Thu Dec 04 1997 - 12:12:29 EST

You wrote:
> > This is off topic for the vectorlist..so please enlighten me in E-Mail
>
> Sure -- leave us all in suspense !

Yeesh! I guess today's schools don't get into Nyquist stuff anymore? ;-)

To have a prayer of reproducing a signal, the Nyquist criteria says you need
to sample at at least double the highest frequency in the source signal,
otherwise you get aliasing.

Practical example:

TV dude has a shirt with horizontal bands. As the camera zooms out, the shirt
goes nuts and large bands start to appear. In this case, the "signal" is the
pattern of stripes on the shirt, and the sampling is the vertical pixel spacing
on the TV. As the spacing between the stripes on the shirt gets to be less
than the pixel spacing on the TV, aliasing starts. Aliasing folds high
frequency components into low frequncy components. That's why the "stripes"
appear to get huge.

This is *extremely* well understood. The mathematics are exceedingly straight
forward, esp. for those of us that still think in Fourier space. Their are
numerous applications that take advantage of this folding phenomena to enable
precision work (take impossibly high frequencies and fold them into a range
where they can be readily observed)

So why did the original CD players sound like a dog shitting in a tin can?
(circa 1984) The sampling rate on CDs is ~44kHz, which is ~2x the limit of
most audio systems (20kHz). The problem was not aliasing, but in the
generation (aka synthesis) of the analog signals. To a crude approximation, a
nice sine wave signal at 20kHz would get reproduced as a nice square wave
signal at 20kHz. Guess what. Square waves sound like a dog shitting in a tin
can.

How to deal with this easily? (at least with state of the shelf DACs circa
1985) Over sampling. In a grossly simplified view, oversampling adds
additional stair steps to the square signal, making it more "analog" like.
Apply a little analog filtering to smooth out the edges, and presumably you
have a nice signal. Apply too much, and things sound like a dog shitting in a
swimming pool (all muddy). This is where the art of audio design comes in, and
why I blew $1k for a Nakamichi CD player back in 1984...everything else
sounded like shit with a bad hang over. DACs today are *amazing* in comparison
and I bow to the circuit gods that make these things work (my wife used to
work as a minion for Analog Devices, and the designers were treated as gods)

My read from the thread so far: Nyquist and aliasing are *extremely* well
understood. Idealized synthesis of analog signals from digital sources with
non-ideal components is less so, esp. for us golden ear types. For retro game
applications, if you notice these things, it's time to crack a beer and get a
life ;-) ;-)

Now to sit back and have the real engineers remind me why I'm in management now ;-)

Ray
Received on Thu Dec 4 09:44:19 1997

This archive was generated by hypermail 2.1.8 : Thu Jul 31 2003 - 23:01:04 EDT