Bill Cornutt writes:
>This looks simple.  Therefor it may not work.
>
>I am making the assumption that is a audio device and is
>intended for "ears only".  And any gliches will not be
>noticed by ears.
>
>The solution is to take ten samples of the input waveform
>at equal time periods, and output nine at a different equal
>time peroid.

        It will work and does work in such systems as speech compressors
which allow one to speed up the speech in a voice recording while retaining
the pitch.  There is a fly in the ointment with a lot of these systems and
here's what it looks like.

        The input to the system is fed to either an analog or digital sampler
where it is briefly stored  in memory like a snippet of audio tape.  I
believe that these devices take around 20 to 50 samples per second and
then play them back either faster or slower than they were originally
recorded to get the pitch change.  The trouble is that it is difficult to
get the next samples beginning to cleanly match the same voltage level that
ended the previous sample especially since one sample is taken at a slightly
later time than its neighbor.  You do definitely hear this as a burbling
or watery sound to the signal, especially if the pitch change ratio is
great.  It is an audio version of the same bad visual effect one gets on a
two-head video tape player in one of its fast view modes.

        It is more than likely possible to use DSP methods to interpolate
one sample in to the next one, but it is not as trivial as it first seems
to get good sound out of such a system.  I really don't know what is possible,
but I have heard many of the conventional sampling type systems and they
don't sound natural at all.  Another problem, assuming that one figures
out how to interpolate the samples so that you get a clean wave form is the
fact that when some of the signal is discarded as it is going to be when
one shifts the pitch lower, pieces of the intelligence get thrown away.
Human speech flows neatly from one sound to the next and gaps in that flow
add their own unique odd sound to the final result so what you end up with
is a voice at its normal pitch, but you know there is something wrong because
the words have funny little glitches in them that sound the way the motion
in a cheap animated cartoon looks;  more like a bunch of stills spliced
together rather than a movie.

        What I am saying is basically that you certainly can do this sort
of thing, but don't count on it sounding normal.  It won't be horrible, but
it won't be mistaken for somebody just talking or playing music faster.

        As a bit of trivia, the first systems that did this sort of
modification to sound did it electro mechanically with helical scan tape
heads that rotated just like in a VCR.  They cost around $50,000 in 1950's
or 1960's Dollars.

Martin McCormick WB5AGZ  Stillwater, OK
OSU Center for Computing and Information Services Data Communications Group