🚀 go-pugleaf

In article <1990Aug06.150222.23167@pmafire.UUCP> tuv@pmafire.UUCP (Mark Tovey) writes:
>
> In article <1990Jul20.223615.4305@portia.Stanford.EDU> ceg@nova.stanford.edu 
>(Chris Gronbeck) writes:
>>
>>I'm looking for suggestions as to the best way to design a circuit to
>>drop music
>>one or more octaves.  A typlical application being making an electric guitar
>>perform like a bass.  Any hints would be useful.  Thanks.
>>
>> 
>   My memory is very dim at this point but I recall reading years ago in
>something like Popular Electronis or something like that about plans for 
>an ultrasonic sniffer. This device utilized the receiver transducer for 
>the old style television remote control units to pickup ultrasonic sounds.  
>This is where my memory gets real hazy. If I remember correctly, the signal
>was passed to a circuit that resembled the front end of a radio receiver and
>reduced the signal down to some intermediate bandwidth. From here it was 
>converted to normal audio and passed into an audio amplifier.

This is perfectly reasonable thing to do; I recall a similar article
fairly recently.  The device is also useful for listening to bats.  You are
correct in your description until you mentioned "bandwidth."  Substitute
"frequency" for a true statement, or read on...

>   I have no idea how well it worked or how to go about building one, but
>it seems to me that the two problems are similar: shifting a signal down
>a known amount in the spectrum. Any ideas or thoughts?

Since no one else has jumped in, I'll try to remember my communications theory
class of a few years ago.  The action of the heterodyne converter you
described above is qualitatively different from the "reduce by one or more
octaves" function.  The heterodyne converter "mixes" the input signal with a
constant sinewave in a nonlinear element, ideally, a multiplier.  The result
is a signal at the frequency (or frequencies) which are the _sum_ and
_difference_ of the input and the constant sinewave.  Note that the only
change in "bandwidth" that has occured is a doubling because both the sum and
difference are involved.  This is essentialy AM modulation. One of the two
"sidebands" can be discarded and the result can still be demodulated
("single-sideband"); the bandwidth doubling is not a required.

For more info:  Any good communications theory textbook for EE's.  (warning:
calculus required) an OK one: Roden, Martin S. _Analog and Digital
Communication Systems_, Prentice Hall 1979, 1985.  ISBN 0-13-032822-7
There are likely better ones.

Summary: mixing with a local oscilator results addition and subtraction of
frequencies.

Now, in music (and elsewhere) the boundaries of an "octave" differ by a
factor of two; the required operation to reduce a frequency by an octave
is a division.  If you were to take an octave-wide chunk of the audio
spectrum, say from 1000 to 2000Hz, mix it with a 500Hz signal and filter
appropriately, an 1000Hz input would be reduced an octave to 500Hz.
but, a 2KHz input ends up at 1500Hz, which is not an octave down.  One
octave has become more than an octave.  That doesn't make for music that
sounds right.

The technique of clip, devide with a flip-flop, and filter will
reduce a single signal by an octave.  It will also introduce a lot of
distortion, which may not be bad if you're playing certain kinds of
guitar music :-)

To do better, you might take a set of notch filters and device up the spectrum
of interest into lots of little pieces. In each one, split the signal two
ways.  One path gets clipped, and has only frequency information.  Rectify and
filter the other to get amplitude information.  Devide the frequency
information by two with a flip-flop, and use a voltage-controlled-amplifier
controlled by the amplitude half of the signal to get the amplitude right.
Finally, sum up all these little parts of the signal and you've reduced the
input by an octave.  For better results, split into more frequency bands.  I
would guess that eight or ten bands per (input) octave might even sound good.

For perfect results, split into infinitely many frequency bands.  This is
essentialy the same as taking the fourier transform of the signal,
changing the frequency information, and taking the inverse transform, which
someone already suggested.  This is done with a digital signal processor.

One last method, for guitars anyway:  Sell the guitar, and get a MIDI
guitar controller, which you play like a guitar and it produces MIDI
signals corresponding to the notes you hit.  Run the MIDI data through a
computer to mess with the note data and drop things down an octave, and then
to a synthesizer set up to sound like a guitar.   Some synths can
probably do the translation without an external computer.  (1/2  :-)

Sorry this got so long, but I hate to see incomplete information go
uncompleted.  Someone may get really confused.  Someday it may be me who
needs the info.
--------------------------------------------------------------------
Steve Tell      e-mail: tell@wsmail.cs.unc.edu usmail:  #5L Estes Park apts
CS Grad Student, UNC Chapel Hill.                       Carrboro NC 27510