Analog Pitch Shifter Idea

[Sean Costello]

Hi folks:

I came up with a fledgling idea for an analog pitch shifter, using
bucket-brigade delays.  Feel free to shoot it down if necessary - all input
is welcome.

My idea for one would be a derivative of a frequency shifter, and would use:

- A quadrature function generator, that generates sawtooth and triangle
waves that are 90 degrees out of phase.
- Two bucket-brigade delay lines (semi-short ones, like the SAD 1024).
- Two VCAs.

The signal is fed in parallel into the two BBD's.  The BBD's each have their
own clock.  The seperate clocks are driven by the sawtooth outputs of the
oscillator, such that each clock is modulated by a sawtooth signal that is
90 degrees apart from the sawtooth modulating the other clock (gee, maybe I
should become a technical writer ;).  Each BBD sends its output to the
signal input of a VCA.  The control inputs of the VCA's are driven by the 90
degree out of phase triangle waves, in such a way that the triangle wave of
each VCA is at its most negative when the sawtooth wave driving the
respective BBD is at its most positive.  The outputs of the two VCAs are
summed together.

As the VCAs are being driven by triangle waves that are 90 degrees out of
phase with each other, the sum of the outputs should be a signal that is
consistent in amplitude.  Each VCA fades in the BBD signal at the beginning
of the sweep of the sawtooth wave - in other words, when the frequency
starts to be shifted up.  As the sawtooth wave sweeps the frequency, the VCA
reaches its maximum value, then begins to fade out.  When the sawtooth wave
reaches its peak and resets, the VCA is at minimum volume, so the glitch
that would result from the sawtooth reset will be unheard.

For changing the pitch of the above circuit, I can think of two methods.
The speed of the oscillator could be changed, as higher frequencies would
generate more extreme pitch shifts.  The drawback would be that any glitches
would be shifted up in frequency with higher pitch shifts, eventually
becoming audio rate.  The other method would be to use a pair of multipliers
to control the amount of sawtooth that goes to the BBD clocks.  With this
method, you could go from positive to negative pitch shifting by varying the
sign of the control voltage fed to the multipliers.  The glitches could at
least be manifested as more of a tremolo-type sound with this method.  

Would this work?  Any idea how much pitch shifting could be realized with
this circuit?  How long of a delay line would be optimal?  Would the
Panasonic delays have wide sweep ratios necessary to get nice wide pitch
shifts, or would the Reticon devices be needed?  

Once again, any and all input appreciated.  If anyone has an idea how the
older pitch shifters worked, I'd love to hear it.

Thanks,

Sean Costello