ODP: thru zero VCO questions

[Michael Baxter]

Hello all,

This a great list! I've been thinking about pitch conversion, picking up
from where Robert Moog/Harold Bode's 1972 JAES paper left off. Also, I'n
working on an exponential DAC circuit -- more on this later.

The system problems in pitch shifter design seem a little like the thread
of discusssion on thru-zero VCOs. Some thoughts on this follow. And more
recent research on devices suitable for use in this pitch shifter is
perhaps relevant to the current discussion on techniques for thru-zero
VCOs (below).

On Wed, 15 Sep 1999, Martin Czech wrote:

> :::How about this:  Run the VCO in the normal frequency range, mix the output
> :::with a crystal controlled oscillator at, say, 1 MHz, then go through a
> :::bandpass filter, 20 KHz wide, centered at 1.01 MHz, then mix again with a
> :::crystal controlled oscillator to mix down to the audio band, then low pass
> :::filter, 20 KHz.  The two cyrstal oscillators should be at slightly
> :::different frequencies, maybe 10 or 20 KHz apart.  That way the resulting
> :::output can pass through zero and into the "negative frequency" (reversed
> :::phase) range.  The 1.01 MHz filter can be done with LC components.  The
> :::stability of the crystal oscillators should be no problem, so the stability
> :::of the whole thing should end up the same as the first VCO.
> :::
> 
> You want to supress a sideband at 1MHz with LC-filters?
> Forget it. It will be difficult enough to have the quartz
> osc. stable enough, they usually spec to 100ppm (Drift, aging).
> 
> HAM people get a (very bad) sideband rejection with mechanical filters
> @ 455 kHz (something like -30dB), that's good for radio, but
> not for a frequency modulation substitute.
> 
> There must be a reason why people use complicated polyphase
> or phase difference networks in order to throw
> away the unwanted sideband.
> 

I recently came across an application note for the Burr-Brown MPY634
multiplier, which has bandwidth into the HF (10 MHz) as a precise analog
multiplier.

You can make a very high-Q filter with only RC filters, by using two
multipliers. The example application in the Burr-Brown application was a 5
MHz center frequecy filter. This is down-converted to DC with a 5 MHz
sinusoidal oscillator, fed to a 1 KHz RC active low-pass filter. The
filter output is then up-converted back to 5 MHz (the IF) using the same
oscillator and a second multiplier. Using a Fc = 1 KHz filter in this way
is functionally equivalent to a 5 MHz bandpass filter with 2 KHz bw,
roughly a Q of like 5000.

This sort of a trick seems useful for the thru-zero VCO. 

Now for audio, as versus RF, imbalances matter more certainly, since any
inharmonic content might be "heard." This is evidently why the Moog/Bode
pitch shifter used quadrature phasing networks; additionally, it offered
simultaneous up and down-shifted outputs. The paper talks about a nice new
stereo effect where a square wave LFO modulates the pitch shift, resulting
in up/down shifted ping-ponging between the left and right channels.

Anyway, the point of this message was to illustrate a workable way to have
very high Q RF frequency filters, as this has potential applicability to
linear FM modulated VCOs, where only ultrasonic frequencies might be used.

The Burr-Brown devices look cool, and are available, if expensive. You can
find the MPY634KP for sale at Digi-Key:
www.digikey.com

Also, a datasheet on the part is available at:
www.burr-brown.com

Best,
Michael Baxter
Silicon Valley California
mabaxter at pacbell.net