Variable ramp generator

[Don Tillman]

   From: "jhaible" <jhaible at debitel.net>
   Date: Tue, 6 Jul 1999 21:32:09 +0200

   > I know, it is possible to independently control the charge and
   > discharge times in integrator by using a pair of diodes or
   > a CMOS switch. AFAIR, Juergen described such operation of KORG
   > LFOs. The problem is if one applies this control scheme to VCO
   > it leads to severe tracking problems.

   I remember Don Tillman once offered an idea how it could work,
   but I've never tried it or followed it further. Don ?

A variable duty cycle sawtooth is easy with a potentiometer
controlling the ramp rates.  1/RampRate1 + 1/RampRate2 = Freq, and you
absolutely want to hold want to hold the frequency constant when you
change the duty cycle, and a pot can split a control voltage between
two inverting opamp stages exactly like that (because the sum of the
resistances of the left and right hand sides of a pot is constant).

The voltage controlled duty cycle sawtooth is much more difficult.
I'm convinced that implementing accurate enough 1/x functions with
transistors is not workable here, so my suggestion was to start with a
regular sawtooth and warp the waveform to make it a variable duty
cycle.  

To do this, take a positive sawtooth going from 0v to 1v, and its
inverse, a negative sawtooth going from 1v to 0v.  Put each through a
current controlled gain stage, and combine them with a simple "min" or
"and" circuit that delivers whichever of the two voltages is most
negative.  Voltage control those gain stages like a panner circuit so
one goes from a gain of 0.5 to some large value (like 20) while the
other goes from the large value down to 0.5.  Inaccuracies will show
as variations in amplitude, not variations in frequency.  

No, I haven't built an actual circuit.  I've been way busy with other
stuff.

  -- Don