Absolute value VCO

[Davor]

A small contribution, but it may help someone with ground balanced PI.
This VCO works for both positive and negative voltages, and assumes 0Hz at 0V. Its linearity depends on supply voltage symmetry, but as seen in the example here, it will work happily regardless, and 0Hz will always be at 0V.

The circuit is a balanced version of a National AN-299 VCO, but unlike the exponential response original, this one is linear. I also tamed the currents so that glitches are below 1mA, and much less to the rails.

If more swing is needed, just increase R2 and R4, but it can not exceed the output swing of an op amp.

I also included diodes D1 and D2 to improve high frequency behaviour.

The only feature where it does not excel is signal symmetry - it almost does not exist.

In this configuration it will oscillate in a range ±1.5V, but if wider range is needed, just increase R3. And vice versa.

Enjoy. (I placed a copyright notice for commercial use)

[6666]

Thanks Davor
the data sheet for the LT1057 says
The LT1057 may be inserted directly into LF353, LF412, LF442, TL072, TL082

so would these other chips work ok in your circuit ?

[Davor]

Actually the opamp is the least critical part. It may as well be a LM324 powered with ±4V, provided you keep the output swing small enough so that it does not saturate. I placed LT1057 simply because it is a working opamp model supplied with LTspice. Its role is integrating input current, and at this place just about any opamp will do.

This VCO works as follows... Voltage is supplied to R3, and because the inverting input is at 0V, it is converted to current I=U/R3. This current charges a capacitor C1. So when a positive voltage is on R3, the opamp output voltage ramps down linearly - it is an integrator. At the voltage level set by Vrail*R4/(R4*R5) the transistors conduct and short-circuit the capacitor, so that the process is repeated.

I placed 1k for a R6. Its role is reducing current glitches. At the capacitor those are below 1mA now. It also makes the switching more sluggish, reducing the top frequency. In effect by increasing R5 the viable voltage span at the input is reduced, but the short-circuit glitches are reduced. Beyond the top viable frequency the oscillator stops oscillating, but assumes oscillation as soon as the voltage returns to viable boundaries. So it does not latch permanently

If wider frequency range is needed, R6 needs to be reduced. At 330ohm the glitches rise to below 3mA, but the viable control voltage span rises to ±4V.

I made a simulation with octave step control voltage (.step param control list -4 -2 -1 -.5 -.25 -.125 -.0625) going from -4V, and halving control voltage in each step. The result is very promising

For the working schematic I reduced R6 to 330ohm, and increased C1 to 15nF to roughly get 1V to 1kHz relationship.

[Teleno]

Thank you Davor. Original ciruit ideas most welcome, even if not used they're sources of inspiration.

Perhaps part count can be reduced using unjunctions or SCR's?

[Davor]

Yes, but comparing availability, BJTs are the best choice around.

I'm not yet in a mood for exploring the rest of the circuit, but most obviously it may be used for gating some feature that depends on signal polarity. Symmetry sucks, so the first step must go in that direction.

[Davor]

I realised there is a way to extract transition spikes as a digital signal. Of the bat I can use those to clock a pair of flip-flops that will provide me with in-phase and quadrature perfectly symmetrical binaural audio.