Hi All,
I'm having some issues with an oscillator startup. I'm using a 2 op-amp quadrature oscillator like this one:
http://www.changpuak.ch/electronics/images/Oscillators-1-4.png
oscillation frequency is around 30kHz.
For R3 on the diagram I have a JFET controlled variable resistance, where JFET gate is driven by an integral controller. the controller has the output of a peak detector which is measuring the peak of one output and a reference peak level to compare it to.
If I make the time constant of the peak detector low enough that the peak detector is discharging significantly per oscillation period, and the controller time constant much slower, the circuit starts up just fine and oscillates stably. However, when the peak detector output changes this much on a per period basis, the system doesn't correctly regulate the oscillator amplitude as it drifts over temperature, as the controller essentially responds to the average of the peak detector output as it sawtooths up and down. I need to slow down the response of the peak detector such that it is constant over many periods of oscillation in order for this behavior to be fixed. However, no matter what I do, I can't seem to get the oscillator to startup on powerup correctly if i try to slow down the peak detector response enough. When I slow down the peak detector like this, the circuit will oscillate between dead and sitting on the rails at low frequency (10-100s Hz range), no matter how slow I make the controller response.
I've tried a few ways (adding different resistor networks to/around jfet load) with no success.
Any advice is appreciated!
I'm having some issues with an oscillator startup. I'm using a 2 op-amp quadrature oscillator like this one:
http://www.changpuak.ch/electronics/images/Oscillators-1-4.png
oscillation frequency is around 30kHz.
For R3 on the diagram I have a JFET controlled variable resistance, where JFET gate is driven by an integral controller. the controller has the output of a peak detector which is measuring the peak of one output and a reference peak level to compare it to.
If I make the time constant of the peak detector low enough that the peak detector is discharging significantly per oscillation period, and the controller time constant much slower, the circuit starts up just fine and oscillates stably. However, when the peak detector output changes this much on a per period basis, the system doesn't correctly regulate the oscillator amplitude as it drifts over temperature, as the controller essentially responds to the average of the peak detector output as it sawtooths up and down. I need to slow down the response of the peak detector such that it is constant over many periods of oscillation in order for this behavior to be fixed. However, no matter what I do, I can't seem to get the oscillator to startup on powerup correctly if i try to slow down the peak detector response enough. When I slow down the peak detector like this, the circuit will oscillate between dead and sitting on the rails at low frequency (10-100s Hz range), no matter how slow I make the controller response.
I've tried a few ways (adding different resistor networks to/around jfet load) with no success.
Any advice is appreciated!
