Oscillator Beating

[breakshift]

I've built the attached circuit diagram on a breadboard for prototyping. It's a variable frequency Wien Bridge oscillator with an automatic gain control circuit. A diode and cap on the output samples the peak output voltage and this is buffered and fed to a standard 5mm LED via a pot to vary its sensitivity. The LED is coupled to a light dependant resistor (via multiple layers of heat shrink) which acts to adjust the gain of the oscillator to control the output voltage.

The oscillator is built to vary between approximately 20kHz and 100kHz.

Op amps are TLC072 on a 5V regulated supply and bypassed by 10uF and 100nF caps.

I initially used C=1uF and R=100k for the peak detect circuit. With these values the circuit oscillates fine but when I use the LED pot to increase sensitivity and thus reduce the output voltage below the op-amp supply rails, the oscillations go into a beating pattern. Screenshot of this attached with oscillations set to 15kHz. The yellow trace is the output and the green trace is the buffered output of the peak-detect capacitor.

I need the output amplitude to be less than the supply rails in order to reduce distortion on the sinusoid.

I'm trying to understand why this beating behaviour happens. I think its something to do with the RC time constant of the peak-detect circuit. Lowering the values of R and/or C seems to settle the oscillations. With C=100nF and R=10k I can reduce the output amplitude and get a nice clean sinusoid.

So the behaviour is related to values of R and C, but also depends on the frequency setting. To get stable oscillations at the higher frequencies I needed to reduce R and C further. So is it something to do with the time constant of the RC network and that of the output frequency?

If it's too high, perhaps it's too slow to react to changes in output amplitude and as a result can't really control it?

Any thoughts?

 

[moffy]

What you have is an overshoot situation when the oscillator starts oscillating in beat mode. The compensation is too large forcing the oscillator off. There is also a variable delay in the LED/LDR circuit, the LDR will turn on faster than it turns off, of the order of several ms which is complicating the whole control loop. You should slow down your amplitude control loop by increasing C, but that's not a nice solution as it will distort the sine wave output. Replace the diode with an NPN transistor e.g. BC547 which will unload the opamp and be faster. I have previously used a sample and hold as a peak detector but I had a quadrature clock source.

 

[crutschow]

Another way to stabilize the loop is to reduce the gain of the feedback. I'd have to see your circuit to tell how to do that.

 

[moffy]

Yeah, thought of that but that's what the pot does. It gets into trouble in the range he wants to control.

 

[moffy]

Crutschow is right. Put a fixed value resistor in parallel with the LDR to reduce the maximum gain.

 

[breakshift]

You should slow down your amplitude control loop by increasing C

Increasing C actually made the beating worse. I found that I could improve things by decreasing C and/or R.

I discovered that the circuit actually works much better without the peak-detect R & C. I've now tied the oscillator output directly to the pot that drives the LED and it's perfectly stables across the whole frequency range. So the LED is driven by the AC output signal and the LDR can't react that quickly so just sits at an average value. No need to drive it with DC, much easier.

So having removed the diode, R, C and voltage follower, everything is fine. The beating therefore must have been a result of the RC. I still don't understand why it was happening? All it was doing was providing a DC voltage to the LED/LDR - it in itself wasn't really part of the actual amplitude control circuit.