# less-than-unity inverting op amp oscillating. Why?

I've got a GPS disciplined oscillator design I'm working on. Suddenly, the prototype I'm working on at the moment seems to be suffering from what is being caused by oscillation in control voltage.

Here's the portion of the circuit of interest:

simulate this circuit – Schematic created using CircuitLab

Not included in the schematic is a 0.1 µF bypass cap across the power supply pins of the amp. The actual amplifier is an AD8538, but CircuitLab doesn't have one of those.

The input to this portion of the circuit is the output of an AD5061 DAC. The output of this circuit goes to the control voltage pin of the oscillator. The input includes a low-pass filter (the 22 ohm resistor and 10 µF cap), as the actual voltage is only expected to change every 100 seconds or so. The purpose of this amp is to reduce the swing range of the DAC and to act as an impedance buffer between the DAC and the oscillator. The 10k resistor on the output is additional loading, as the oscillator actually has a 100k input impedance.

What I'm seeing is around 6 mV P-P of an approximately 22 MHz sine wave being imposed on the output.

No me gusta.

The power supply has around 2 mV P-P of ripple, but it's not anywhere near the 22 MHz of the oscillation I'm seeing. The oscillator's output frequency is 10 MHz, so it doesn't seem like it's coupling that to me.

What's going on?

• I think you did all the right things and are just unlucky that it oscillates. Have you tried adding a 22pF cap across R6 and/or R4 ? Maybe a 100 nF cap across R1 to decouple that commonmode voltage just a bit (and also supress feedback through the supply lines). Commented Oct 6, 2015 at 15:57
• The closed loop gain from a stability standpoint is really the non-inverting gain, which is >1 in this case. (1+Rf/Ri). This is a chopper stabilized amplifier with a gain-bandwidth of less than 500kHz, so it's not likely a classical stability problem with the feedback loop. It might be you are seeing the chopper frequency coming through the output. In that case a low-pass filter could solve the problem since you clearly don't need gain at 22 MHz given the op-amp choice. Commented Oct 6, 2015 at 16:00
• FWIW, I did try swapping out R1 and R2 with 10k to reduce the impedance to the virtual ground. Nothing changed at all. That suggests to me that decoupling the virtual ground won't help - simply because if that were it, I'd expect some sort of effect from altering the impedance (but I'd not be at all surprised if I were wrong about that). Commented Oct 6, 2015 at 16:15

Check the data sheet on the op-amp. Above about 1 MHz, the closed loop output impedance is about 400 ohms and if there your VCO uses a varicap diode then bingo, that's where the unwanted/unexpected signal is coming from. Try feeding it via a resistor and capacitor with the cap closest the VCO control port.

I'm aware you say that the oscillator should be 10MHz but could it be slipping up to 11MHz and you are seeing an overtone on the vco input?

Here's the section in the data sheet: -

Maybe try a different op-amp.

• It's not a VCO. It's a VCOCXO. The trim range is +/- 10 ppm, so there's no way it could be shifting around that much. However, other commenters have suggested an RC LPF on the output. What do you (or anyone) think of the idea of moving the existing LPF to the output of the amp rather than the input? Commented Oct 6, 2015 at 17:54
• @nsayer - try feeding the "VCO" via a 10 kohm resistor and seeing what the realtive amplitude of the voltages either side look like. Of course 10k may be too big but it depends on the input impedance. Maybe a link to the device would be useful. Commented Oct 6, 2015 at 18:08
• It's a Connor Winfield DOC020V. conwin.com/datasheets/cx/cx207.pdf The control voltage input impedance is 100k. Commented Oct 6, 2015 at 20:29
• All I can say is try biasing the input to midrail from 2x 10k resistors and see what the signal on the input looks like (if there is one). Commented Oct 6, 2015 at 20:39
• Without adding the biasing, the signal on the input is DC (the output from the DAC) with a small (~1 mV P-P) imposition of the power supply ripple, which is at a different frequency than the oscillation seen. Commented Oct 6, 2015 at 22:37

I managed to hack in a 10k resistor between the output of this circuit and the input of the oscillator.

The bizarre sine wave stays on the oscillator side of the resistor. The amplifier side is clean.

So... I'm beginning to think this has nothing to do with the amp.

• Input impedance may be quoted at 100 kohm but in all likelihood there will be a capacitive element due to possibly the input being a varactor diode that is directly controlling the tank circuit in the oscillator module. Leak back from the oscillator tank is common-place - can you confirm that the sine wave is directly related to the oscillator's output frequency? Commented Oct 7, 2015 at 7:57
• BTW it's still a problem with the op-amp - this particular op-amp cannot be expected to drive this sort of circuit due to the issues mentioned in my answer. But, on the other hand, you might find that what appears to be a problem isn't a real issue. Commented Oct 7, 2015 at 7:59
• One of the things I tried to counter any potential capacitative effects was adding a 0.1µF cap in parallel to the feedback resistor (51k). That did nothing. The oscillation is around 22 MHz, while the output frequency is 10 MHz, +/- a few ppb. The power supply ripple is around 2.5 MHz. The only possibility I can envision of coupling is some sort of mixture of power rail ripple and the 1st overtone of the output. Commented Oct 7, 2015 at 13:56
• I would expect it to do nothing because, in effect it would be in series with the 100k (R3). There's no getting away from the fact that the op-amp output impedance is too high. Try putting 100nF on the VCO input and drive it from the op-amp via 330 ohms (estimate). Commented Oct 7, 2015 at 14:46