Opamp Oscillation

[Evan.Cornell]

I have a amplifier+filter signal chain comprised of AD4807-1 opamps. Singled ended supply of +3.3V & reference of 1.65V.

My original simulation was done with TINA-TI and didn't see any problems, but now that I've got the prototypes back, I'm getting a ~12-13MHz oscillation on the output of the first amplifier stage. I tried to re-run my TINA-TI simulation and wasn't able to get it to run fully. Possibly the original simulation was suspect.

I re-drew it up in LTSpice (attached) and do observe a ~35MHz oscillation in simulation with no input signal. Any ideas on how I can resolve this?

Thanks!

[KT88]

Hi Evan,

it looks lie a bug in LTSpice or the model of the ADA4807. If you delete the rest of the circuit, it's stable...

Cheers

Andreas

[Andy Watson]

Too much phase shift in the feedback - is always the answer

I'm getting a ~12-13MHz oscillation on the output of the first amplifier stage.

Is it genuine?
I note that the datasheet says "Capacitive Load = 15pF" for a 30% overshoot. I read that as suggesting that more than 15pF will push the op-amp into oscillation. A few pF due to the following circuit, plus a scope-probe could easily reach 15pF. When probing such sensitive circuits I add a 100 resistor at the tip of the scope probe to "isolate" it from the op-amp.

Any chance you could provide the schematic in a more universal form.

[Zero999]

Any chance you could provide the schematic in a more universal form.

LTSpice is a free download, but here's a PDF of the OP's file. I haven't done any calculations or simulated those filters, but presumably they're stable and shouldn't oscillate. It looks very convoluted. There's probably a way to simplify it.

[Andy Watson]

LTSpice is a free download,

May be, but as I don't normally use it ...

It looks very convoluted.

Indeed!
I've not done the calculations but a cursory glance at the component values would suggest filters operating at audio-frequencies, but the op-amps boast 180MHz banwidth!  I would suggest that the stage one op-amp is not going to like driving 500pF (C6 & C8) into the output of U2.

[Pulsepowerguy]

The layout might be the problem. Please share an image of that.

[KE5FX]

Tip: don't use rail-rail amps unless you actually need them.  I had to learn this the hard way myself.

Does the oscillation look like this, or does it dominate the whole waveform?

http://www.ke5fx.com/1677_osc_sine.png

[mikerj]

There is no supply decoupling at all in the LTSpice netlist, does your real circuit have some?  These are fast op-amps, stiff supply rails are going to be important.  How is your Vref voltage being generated?

[Evan.Cornell]

Does the oscillation look like this, or does it dominate the whole waveform?

No, it dominated the entire waveform.

Using https://www.analog.com/media/en/technical-documentation/application-notes/an148fa.pdf, I calculated that the feedback resistor network along with assumed 4pF parasitic capacitance works out to be right around the 12-13MHz mark.

I've changed the simulation to use ADA4805-1 instead of ADA4807 and things seem to behave. I will prototype it and see if that helps.