3
\$\begingroup\$

For whatever reason, I seem to create a lot of oscillators in the course of developing circuits involving op-amps.

I'm wondering if the frequency of oscillation, which normally appears to be quite stable, can be used for diagnostic purposes?

For example, could I assume that the open loop gain is 0dB at the oscillation frequency and then back-plot a 40dB/decade line from that point to identify the frequency of a pole I need to compensate out by where it crosses the open loop gain plot from the datasheet?

\$\endgroup\$
3
\$\begingroup\$

The terminology is a bit confusing - but you must distinguish between "open-loop gain Aol" (gain of the opamp without feedback factor B) and "loop gain (Aol*B)" (gain of the complete loop). Thus, if an opamp with feedback oscillates, you can assume that at the oscillating frequency the loop gain (AolB) is unity. More than that, it is correct that at this frequency the loop gain magnitude roll-off will be app. -40dB/dec (phase shift -180 deg). However, at which point of such an asymptotic line you will find the pole frequency? This is possible only if you know the first (lowest) pole frequency

Therefore, the best method for identifying the second pole is to record the magnitude function vs. frequency.

\$\endgroup\$
1
  • \$\begingroup\$ Okay, that's really good to know. So if I have the open-loop gain frequency response from the datasheet, that shows the position of the "first" (lower frequency) pole, then back-plotting a 40dB/decade asymptote should intersect at the pole to be compensated, it sounds like. Of course assuming unity gain etc. (β = 1, so Aβ = A). This should help me narrow things down a bit. I can adjust the data sheet Bode plot to incorporate non-unity β etc. if present. Thanks @LvW :) \$\endgroup\$ – scanny Jul 28 '15 at 23:49
2
\$\begingroup\$

Another cause of oscillation is caused by inadvertent coupling rather than because of problems in the feedback compensation network.

One common source of inadvertent coupling is via the power supply. The power supply rejection ability of opamps can be very poor at frequencies higher than a few 10's of kHz. If you have inadequate decoupling. In general you should have some large electrolytic capacitance and also some small ceramic decoupling to take care of both low and high frequencies.

In amplifies with discrete devices this can even happen at low frequencies and in audio amplifiers was often called motor-boating because of the sound it makes.

Be careful with your grounding - sharing a common section of wiring between a high power output stage and a low level input stage will often cause problems. With audio amplifiers it can cause bad distortion if the current from a class AB output stage gets into an earlier stage.

At higher frequencies the inductance of the grounding system can be very important. I have even seen this problem with some opamps with the way that the wiring was arranged to the decoupling caps. With high bandwidth amplifiers good construction is required, either a PCB or for a hand constructed prototype use a blank piece of unetched PCB as a ground plane with Ugly construction. (You should be able to find lots of articles on the web about that).

Keep the capacitance low at the summing junction of an opamp. One time I had a simple opamp inverter with 10K input and feedback resistors that was oscillating at a few MHz because there was about 3pF of capacitance at the summing junction - that caused enough has phase shift to make it oscillate. I usually put a small feedback capacitance (1pF or so) between the output of the amplifier and the inverting input to tame those types of problems - of course in that case I forget to do that!!!

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.