OP Amp Oscillations/Feedback

What kind of "feedback" are you using? If it is just resistive it is unlikely that the circuit becomes unstable because there is only one pole, created by the OPAMP transfer function characteristics (-6dB/oct slope) and the resistor in the feedback. Gene

Google for "opamp stability criterion" or similar. You can do that by analyzing the actual circuit inside the opamp together with stray capacitances from the PCB layout, or by the bode-plot resulting from it. Whenever the open loop gain is greater than 1 *and* the phase-shift gets

Thanks for your response Gene- I have the inverter-input connected to ground. The non-inverter is connected to a photodiode- a little tricky here-the photo diode receives light from a laser on the op-amp output to create the loop optically. When I put the scope on either the diode or laser I get oscillations-I think around 1 micro-second but I need to check when I get back to the lab. I presume that the output, when attempting to re-zero the difference on the inputs, overshoots, resulting in oscillation. Further, I would expect laser and photo diodes to have some inductance and capacitance (they are in close proximity so there is no lightspeed lag here). I can play with a decade capacitor box but my main interest is setting up the diff-E for this loop so I can estimate the required compenstation analytically. I thought perhaps this is described in a textbook somewhere-I tried to do it with taylor series but didn't come out the way I expected, and I am sure there is more to the picture that a good text would add.

Thanks Fritz

Thanks for your reply Ban- as I mentioned I am trying to analytically calculate the oscillation rather than resort to "tricks" or trial-and-error here. I did not find anything on the web which describes negative feedback in terms of a second order differential equation however I presume the electronics terminlogy here is "transfer function" and maybe I am looking to much at the mathematical approach to this kind of problem. I will head out to the library and see what I can find on feedback loops.

Thanks Fritz

Hi Fritz,

the discipline that studies this topic is called "Control Theory" and it is quite complex. A book I could recommend is "Automatic Control Engineering" by Francis Raven. In any case, a loop that is unstable( oscillations) is indicative of excessive phase lag and gain above unity. I do not know if you have the capability to perform an open-loop Bode-plot, but if you were it would tell you the degree of instability of the loop and the way to stabilize it. Adding capacitors will not solve the problem. Considering your lack of knowledge in this field I recommend the following:

1. Reduce the gain of the OPAMP until the loop has good stability.
2. Is the performance accettable? if yes, leave it. If not go to the next.
3. Is there a tracking error problem? If yes you could try to to put a capacitor in SERIES with the feedback resistor that sets the OPAMP gain. Start with a big one and step by step decrease its value until the loop become unstable. Mark the capacitor value when this happens. Intall a cap that has double of the capacitance. Now the loop should be stable, with virtually zero tracking error.
4. If you have dynamic tracking problem, then it is problematic to improve the situation without having control theory knowledge.

Das Glueck ist dem Kuehnen hold!

Gene

but

such

I don't know about the rest of the world out there, but before spice simulators, I never ever found a circuit that could be nicely stabilised by the application of cold theoretical methods. Disregarding of course, those trivial textbook examples, or the cop-out of slugging the response to that of treacle, or having available a roomfull of gain/phase measuring equipment, or having a pocketfull of those nice, dominant poles that everyone except me seems to have in their toolkit. The impossible to define parasitics seem always a major factor. The semiconductor data sheets give too little info. The source and load Z's are usually nothing like those estimated. That 20nH of unseen, inductive cross-coupling may be impossible to analyse, measure, estimate, simulate or (sometimes) even understand but is resulting in 200megs oscillation, massive current drain and signal distortions, random device failures and exciting artefacts in the preceding electronics. Any solution must come from an in-the-flesh-on-the-bench-suck-it-and-see approach. With experience it gets easier to home in on the sweet spots but it's still ad-hoc and unscientific.

As the best *fixes* usually seem to involve the odd strategic R or C (or god forbid, an occasional L), to me it still most definitely feels as if I'm cheating or playing a dirty trick on the circuit, when just by adding that critical 2p cap' the response instantly changes from a monstrous nightmare, to that of pure sweetness and light.

Unless others know better :-)

regards john

I agree 100%. The theory can lead to the "ballpark", but then good lab investigation and analysis finalizes the design. Gene

Fritz, the EEs do not use differential equations for their transfer functions, but the Laplace transforms with the frequency operator S= j f/f_nominal. This way the whole thing can be solved with simple algebra and can be overlooked easily.

OK and I thank you all. I checked back and I see 50 microsecond Oscillations. Again the feedback is optical (e.g. there is no feedback resistor). I found that placing a .04 uf capacitor from the output to input removed the osciallations and I am now able to modulate the laser as needed. I appreciate the book reference from Gene and will get a copy (I have D'azzio but I presume it is somewhat dated).

Viele Dank, Fritz Woods Hole Oceanographic