Just use a 100Ω resistor: Understanding a rule of thumb for oscillating amplifiers
A capacitor and a resistor that hang on the output of an amplifier change the amplifier's open-loop-gain curve.
By Bonnie Baker -- EDN, November 27, 2008
So there I was, a new hire for a leading-edge analog-electronics company. As it was my first job, I was wide-eyed and excited—confronting new problems left and right. One problem I ran into involved the stability of an amplifier circuit. In this application, a buffer-amplifier circuit, with a capacitive load, sang like a bird. Because I had a huge community of experts around me, I ran around and asked for advice. The words of wisdom that came down to me were, “Oh, put a 100Ω load resistor between the amplifier output and the load capacitor.” When I asked why, the engineer said, “Just do it. Trust me; it will work.”
So, I built a new circuit as suggested, and, lo and behold, the circuit still oscillated. This new circuit was still singing, but it was producing a new frequency. I returned to the engineer who gave me the first bit of great advice. His recommendation: “Change the 100Ω resistor to a 500Ω resistor,” still with no explanation. It solved my problem, but, given my work load, I did not return to his suggestion for several years. Now, it has come back to haunt me. I need to know what is really going on!
What I did not understand then but understand now is that a capacitor and a resistor that hang on the output of an amplifier change the amplifier’s open-loop-gain curve. The combination of the load capacitor, CL; the load resistor, RL; and the amplifier’s open-loop resistance, RO, introduces a pole to the open-loop-gain curve, and CL and RL then introduce a zero to the open-loop-gain curve (Figure 1). Creating this pole and zero does not disrupt the amplifier’s stability as long as they cancel each other out before the open-loop-gain curve crosses the closed-loop-gain curve. If the open-loop-gain and closed-loop-gain curves cross with a 40-dB/decade closure rate, the amplifier circuit will be marginally unstable or, worse yet, will oscillate.
|
You can find the pole and zero locations in this circuit in the following equations:
What have I learned from this situation? It pays to understand why an engineer’s rule of thumb works. If you comprehend the general guidelines, you will be OK. If you are not on top of the explanation, however, it will come back to bite you.
| References |
|
-
This might explain why most engineers need to know, not just what, but also WHY !
Shane Kennedy - 2008-2-12 14:49:00 PST -
Stevie Wonder said it best: When you belive in things that you don't understand, you will suffer.
Carl Spearow - 2008-2-12 12:37:00 PST -
Yup - it's a good one and it happens to power supplies too. They too have an increasing impedance with frequency so the addition of the 'right' capacitor can cause instability. I finally took the time to understand the problem too and wrote about it here in EDN in the Sept 1, 1989 issue. Good job Bonnie for reminding us all about this never ending issue.
Steve H - 2008-2-12 12:28:00 PST -
What is interesting about this: many engineers have come to regard opamp outputs as very low impedance---I mean after all, they are emitter followers (usually) aren't they?
Ah yes---but what is driving those, and what is its open-loop Z? Your e-followers' output Z is nominally the base drive Z divided by follower Q beta (and beta at what frequency?). Rare is the amplifier with an open-loop output impedance that is all that low.
Brad Wood - 2008-2-12 12:16:00 PST -
I can relate. Right out of college I was working for a small power supply company and told my boss that we should be building oscillators instead since it seemed like it was a lot easier to get the error amp to oscillate than to be stable. As I recall, that idea was not well accepted.
Neal E. Naumann - 2008-1-12 05:31:00 PST
