What is opamp overshoot and will I encounter it with audio frequencies?

[Max999]

I have been looking at the Opa1641 datasheet at the "small signal overshoot vs capacitive load" graphs. Since we drive pretty big capacitive loads with our guitar cables I would like to understand these graphs:



So the graphs are showing that during a small signal output step of a non inverting buffer the overshoot varies with the value of capacitor Rout.

Does this means the overshoot will only occur during a step response test or will I also have overshoot feeding this circuit audio frequencies?
My understanding is that overshoot is a signal that is a bit too loud at the start of the waveform but is later corrected by the opamp. Is this correct?

[PRR]

Well, how fast is it?

Fig 13 shows the overshoot in time. The overshoot is done in about 100 nS. In audio, anything under 50 uS is supersonic. So this blip is 500 times higher than the top of the audio band.

I've plotted a sine-wave against the transient response. I had to scale for a 1 MHz sine to show well; at 20KHz we would see about 2% of the sine, so it wouldn't even look slanted/curved. The actual overshoot is like a part of a 10MHz wave: that won't get far in a guitar system.

In addition, good audio hygiene suggests an output resistor on ANY output that goes to the Outside World. Who knows what will get connected?? Shorts happen; resistor limits the stress on the chip. As you found, NFB amps are unsettled with some capacitive (and 50-1000pFd are very typical of 2'-30' cords); with a series resistor the amp is not facing a bare-naked capacitance.

And there's no objection to "small" resistors in guitar-cord work (may be ugly in loudspeaker work). 50, 500, even 5K makes about no difference in on-stage work. (Long runs to the mixer in the bleachers should be few-hundred ohms not thousands-ohm.)

[Max999]

Thank you PRR for taking the time again to answer my questions. I see now that the graphs show that the chip will overshoot at a 10MHz wave with a capacitive load.
But they do not show that 10MHz is the onset of overshooting, correct?

Can I find in a regular datasheet which frequency will be the limit of what an opamp can deliver without overshooting with a capacitive load?

From a purely practical standpoint: have you ever encountered overshoot while using an opamp for audio frequencies ( 20Hz-20kHz) with a long cable ( say 1000pFd)?

[PRR]

I would just put 100r between opamp and jack, and not worry about it.

(TL072 might prefer >200r.)

[amptramp]

An op amp operates by applying feedback from the output to the inverting input so that any tendency to overshoot is minimized.  But the output impedance of the op amp and the capacitance forms a lowpass filter that prevents the feedback from reaching its final value immediately.  With the inadequate feedback, the output will continue to climb, resulting in an overshoot until the feedback is sufficient to settle the output at its quiescent state.  Often, you will see designers add a feedback lead - a capacitor in parallel with the feedback resistor which limits the frequency response of the stage but tames any tendency to overshoot.

[Transmogrifox]

The usefulness of the overshoot graph doesn't have a direct application to using an op amp for audio purposes, but it does have some implications.

If you're designing an optical receiver circuit that decodes square waves at 2 MHz then you might want to pay attention to the overshoot an understand what it might do when you send a pulse with a fast edge into the op amp.

For audio purposes, this overshoot behavior doesn't have any direct meaning.  Looking at these graphs vs capacitive loads gives you an idea of how much capacitance gets uncomfortably close to stability margins.

An op amp that is unstable at 10 MHz might still appear to be working like an op amp at audio frequencies, but it will exhibit odd behaviors that you don't expect (more noise than expected, distortion, and strange "floating" bias behavior.  This type of instability is hard to troubleshoot.  If you attach an oscilloscope probe to the circuit it will change the behavior or it might even stabilize it if connected to the right spot on the circuit.

Don't get hung up on 10V/us transient behavior when considering 10V/ms edge rates. There's a factor of 1000 here.

The best practice is to do what PRR says and use a series output resistor to help prevent instability.

[Perfboard Patcher]

The overshoot phenomenon isn't something bad that has to be avoided at all cost. A cabinet simulator with a 3rd order or higher low-pass filter at say  5kHz will most likely create overshoot in the audio frequency range. It's the price that's to be paid for applying heavy EQ.

[Max999]

Thank you amptramp, Transmogrifox and Perfboard Patcher.

@ Transmogrifox How do you know a circuit has a problem with distortion when you cannot hook it up to an oscilloscope?

@everybody

Are BJT's in simple configurations ( not discrete opamps) also unhappy with capacitive loading or is this reserved to op amps with huge amounts of "global negative feedback"?

I see that Boss usually ends the circuit with a BJT ef, even though there are sometimes opamp de-emphasis filters before it that to my understanding can already do the job ..