Overshoot and Ringing

[Captn Dave]

I need some help understanding the difference between overshoot and ringing. I read about the two as if they were related so I'll just mention that I understand that overshoot is the tendency of the amplified wave form produced from a square wave to overshoot and then oscillate a bit at the beginning of the square wave. I understand that by experimenting with some silver mica caps bypassing the negative feedback resistor one can eliminate the overshoot.

So assuming that's a relevant starting point, can someone please explain the concept of "ringing" to me?

Thanks

[nigelwright7557]

Ringing is an oscilation slowing decaying.

Overshoot is a signal passing past where it should stop.

In a class AB amplifier (none quasi) you usually need a capacitor in the VAS to stop overshoot/ringing on the output.

In a class AB (quasi) you also need a capacitor on the lower driver CB to stop overshoot/ringing.

Thats is certainly my experience.

Ringing/overshoot are often found where inductors are being driven like a speaker.

[smoking-amp]

Both phenomena are caused by an almost resonant response (oscillator with insufficient loop gain to oscillate steadily, but gets fired off by a transient still) from the phase shifted feedback (phase delay from the OT leakge L mainly) . If the resonance is critically damped (loading Z = to omega L = 1/ omega C), then it only overshoots once and settles back to the new state. If under-damped (insufficient loading Zload > omega L...) then it "rings" or oscillates with a decaying amplitude. The overshoot only case simply decays fast enough to avoid another cycle.

The "speedup" capacitor advances the feedback phase enough to eliminate the initial phase shift. No phase shift, no oscillation tendency.

[HollowState]

Overshoot on a square wave or pulse (that's what a square wave really is) refers to the rising (or descending) leading edge where it exceeds the average amplitude. It is indicative of excessive or uncontrolled high frequency response.

Ringing, with reference to the above, is a series of oscillations that follow the initial overshoot. This is usually do to the resonant effects of a coil's inductance.

[Captn Dave]

Thank you all for the most-helpful responses.

Please allow me a couple of follow up questions.

Originally Posted by smoking-amp

If the resonance is critically damped (loading Z = to omega L = 1/ omega C), then it only overshoots once and settles back to the new state. If under-damped (insufficient loading Zload > omega L...) then it "rings" or oscillates with a decaying amplitude. The overshoot only case simply decays fast enough to avoid another cycle.

The "speedup" capacitor advances the feedback phase enough to eliminate the initial phase shift. No phase shift, no oscillation tendency.

You say that "If the resonance is critically damped". Can I substitute the word correctly for critically or is there a distinction?

Your explanation covered two cases, the critical and the under-damped. It seems to me that the third case is over damped, and in my experience that causes the phenomenon described in HollowStates drawing as rounding. Have I got that right? If so, what are the sonic effects of these three cases?

I experimented with a little single ended amp I'm building and for the life of me I couldn't discern any difference; granted my bench speakers are not the right tool for that work.

Originally Posted by HollowState

Overshoot on a square wave or pulse (that's what a square wave really is) refers to the rising (or descending) leading edge where it exceeds the average amplitude. It is indicative of excessive or uncontrolled high frequency response.

Ringing, with reference to the above, is a series of oscillations that follow the initial overshoot. This is usually due to the resonant effects of a coil's inductance.

So I take it that overshoot is possible without ringing. I have yet to see that in the flesh. (I can't conceive of ringing without overshoot - I read your explanation to indicate that overshoot is a prerequisite). That would be from excessive high frequency response without the resonant effects of the coil. That leaves me wondering what happens with resonant effects of the coils inductance without the high freq problems. That wouldn't be motor-boating would it?

Lastly, is the addition of a few hundred picofarads of capacitance bypassing the feedback resistor the only tool in our box for correcting overshoot and ringing? It occurs to me that the resistor value could be monkeyed with, but after reading Morgan Jones on the subject I was left with the idea the value is best determined mathematically. Is that a fair statement?

Finally, is that drawing you posted part of a source material available on-line. If so, I'd go for a link.

Originally Posted by nigelwright7557

In a class AB (quasi) you also need a capacitor on the lower driver CB to stop overshoot/ringing.

Nigel, I don't understand the reference to "CB". I'm drawing a blank. The rest I got.

Thanks for the assistance, professors. It's greatly appreciated.

Dave

[DF96]

"Critically damped" is a technical term with a precise meaning. Whether this is the correct damping depends on the situation, and what you are trying to achieve.

Motor-boating is low frequency oscillation. It is the LF equivalent of HF oscillation, and is usually caused by different circuit elements (apart from the OPT which tends to figure in stability issues at both ends of the spectrum).

[mige0]

Dave - all your questions are answered by a (online) filter calculator.
Have a look at the step response and you are there.

There simply is no way of being "correct" - its all a question of bandwidth and filter type involved.
step response plots tell you what you get.

Sadly - no one will be able to tell you what sounds best (besides avoid excessive overshot/ringing) - and this is true for *any* part of audio design where filter slopes appear (and there are many many many - down to the PSU and even the power grid behaviour at current demand)

Michael

[HollowState]

Originally Posted by Captn Dave

So I take it that overshoot is possible without ringing. I have yet to see that in the flesh. (I can't conceive of ringing without overshoot - I read your explanation to indicate that overshoot is a prerequisite).
Lastly, is the addition of a few hundred picofarads of capacitance bypassing the feedback resistor the only tool in our box for correcting overshoot and ringing?
Finally, is that drawing you posted part of a source material available on-line.

Yes, overshoot is very possible without ringing in any case where there is no inductance involved, or virtually none. A simple example would be the compensation of an oscilloscope 10X probe. The probe is connected to a 1KHz square wave and the compensating capacitor adjusted for a sharp square leading corner without rounding or overshoot.

Technically speaking, overshoot is not a prerequisite for ringing, although it may sometimes be there. Outside of audio amplifiers, ringing can be a useful thing. Sound effects, frequency calibration and transformer testing to name a few. From my TV servicing days, ringing high voltage transformers was a good test of the winding's integrity. Loosely coupling a sawtooth waveform to the scope vertical input along with the winding in parallel produced a damped ringing waveform something like the one shown below if it was good.

Overshoot and/or ringing from the OPT can sometimes be helped with a small amount of C across the primary.

The picture I posted was to depict some common pulse terminology. I scanned it from Tremaine's Audio Cyclopedia. But any good text on pulse technogoly would have it.

[tomchr]

Wikipedia has some pretty decent articles on overshoot and ringing as well.

~Tom

[mige0]

Originally Posted by HollowState

Yes, overshoot is very possible without ringing in any case where there is no inductance involved, or virtually none. A simple example would be the compensation of an oscilloscope 10X probe. The probe is connected to a 1KHz square wave and the compensating capacitor adjusted for a sharp square leading corner without rounding or overshoot.

Compensating a probe is a completely different task.
Its the task of *removing* a bandwidth limiting slope by making a divider resistive only.
Certainly the only "perfect" filter is no filter at all.

Again - either play with a filter simulation tool or read a book about filter theory to get a feeling regarding step response and overshot / ringing. The basics are pretty simple - but complexity of details easily sky rock with higher order

Michael