Indeed, I've done that since the 1970s and it also is very useful in sims. No need to apply full power square wave.
It's also a good place to attach a window comparator for a clipping detector. It will detect clipping, slewing, current limiting (if applicable) and anything else that causes gross differences between the input and feedback signals.
Yes, and add a pulse-stretcher to the comparator to give you time to actually see it go off ;-)
Jan
Jan
This is a TIM Otala/Curl method described 50 years ago. It works only in case of low OLG gain. In case you have OLG of 120 dB and 100dB of feedback, it is more than tricky to insert a probe tip to measure input Vdiff, not speaking about tiny amplitude.
A very efficient test to show SR limiting is a 2-tone CCIF test, 13+14kHz of 19+20kHz. Increasing input amplitude one easily finds the point of limit. And the dv/dt of twin tone is well known, it is a textbook school case.
I use a monostable multivibrator for that, it stretches the pulses to about 1 second. See Figure 4 of
"Audio power with a new loop", Electronics World February 1996, pages 140...143, https://worldradiohistory.com/UK/Wireless-World/90s/Electronics-World-1996-02-S-OCR.pdf
The six transistors on the bottom left are the window comparator (with current input).
Another way to watch slew rate limiting. Yes, LT1006 is one of the slowest opamps in the world.
@MarcelvdG Yes that will certainly work, although a simple 555 or even an RC time will also do for the purpose.
BTW Nice amp design; I saw it at the time, forgot about it and now it's in my face again ;-)
Jan
BTW Nice amp design; I saw it at the time, forgot about it and now it's in my face again ;-)
Jan
My experience with measuring slewing (a long time ago) was to measure the output of the VAS. Any crossover notch became quite visible on an analog scope by the discontinuity around the zero-crossing. The trace disappears, leaving two disjointed halves.
Ed
Ed
But xover distortion is not the same as slew rate limiting.
You can either one or both or none.
Completely different phenomenon and it also looks quite different.
Jan
You can either one or both or none.
Completely different phenomenon and it also looks quite different.
Jan
They are not the same, but crossover distortion "fixed" by negative feedback leads to slew-rate limiting. This was a common problem in early solid-state amplifiers.
Ed
Ed
Last edited:
I understand this problem completely. 🙂
A severe crossover notch causes slew-rate limiting when negative feedback is applied. Inside the notch, the amplifier is running open-loop. An amplifier usually has more open-loop gain than its slew rate can support. Hence, it slews (momentarily). The moral is to avoid having a crossover notch.
Ed
A severe crossover notch causes slew-rate limiting when negative feedback is applied. Inside the notch, the amplifier is running open-loop. An amplifier usually has more open-loop gain than its slew rate can support. Hence, it slews (momentarily). The moral is to avoid having a crossover notch.
Ed
I think you could still solve the slew rate limiting by making the linear range of the input stage large enough (although you still need to do something else against residual crossover distortion then, current dumping maybe, or the Elektor Edwin principle?).
Suppose your output stage is a class-C voltage follower of which the output voltage is sometimes 0.7 V too low and sometimes 0.7 V too high, depending on the direction of the current. You could then get a 1.4 V step in voltage when the current direction changes. With 20 times attenuation in the feedback network, that's a 70 mV step in the error voltage that drives the input stage. +/- 70 mV is not impossible to handle with some local feedback in the input stage.
I think this line of reasoning also holds with a resistive load, even though around the zero crossings, the output voltage will then change gradually and the voltage driving the output stage will have to go quickly from -0.7 V to +0.7 V or the other way around. You can still model the output stage as something that either adds or subtracts 0.7 V and the worst that can happen is that it switches between these modes.
Suppose your output stage is a class-C voltage follower of which the output voltage is sometimes 0.7 V too low and sometimes 0.7 V too high, depending on the direction of the current. You could then get a 1.4 V step in voltage when the current direction changes. With 20 times attenuation in the feedback network, that's a 70 mV step in the error voltage that drives the input stage. +/- 70 mV is not impossible to handle with some local feedback in the input stage.
I think this line of reasoning also holds with a resistive load, even though around the zero crossings, the output voltage will then change gradually and the voltage driving the output stage will have to go quickly from -0.7 V to +0.7 V or the other way around. You can still model the output stage as something that either adds or subtracts 0.7 V and the worst that can happen is that it switches between these modes.
Well, not to split hairs, but technically during the xover OL interval the amp is overdriven.
So it is driven into SR limiting by massive overdrive.
The SR limit is the same in OL as it is in CL.
But I think we agree.
Jan
@MarcelvdG - Yes, that is equivalent to setting x<1 in my post #67. It is not commonly done because higher loop gain along with a mostly linear output stage is better.
Another way to preclude slewing is to put the frequency compensation capacitor across the differential input. I have seen that in a few oddball amplifiers on this board.
@jan.didden - We agreed back on post #72. This thread has too many pages... 😉
Ed
Another way to preclude slewing is to put the frequency compensation capacitor across the differential input. I have seen that in a few oddball amplifiers on this board.
@jan.didden - We agreed back on post #72. This thread has too many pages... 😉
Ed
I don't think so. Your x < 1 means that the input stage can handle the full input signal, so it would not even be driven into clipping without any feedback (which would still be inadequate for zero risetime square waves with feedback, you may need x <= 1/2 for those). My criterion relates to the size of the dead zone and the feedback attenuation.
I meant "similar idea" rather than "mathematically equivalent". My formula is a rule-of-thumb since the gm of the differential pair is not constant.
Ed
Ed
Ed, made a long walk today and something got bothering me. We discussed xover situation, where gain collapses to (almost) zero.
That means that the amp internally will be hugely overdriven by feedback action. So far so good.
But I now think that this doesn't automagically mean the amp goes into slew rate limiting.
It will surely go into internal clipping, but whether it goes into internal SR limiting too depends on the actual SR capability of the amp and the amount of overdrive, ie the feedback factor.
Makes sense?
Jan