John I can agree to most of this, and it did flash my light to something I tend to forget.
There are actually two mechanisms (at least...) at work that increase distortion in a feedback amp with frequency.
Firstly, with rising frequency, the open loop gain drops and thus the amp needs more input signal for a specified output level. That, as you say, means more non-linearity from the input diff pair.
Secondly, because the open loop gain is lower, the feedback becomes less effective at suppressing this distortion.
If you look at the rising distortion with freq in a common amp, it rises pretty fast - and this dual mechanism could be the reason.
So, let us degenerate the input pair to extend the linear range. That means less distortion with rising frequency. BUT, the degeneration (all other things being equal) also decreases the loop gain. So what will the result be: more distortion, less distortion, no change? ;-)
But in any case, my question was, if you have an amp that does not under any circumstances (clipping excluded) become overloaded internally in an open loop situation, does that then mean, it will also never be overloaded when you close a feedback loop around it?
Remember, the premise is that it will be internally linear within the audio band and below clipping. Is there a scenario where feedback will cause internal overload? I can't think of any, but that may be my limited imagination ;-)
Jan
I don't know about Leach, but I (and others) have a perfectly linear amp (in this context) that rolls off open loop at 10Hz. You say I need at least 100kHz. That's a cognitive dissonant if I ever saw one. You can't have it both. Make up your mind ;-)
To paraphrase Meatloaf - What's it gonna be boy ?
Internally, an amp looks like a low-pass filter.
High frequencies are not time-delayed by this filter but only attenuated.
Bar the propagation delay due the electrical path, the output of the amp starts to react as soon as a signal arrives at its input.
"Remember, the premise is that it will be internally linear within the audio band and below clipping. Is there a scenario where feedback will cause internal overload?"
Yes. If you switch to time domain and consider transients . If open loop is as slow as 10 hz you can get internal overload. Just read Leach on low TIm amps.
Yes. If you switch to time domain and consider transients . If open loop is as slow as 10 hz you can get internal overload. Just read Leach on low TIm amps.
Exactly! People make the same mistake with capacitors; a capacitor ´delays the signal´. Not so! A capacitor voltage starts to change at the very instant the current into or out of it starts to change! But it does change the phase relationship.
Do a sim of a square wave current into a cap, you will see a triangular cap voltage with inflection points coinciding with the square wave rising and falling edges.
Jan
If your transients are outside the audio bandwidth, they violate the premise.
If the transient is within the audio bandwidth, please show how that is different from a ´normal´ audio bandwidth signal.
BTW I have corresponded enough with Leach to know what he thought. I don´t think that route leads to convincing me.
BTW2 You still come up¨with arguments that are not evident in real amps. My 10Hz OL opamps never overload internally as long as I don´t clip them and don´t hit them with supersonic signals.
BTW3 Time domain and freq domain are absolutele identical twins. You can´t have a problem in one and not in the other and vice versa.
Jan
Last edited:
I don't are a electronic designer so, if this is a stupid question, please, don't shoot me.
But, if the amp work well in open loop, what is the reson for use feedback? In other words, if I are healthy, for what to eat medicines?
The valve amps work nicely without feedback.
Thanks
Last edited:
If bandwidth is increased, the upper pole where 180 degree phase shift sits will rise. This delivers a higher open loop gain cutoff, because with less compensation the lower pole moves above the audio band. Put another way, the loop gain is high even at 20KHz, and therefore the THD will not increase within the audio band.
As a matter of fact, Leach would, and did. Please see the attached screen shot of his Low-TIM amp page. I built my flavor of Low-TIM amp because of it exactly.
Attachments
Last edited:
You keep confusing signal transit time, phase accumulation and open loop bandwidth.
And switching between time and frequency domains won't solve your basic misunderstanding.
A short but very informative paper on this point has been proposed by Marcel van de Gevel :
http://linearaudionet.solide-ict.nl/sites/linearaudio.net/files/volume1ltemvdg.pdf
There has been evolution of thinking about TIM since Ottala and Leach.
I have an issue with the first highlighted sentence. It assumes that bandwidth is linearly correlated to phase (aka delay) profile. It ain't necessarily so. A cursory inspection of "delay lines" would show a strong counterexample.
Another standard beginners error. People confuse the dead zone in a class B amplifier (leading to a zero loop gain, therefore no feedback correction is possible, as you said) with the gm doubling effect in class AB amplifiers (causing nonlinear distortion) and call them both "crossover distortion".
Hence the wrong analogy with digital circuits. Digital propagation time is always defined in a capacitive load, however what really matters there is the time required to reach a certain output level. This is of no concern in analog circuits, up to a phase shift. Now, if you want an audio amplifier with a close to zero phase shift, then you may be concerned by the signal "delay" in the Miller compensation cap.
Yes, it is called phase delay. Of no concern, if not confused with the group delay.
https://en.m.wikipedia.org/wiki/Group_delay_and_phase_delay
Like it or not, they are correlated. Audio amplifiers are minimum phase systems. Therefore, gain and phase are not independent.
This thread has the highest concentration of common errors, pitfalls, myths, that I have had seen lately. Beats by far the BT thread.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.