You may say whatever you want, but I would prefer to compare results of your work with results of my work. To hear the difference, which power amplifier driving speakers cleaner and more realistically reproduces the recorded material, which better fools imagination as if no amps nor speakers present.
As I said, minimization of THD per se to me does not matter. It probably was a problem back in 1960'Th. What matters, how to get minimum of audible distortions, instead of going for something that does not matter. And in this quest, my variables are more flexible than just hypothetical stages in symbols; my variables include versions of topologies, with certain properties of certain components, that in your "common theoretical" case can be parasitic, while in mine they are usable.
After half a century of trials and errors I came to very simple topology. It is like to realise finally that a truck is more convenient to move a grand piano than forty bicycles tied by ropes.
https://youtu.be/skCCd4O-fVQ
As I said, minimization of THD per se to me does not matter. It probably was a problem back in 1960'Th. What matters, how to get minimum of audible distortions, instead of going for something that does not matter. And in this quest, my variables are more flexible than just hypothetical stages in symbols; my variables include versions of topologies, with certain properties of certain components, that in your "common theoretical" case can be parasitic, while in mine they are usable.
After half a century of trials and errors I came to very simple topology. It is like to realise finally that a truck is more convenient to move a grand piano than forty bicycles tied by ropes.
https://youtu.be/skCCd4O-fVQ
It depends on which end result you want. May be on certain power level THD will be the same, but when stages misbehave, they misbehave differently, including RC coupling, transformers, resulting in different kind of distortions, that are differently audible.
Edit: I spilled enough of beans. I have no interests in proving my point to theoreticians operating a ball-shaped horse in vacuum. Want to prove your point? Welcome to contest, let's music speak for us!
Last edited:
Like Waly said, in a global loop system if one of the internal stages overloads, you can lose one of the zeros in the feedback loop (or gain a pole, and any number of weird things) and then your amp is unstable. In very complex designs often you can't get past the simulation stage because there is no way to prevent a certain stage from latching into a bad phase condition.
Which is probably one of the reasons miller compensation is by far the dominant compensation strategy, it is resistant to such mishaps with conventional designs and devices.
To get around the problem you must somehow ensure that no stage when overloaded causes phase issues. But every stage has lots of quirks such as asymmetrical saturation points, overload recovery times, variable input/output impedance, etc. You can try clamps, shorting diodes, collector resistors, all sorts of stuff trying to get a stage to behave. The problem is that at 1MHz or 10MHz, whatever your ULGF, many stages just can't be made to behave under all conditions.
If Bruno used nested feedback loops to achieve what he did, it doesn't necessarily mean it was better, it may just mean that was the only apparent way it could be done.
Didn't Bruno have a whitepaper or something about this?
Which is probably one of the reasons miller compensation is by far the dominant compensation strategy, it is resistant to such mishaps with conventional designs and devices.
To get around the problem you must somehow ensure that no stage when overloaded causes phase issues. But every stage has lots of quirks such as asymmetrical saturation points, overload recovery times, variable input/output impedance, etc. You can try clamps, shorting diodes, collector resistors, all sorts of stuff trying to get a stage to behave. The problem is that at 1MHz or 10MHz, whatever your ULGF, many stages just can't be made to behave under all conditions.
If Bruno used nested feedback loops to achieve what he did, it doesn't necessarily mean it was better, it may just mean that was the only apparent way it could be done.
Didn't Bruno have a whitepaper or something about this?
Yes, on this exact forum his paper was discussed. His major achievement was, when he surrounded an output filter by a feedback loop.
Fair point, as long as you stop glossing about the magic properties of nested feedback loops and other pseudo-technical arguments. You (and perhaps many others) like the sound of your amp, I guess nobody will debate this, as much as nobody debates it is overall a good amp.
A lot of feedback theory discussion spins around the assumption that you can just choose a compensation scheme that affects the stage gain in a specific way and then distortion will be divided by that gain when the loop is closed. It hardly ever works that way after you get under 0.01% distortion. Every compensation component in the feedback loop shunts a stage and reduces gain, but as well as that, it conducts distortion through it which depends only on the type of distortions present at the nodes it is connecting. This cannot be dealt with theoretically in a block diagram, only empirically in individual designs.
Another way to say it is that the Miller capacitor is a compensation component, yes, but it is also a capacitor shunting the input of the output stage, and which conducts distortion from the output stage to the VAS input. The consequences of the capacitor on output gain and phase can be calculated without knowing the exact circuit, but the affect on distortion cannot. Most feedback discussions seem to omit the latter as if it doesn't exist, and carry on expecting gain/pole/zero calculations to reflect what should be optimal for a given design.
Nothing is magic, just usage of properties of components and topologies for the final result, minimizing audible distortions. I repeat again and again, that I am minimizing audible distortions, so amps and speakers as if disappear, instead of trying to build a truck from 40 bicycles with global feedback loops and compensation that leads to nasty behaviour of amps. Audiophiles are not lying. They just can not explain what they hear using terms that you understand. Synaesthesia...
And, let me repeat, no way I like the sound of my amp. What I like, almost absence of sound that it adds.
Last edited:
To continue my previous post...
So the answer to your question is really, whether or not nested or global feedback is better depends on which configuration conducts the least distortion across the stages of the amplifier. Maybe a global loop will work best with the given topology, or maybe a nested feedback design is better.
Furthermore, after you have closed a global feedback loop you can often make improvements adding extra loops such as bootstraps, which simply cannot be made global loops. It is physically impossible. It doesn't make any sense to omit these extra improvements just because "global feedback is superior to nested".
Ultimately feedback theory is constrained and broken up by topology. You cannot do all the same things with a Blameless amp that you can with objects in a feedback block diagram.
Very good point!
In this design I used nested positive feedback loops by current.
https://www.youtube.com/watch?v=iZ4BDSrHkxI
https://www.youtube.com/watch?v=qsb2Hzq7sD8
Ok, I'll bite: what are "audible distortions", how to you define this concept for the use of a circuit designer, and how is it different from the known harmonic distortion concept? Inquiry minds want to know.
I don't think Bruno published on that particular subject. I asked him once to write it up for Linear Audio and he replied in his friendly manner that 'I have no desire to educate the competition'. 😎
He did indeed include the output filter and it was a 6th order loop iirc.
Edit: could have been 5th order, I will yield to better memory.
Jan
Last edited:
Oh yes, the technical and audio qualities are so obvious from these videos. Are you serious, or just trolling around?
Mark, I do actually have that book, the section is on page 49. I was hoping that the separate paper has some more detail about it.
Rosenstark, S., Loop Gain Measurement in Feedback amplifiers, Int J. Of Electronics, vol. 75, no 3, pg. 415- 421, March 1984.
Jan
Obviously it's impossible to prevent all forms of clipping, but I'd love to see a count of y'all that regularly clip material when listening at home. Maybe my source material is too low crest factor, but I am usually attenuation below a commercial (commonly g=20) amplifier. I.e. headroom galore. Not going to clip.
The amps I'm designing are going to be as close to unity gain as I can stabilize. And run off some 12v transformers I have laying around. 10 watts or bust.
The amps I'm designing are going to be as close to unity gain as I can stabilize. And run off some 12v transformers I have laying around. 10 watts or bust.
yes. that is true. and can be tested in SIM of VFA model to compare the OLG results with and without the big C bypass. However....
In practice and esp with an unknown amp design to quickly get an idea of the OLG it is fairly close over most of freq range. That is how I used the suggestion.
[ I haven't seen any instabilities but some designs may not like it, I imagine.]
If only talking about exact OLG results, esp at high freq end then its back to theory; Middlebrook or others.... http://www.arrl.org/files/file/QEX_Next_Issue/Mar-Apr_2011/QEX_3_11_Post.pdf
THx-RNMarsh
Last edited:
- Status
- Not open for further replies.