Amplifier recommendations for 3A Signatures

[atmasphere]

On 9/18/2022 at 2:26 PM, Richard Vandersteen said:

While this is true there are some similarities.  In a linear amp class A/B you have a continuous signal with power gain.  In class D the signal is sampled at a high speed then fed to the output stage as a high-speed square wave which is modulated towards a positive signal or negative signal by changing the on time and off time.  The signal is then filtered to remove the high frequency square wave carrier and what is left is the reconstructed music waveform with some inherent errors, timing especially. Very efficient, cost effective, great sounding but lacking the emotional connection of many linear feedback free amplifiers on analog sources, YMMV.  RV

Class D can be done zero feedback, some feedback, or really a whole lot of feedback (usually the latter will be a self-oscillating design). You do want to include the output filter in the feedback loop so the feedback can correct phase shift introduced by the filter.

The problem with feedback in traditional solid state and tube amplifiers is that the point where the feedback is applied (in a tube amp, often the cathode of the input tube, in a solid state amp, often the base of a transistor in a differential pair) is not linear. What this means is that the feedback signal gets distorted before it can be mixed with the input signal. This is why its introduction in most amps causes higher ordered harmonics to show up, as well as intermodulations, all caused by the feedback point itself.

That is why feedback has a deserved bad rap in high end audio. But there is a solution, a brute force approach, which is to apply so much feedback that this problem is overcome. To do that you need in excess of 30dB. That might not seem like a lot but if you look at most conventional amplifiers the typical amount is 15-25dB. That isn't enough; 15dB puts you in a bad place insofar as all that distortion I mentioned is concerned.

You have some challenges to adding 30dB or more of feedback! First, you need that much gain you can blow off, plus the gain you need to make the amplifier practical; most amps need at least 25dB of gain for that. So at a minimum to make this work you need the 30dB for the feedback plus 25dB more to function as the amplifier's gain. That's at least 55dB which is quite a lot. BTW that value is known as 'loop gain'. To further complicate matters, all amps exhibit phase shift due to frequency poles in the design. The phase shift will cause the feedback at some high frequency to become positive rather than negative. This is the 'phase margin' of the amp and to avoid oscillation it must not be exceeded. When you have lots of loop gain and +30dB of feedback that's really hard!!

With all that gain you have trouble making bandwidth- the more gain, the harder it is to do. There is a design consideration called Gain Bandwidth Product that is the a mathematical derivation of both gain and bandwidth of the circuit, just as it sounds. If there's not enough, feedback will fall off at a certain frequency and distortion will rise starting at that frequency (resulting in harshness and brightness). For this reason, since getting enough GBP is impossible in a tube amp, we never ran feedback in our OTLs. Nearly all transistor designs in the last 70 years lack the GBP as well. 

Class D offers a simple solution: go ahead and have all that feedback, and exceed the phase margin. The amp goes into oscillation right away as soon as its turned on; just make sure that the feedback loop only allows for one frequency at which the oscillation occurs. That is used as the switching frequency of the amp! This technique allows for a simpler design (killing two birds with one stone) and a whole mess of feedback, plus  you can get sufficient GBP to support that feedback at all audio frequencies.  Now the distortion is lower and doesn't rise with frequency.

If you paid attention to your Ps and Qs in the design, the result can be a class D amp that is just as involving as any tube amp. Obviously there are a lot of class D amps out there where the Ps and Qs didn't happen and have turned a lot of people off to the technology! It is a mistake though to assume that if you heard one you've heard them all. IME the 'sound' of class D amps varies more than it does from the worst tube amp you can think of to the best. That's a pretty wide variation!!

18 hours ago, nrenter said:

Any amplifier that uses feedback to stabilize the the circuit is introducing “timing” issues into the signal, as the (negative) feedback is attenuated and added back to the source signal (180 degrees out of phase).

This statement is problematic! First off if a circuit is unstable feedback is only going to make it worse. Negative feedback is a destabilizing factor; if you want to see a circuit in a stable state, put positive feedback on it so it oscillates (or don't use any feedback at all). That will be a stable condition... and that is how a self-oscillating class D amp does it. Since its already oscillating, weird loads can't somehow make it oscillate further.  At any rate, 'timing' is really an aspect of phase shift. On an oscilloscope it will appear that the phase shift has altered a particular frequency in time. But if you have enough feedback it can alter the phase of the signal since if it does not agree with the input, it gets correction. But at lower feedback levels this does not occur so much.

 

 

Edited September 21, 2022 by atmasphere

[GdnrBob]

@atmasphere, Thanks for clearing this up. I don't have the vocabulary, nor tech savviness to explain it as well as you have done.

 

But, 

3 hours ago, atmasphere said:

just make sure that the feedback loop only allows for one frequency at which the oscillation occurs. That is used as the switching frequency of the amp! This technique allows for a simpler design (killing two birds with one stone) and a whole mess of feedback, plus  you can get sufficient GBP to support that feedback at all audio frequencies.  Now the distortion is lower and doesn't rise with frequency.

Does this affect time and phase? Or, does it matter?

Bob 

Edited September 21, 2022 by GdnrBob

[Holmz]

25 minutes ago, GdnrBob said:

,,,,

Does this affect time and phase? Or, does it matter?

Bob 

 

I think this says that the feedback corrects it??
(Provided that the feedback is enough)

 

3 hours ago, atmasphere said:

At any rate, 'timing' is really an aspect of phase shift. On an oscilloscope it will appear that the phase shift has altered a particular frequency in time. But if you have enough feedback it can alter the phase of the signal since if it does not agree with the input, it gets correction. But at lower feedback levels this does not occur so much.

 

And it differs from Goldilocks in that there is no “too little”, “too much” and “just right”
We either want none or enough… much in the same way that the squirrel is OK on the left side of the road or the right side. (But the average place of the middle of the road is not the place to be.)

Edited September 21, 2022 by Holmz

[nrenter]

Thanks @atmasphere! You are one of the few contributors make me still peruse the various audio forums. I appreciate you.