Fully balanced topoligies mean that we have in fact two standalone amplifiers working in parallel, one for the positive, and the other for the negative part of any sine wave.
This topology has some advantages, like being able to work off lower voltage supply lines, which is always good for the output devices, and the fact that it doesn't rely on the ground for a reference point. Hence, any junk in the ground cannot influence it, nor can any change of ground potential influence its operation.
But it has some serious drawbacks too. One is that for it to realy work well, you have to have almost perfectly matched amplifiers, which is never easy to do. However, I must note here that some very interesting topologies have been shown to work, and work well - for example, James Bongiorno's Sumo 9+, made way back in 1982. Exemplary work, that one - simple, ellegant and efficient, and above all, it sounds fabulous even after all these years.
The second drawback is that it is usually (but not always) more complex than a single amp. More parts mean less reliability by default, more chance of something going wrong.
Fully complementary is somewhat similar, but not the same. This assumes that is every NPN transistor has a complementary PNP transistor, their inherent differences will tend to even out and significantly reduce even order harmonic distortion (2nd, 4th, 6th, etc harmonic), leaving only odd ordered harmonics (3rd, 5th, 7th, etc) to be minimized by the negative feedback.
To the best of my knowledge, this topology was first used by - hang on, now! - James Bongiorno in 1971.
It has been shown that the basic assumption of reducing even harmonics is, or can be, true.
This is the only place where Dan Banquer and I part company (figuratively speaking, of course), because Dan uses this topology, and I don't, obviously because he thinks it's better, and I don't. I may have the greatest respect for Dan's work as I do, but I do not believe this approach is inherently better than many others. It does reduce the even harmonics, but in my view, the best of amps is the one in which all forms of distortion are equally present. Feedback is essentially non-selective, it feeds back EVERYTHING, although of course, it can be tweaked by altering its characteristics, using dual feedback lines (what Luxman in its day called Duo-Beta) with very high feedback for DC to keep the amp's DC balance in good shape and less feedback for AC, adding non-linearizing elements (the simplest of which is bypassing the feedback resistor with a capacitor, so feedback rises above a certain point, thus enlarging the amp's linear bandwidth), etc. Heck, DC servo is essentially a second feedback path, with tremendous feedback below a certain point, typically below 1 Hz, and zero feedback above say 5 Hz.
I believe that the inherent differences between NPN and PNP transistors will always be such that a fully complementary topology will never really improve matters to the point of being clearly superior. For one thing, it uses more transistors, each one of which is a possible source of problems. To me, it's a classic case of quid pro quo, gain here, lose there.
Mind you, this is just my personal view, and if truth be told, then we must note that some of the best amplifiers of our times have used this topology - just as other best amplifiers of our times did not use it.
Another interesting topic here is feedback - yes or no? I refer to overall feedback, which is taking a part of the signal from the amp's output and feeding it back to the input stage.
Some, and particularly the modern Scandinavian school, maintain that zero feedback is better becouse of superior transient response. The idea is to use local feedback around each stage, the rationale being that distortion should be dealt with where it appears, right there and then. It is possible to obtain large bandwidths, low distortion, good damping factor and high slew rates without overall feedback, and these are usual reasons for feedback, which makes one wonder why do we need it, if we can achieve our goals without it?
Superficially, this is all true as far as it goes - but it doesn't go far enough. True, you can have low overall distortion, high slew rates, good damping factor and a large bandwidth without it by applying local feedback. Therein lies the first potential danger - to get it, you may have to ovedo local feedback, do on micro scale what you are trying to avoid on the grand scale. But if you spoil the sound of any one stage of the amplifier, the end result will invariably be a poor amplifier.
The second problem lies in subtle mismatches between stages, between transistors, components, etc. Since this is dealt with locally, formal distortion is taken care of, but slight mismatches will still be in our output signal, with no overall feedback to cancel them, or at least reduce them.
Personally, i don't like the sound of zero feedback amps. I have heard quite a few, and not one sounded natural to me, all had one fault or another. If they had anything in common, it would be a two dimensional sound stage, sort of flat. Detailed, precise, but flat. And that I just can't stand.
My ideal are the so-called low feedback amps. What is high and what is low is of course relative, but over the last few years, it has been accepted that overall feedback of 20:1 (26 dB) or less is low feedback. In 1993, Harman/Kardon's 6550 integrated amp had an overall feedback of just 14 dB (5:1), and that was a commercial unit not even pretending to be high end, though it was more expensive than most in its power envelope. In 1999, Harman/Kardon's 680 integrated amp had an overall feedback of 10 dB (3:1). Both sound really good, believe me, I have both at home. But in the late 70-ies, feedbacks of 40-60 dB (100-1,000:1) were quite common.
In my view, the PROPER way to use feedback is to make the amp as linear, as wideband and as low distortion as possible without any overall feedback - then add just enough feedback to make something good even better yet. That H/K 680 is a perfect example - without any feedback, its distortion is about 0.5% and it goes above 85 kHz open loop (no feedback). Then they add just 3:1 feedback to get the distortion to below 0.09%, and the bandwidth out to beyond 250 kHz. Just what the doctor ordered.
To know how it's possible, remember that their head designer, Richard Miller, worked with Prof Matti Otala for 12 years, 1974-1986, producing such outstanding designs as Citation XX. Matti Otala identified Transient Intermodulation Distortion (TIM), and proposed methods to measure it, and developed tests for it and showed ciruits rid of the problem in 1972. Then he talked about loudspeaker load complexity, how it works and what to do about it. In short, together with James Bongiorno, he's the most influential person on this planet alive regarding power amplifiers.
Which brings us to the most crucial issue - what do we design the amplifier for? Speakers, you say? Ah-ha, but how do speakers behave? What will the amplifier actually see as its workload?
Will the amp make it? Will the Speaker triumph over our hero? Watch this space for more revelations, same time, same channel.
Revealingly,
DVV
