I have always thought that there was an interesting difference between negative feedback (nfb) and error correction (ec) in an amplifier loop1). But it's more subtle than that.
In the article I give some equivalent conceptual circuits that show the embedded positive feedback loop within a H.ec loop. At a cursory glance that may give the impression that there is infinite loop gain. However, that is only the case if the forward gain would be zero which is not a realistic condition. In fact, it can be shown that the loop gain varies directly in proportion to the forward gain error (thank you Mike). Thus, when the latter would be zero, there would be no feedback at all.
Negative feedback relies on excess gain in the main amplifier (forward)
loop. By returning a part of the output signal to the input, the amplifier
gain becomes much smaller, but also distortion becomes much smaller. Output
impedance becomes much smaller, and input impedance becomes much higher. In
other words, the amp becomes much more ‘ideal’. So this is a kind of
trade-off: we give up most of the large (open loop) amp gain to get a much
more linear and ideal result. Sounds pretty neat, but what about that (very)
large open loop gain we need to make nfb work? That in itself can be a
source of problems.
Nfb is quite old, it was invented by Harry Black (see pic) in 1934. If you are interested in how this works in more detail, here’s his original paper.
My implementation is different, both for the output stage as well as for the voltage amplifier stage. The output stage circuit is provided courtesy of Elektor Electronics.
Error correction is, in my
view at least, more ‘elegant’ than negative feedback.
Error correction looks at the difference between the amp output and the amp input (taking into account of course the amp gain). Any difference is returned to the input and added to that input, in precisely the right amount to make the output identical to the input. Because the output is now identical to the input, distortion becomes ideally not just smaller, but zero. (In practice, this will not happen, but more on that later). Also, whatever the load, the output stays correct, so it looks as if the output impedance is really zero (or close to it). This technique does not require that the amp (open loop) gain is very, very high. It is enough if the amp has a gain close to the final required gain, and then the ec fills in the missing fraction in the signal. See here for a conceptual approach.
Next Running the amp at lower power
1) Of course nfb can
also be classified broadly as a specific way of error correction. But when I
refer to ec I mean the topology as presented by
Malcolm Hawksford in his
original
publication on error correction.
