High levels of overall negative feedback are regularly criticised for several reasons. The most common idea is that relatively innocuous second harmonic distortion without feedback is converted into more unpleasant higher order harmonics when feedback is added. This was analysed in detail by Peter Baxandall, 'Audio power amplifier design-5', Wireless World Dec.1978 p53-56. (An even earlier analysis by M.G.Scroggie appears to give a very different result, but see footnote at the end of this page.) For a square law fet amplifier at a certain signal level the levels of various harmonics were plotted as a function of feedback loop gain. Rather than copy the graph (and infringe copyright) here is a table of harmonic levels, based on both calculations and measurements:
A real fet is not a perfect square-law device, and so even with no feedback there are low levels of third and higher harmonics. It can be seen that the original second harmonic reduces in proportion to the feedback loop gain, and 40dB feedback reduces this distortion by 40dB. The 3rd harmonic however, originally at -55dB, increases by 5dB up to -50dB with just 10dB feedback loop gain. 4th and 5th harmonics are increased even more. If this small amount of feedback can do such damage it would be easy to stop here and conclude that feedback is a bad idea. If, however we increase the loop gain further we find that at 20dB gain the 3rd harmonic is now 2dB lower than at zero feedback, an improvement at last. The 4th and 5th harmonics however are still higher than at zero feedback. Onwards up to 40dB feedback, and here all harmonics listed are now below their original level, and if we continued increasing feedback they would all now decrease in proportion to the increase. i.e. another 40dB feedback would reduce all these harmonics by close to 40dB. If we had listed harmonics up to the 10th we would find that it was originally at a very low level, maybe -200dB, but would only fall below this original value at some high level of feedback, maybe 50dB. For any harmonic there is some level of feedback at which it falls back to the zero feedback level, and beyond which it continues to fall in proportion to added feedback. The higher the harmonic the greater the feedback needed to reduce it, but in the example used here the 6th and all higher order harmonics appear to remain below -120dB for any feedback level over about 40dB.
The conclusion from this is that for this example any low level of negative feedback will increase high order harmonics significantly, and the idea that feedback loop gain should be kept low is the exact opposite of the truth. There is a reduction of the significant high order harmonics from feedback when it is at a high level. Avoiding feedback altogether even for a theoretical perfect square law amplifier does not prevent high order harmonics being added, for reasons explained in part 2 of this series.
In reality open-loop audio amplifiers are rarely square-law devices with -22dB second harmonic. It is only when a high level of output distortion is fed back to the input and the amplifier non-linearity mixes it with the original input that significant higher order components can be generated. If we start with sufficiently low open-loop distortion low levels of feedback may have little if any damaging effect. Starting with higher open-loop distortion (within limits, see the footnote) we can use very high levels of feedback to very effectively reduce it to insignificant levels. There are good designs using both approaches, but the results from my simple mosfet designs convince me that the high feedback option is by far the simplest solution.
My own high feedback mosfet amplifier uses over 80dB feedback loop gain, but extracted distortion when driving a speaker load with a music signal revealed absolutely nothing unexpected. The distortion when amplified and listened to alone was inaudible somewhere below the noise and uncancelled music signal.
1). In 1961 M.G.Scroggie published what at first sight appears to be an analysis of almost exactly the same circuit arrangement used by Peter Baxandall. This was reprinted and updated in Wireless World, Oct 1978, p.47-50, the author writing under the name 'Cathode Ray'. A square law device has overall feedback applied, but with 40dB feedback there is now over 7% third harmonic, 3% 4th, and so on. I have seen this result used as a justification for using only low levels of feedback, so the question of why this version was so much worse deserves some consideration. Reading the entire article rather than just the distortion figures the explanation is clear. First, the open-loop distortion is higher, at 20%, i.e. -14dB. One result of this is that the gain varies considerably with signal amplitude, and the feedback loop gain is only 40dB at zero input signal. For the range of signal voltage at which the distortion figures are quoted the loop gain actually falls to zero over part of the sine-wave, and what we are really looking at is the effect of asymmetric clipping. Of course it is impossible to reduce the distortion of an amplifier once it has been driven beyond its output clipping level, no matter how much or how little feedback we use, and it is well known that high feedback makes the clipping more abrupt, so that more unpleasant high harmonics are produced. Any form of clipping is of course a bad thing to be avoided, and the cure is just to turn down the volume control, or if this is unacceptable use some form of 'soft clipping' ahead of the amplifier. The summing up at the end of the article made 7 points, 3 of these being about clipping. Other points were that below clipping and for more moderate open-loop non-linearity the idea that feedback reduces all distortion equally well was 'fair enough', and negative feedback 'works like a charm.'
2). I was puzzled by the distortion curves, because at a level of negative feedback where both 2nd and 3rd harmonics were falling the 5th harmonic, generated partly from intermodulation of the 2nd and 3rd, was still increasing. The explanation is that the feedback is being increased by increasing the gain of the feedback network, so that more of the output is being fed back. Although the output 2nd and 3rd harmonics are reducing, the levels fed back to the input are not. At first I thought that increasing the feedback by increasing the amplifier open-loop gain and keeping the feedback network fixed would prevent this. Further thought revealed that the effect depends on where in the amplifier we add the gain, and where the distortion is generated. (For example if we increase the driver stage gain the increased feedback needed to reduce higher harmonics applies to output stage but not input stage distortion.)HOME.