Background of Relatedness
For the last thirty years I have been working with low feedback and zero feedback amplifiers. Initially I was convinced that audio power amplifiers had to include certain elements (such as negative feedback) to control distortion. After some exposure to Robert Fulton of FMI (Fulton Musical Industries) in the late seventies, a larger picture began to emerge. Robert Fulton was emphatic that the quality of the topology came first, then quality of components, and if everything was right, the distortion would already be low. Negative feedback could be reduced or eliminated.
Back in the days when feedback was being experimented with, it was common knowledge that this was a compromise. In the succeeding decades it became accepted as a fact of audio design. More recently the negative feedback idea has been challenged by elements of the high-end audio community.
My own explorations resulted in an amplifier that was designed to reduce distortion in every way possible without feedback. In this way it was possible to examine the effects of feedback, since the amplifier was very functional without it. During the process, I also learned about common engineering practices that tend to hold back development in fields such as audio. In their recent book "Control Design And Simulation", Jack Golten and Andy Verwer discuss this phenomena in chapter two, with regard to applying mathematical models to the real world: "...mathematical models invariably involve simplification. Assumptions concerning operation are made, small effects are neglected and idealized relationships are assumed."
It is the mark of a good engineer to know when and which things should be assumed, neglected or idealized, and we see this in audio all the time. The problem here is human nature. We tend to stay within the limits thus set by the existing paradigms and to resist changes that threaten one's viewpoint of the world. When someone else creates challenges to the paradigms, it is normal also to try to protect one's world view by preventing the new idea from gaining ground.
As alluded earlier, negative feedback has been found to be an inexact solution to amplifier distortion. This is due to propagation delays (the very small but measureable amount of time it takes for a signal to move from the input of an amplifier to the output) which are a normal phenomenon of amplifiers. In order for negative feedback to work according to the math, it must be applied to counter the input signal in real time. Propagation delay in the amplifier circuit prevents this; the negative feedback will always be lagging the original input signal. This lag results in ringing effects and enhancement of the odd-ordered harmonics that the human ear/brain system uses to measure the loudness of a sound (specifically the 5th, 7th and 9th harmonics).
In the 1960s, General Electric's conducted a variety of tests, confirming that amounts of barely hundredths of a percent distortion were not only audible but also irritating to the human ear (conversely, they also found that the ear is quite tolerant of lower ordered harmonic distortion). In other words, a disparity exists between the mathematical proof for negative feedback and its actual application, and is an example of the engineering phenomena to which Golten and Verwer refer. Despite this, negative feedback is commonly accepted in the audio world, causing the reigning design, test and measurement paradigm to have a built-in weakness.