[Part 1 begins with an overview of amplifiers, and discusses basic requirements for current and voltage output, and transient response.]
6.4 Non-linear distortions
Until now we have been speaking about the linear distortions of the amplifiers. A linear distortion involves a change in the waveshape, or the frequency response, but where no new frequencies are introduced that are not present in the input signal.
If either the amplitude or phase of the frequency response are distorted, they are linear distortions, and will affect the waveform. Academically speaking, the full frequency response includes both the amplitude and the phase responses, although in more general conversation the term 'frequency response' usually simply relates to the amplitude portion, variously known as the magnitude or modulus of the frequency response.
Unlike the linear distortions, a non-linear distortion is one where new frequencies are introduced into the output of a system. In the case of harmonic distortion, these extra frequencies are multiples of the input frequencies. Intermodulation distortion occurs when the various frequencies in the input signal interact to produce sum and difference frequencies which may have no harmonic relationship whatsoever to the input frequencies. Noises and rattles also constitute forms of non-linear distortion in mechanical systems. In general, the non-linear distortion in well-designed electronic systems is way below the levels produced in loudspeakers.
This situation has given rise to suggestions that the non-linear distortion in amplifiers will be insignificant when heard below the much higher levels of loudspeaker non-linearities. However, electronic and electro-mechanical distortions are produced by very different mechanisms. Loudspeaker distortion tends to be benign compared to electronically-produced non-linear distortion, which tends to be much more subjectively disagreeable.
Intermodulation distortion has always been an elusive property to measure. It depends on:
1) the signal level
2) the bandwidth of the signal
3) the complexity of the signal
4) the peak to mean ratio of the signal
5) the signal waveform
6) the interaction between any of the above, and a number of other factors
Historically, and currently, harmonic distortion is still the measured quantity, but harmonic distortion, alone, is not necessarily either unmusical or unpleasant. Referred to the above list, harmonic distortion for any given frequency is only dependent upon level, because a sine wave:
1) has no bandwidth
2) has no complexity
3) has a fixed peak to mean ratio
4) has a defined waveform
5) cannot interact with itself
Chapter 9 will deal with the deeper aspects of the relevance of distortion more closely, but these points need to be made now because the ways in which amplifiers produce non-linear distortions is much more varied than would be expected simply from reading the brochures, where all good quality transistor amplifiers tend to produce approximately similar levels of harmonic distortion in the range of 0.01%. As such low levels of purely harmonic distortion are almost certainly not detectable by the human ear, it implies that other forms of non-linear distortion are the culprits where there is harshness, lack of clarity, lack of transparency, lack of 'air', or where other similar descriptions are used to qualify the less than desirable sound of any amplifier, or the difference between amplifiers.
The situation is that from published distortion figures alone, there is little that can be implied about the musical accuracy of an amplifier. Conclusions can only be drawn about the sound of an amplifier by listening to it in the specific system with which it is intended to be used. However, the 'biasing class' can give some guide to possible performance under specific circumstances.
6.5 Amplifier classes and modes of operation
There are many different designs of amplifier output stages, but they are all usually grouped into a system of classification relating to their output biasing or switching. Classes A, B and C are unswitched stages, whilst Classes D, E, G and H are switched stages.
As with so many things, there are pros and cons to each design, and amplifier designers must use their experience to decide which class seems to be most appropriate for the intended use of the amplifier. Although this is an enormous subject in its own right, we need to at least outline the concepts here, because the choice of the type of amplifier can significantly affect the characteristics of a loudspeaker system in ways which may not be apparent from simple, traditional measurement techniques.
As discussed in the last sections, things such as intermodulation distortion and instantaneous current capacity into reactive loads can give rise to distinct sonic differences between amplifiers of different design implementations. As the above two problems are unlikely to be shown up by any normal static tests of amplifiers performance, they may well
not appear on any specification sheets.
Until now, there has not arisen any standardised test for intermodulation distortion with any close correlation to subjective listening assessments. Although proposals have been discussed11,2, the number of variables listed in Section 6.4 may still mean that, in many cases, only certain types of music at certain levels and with certain combinations of instruments will lead to problems. The question then arises as to how to cover all eventualities.
Clearly, amplifiers should be free of problems to the greatest degree possible, but at what cost? One approach may be to build an amplifier with big reserves of performance, but if its construction is too big and heavy to fit in a small, self-powered monitor system, it is no option at all. However, in this discussion we will try to stick to performance, and only mention practical considerations when they can be seen to affect the overall thinking.