Guide to Tube and Solid-State Amplifiers: Theory and Practice

An old friend of mine was in the habit of remarking that an engineer is someone who can do for $10 what any fool can do for $100. Needless to say, he was an engineer. I suppose, in a sense, that this is a fairly good description of the major aims of engineering practice; that the designer should seek elegant, efficient and cost- effective ways of achieving a clearly specified objective. Only when satisfactory ways of doing this cannot be found is there a need to seek more elaborate or costly ways to get this result.

In the field of audio amplifiers there has been a great interest in techniques for making small electrical voltages larger ever since mankind first attempted to transmit the human voice along lengthy telephone cables. This quest received an enormous boost with the introduction of radio broadcasts, and the resulting mass-production of domestic radio receivers intended to operate a loudspeaker output. However, the final result, in the ear of the listener, though continually improved over the passage of the years, is still a relatively imperfect imitation of the real-life sounds which the engineer has attempted to copy. Although most of the shortcomings in this attempt at sonic imitation are not due to the electronic circuitry and the amplifiers which have been used, there are still some differences between them, and there is still some room for improvement.

In this guide I have looked at the audio amplifier designs which have been developed over the past 70 years, in the hope that the information may be of interest to the user or would-be designer, and I have tried to explore both the residual problems in this field, and the ways by which these may be lessened.

I believe, very strongly, that the only way by which improvements in these things can be obtained is by making, and analyzing, and recording for future use, the results of instrumental tests of as many relevant aspects of the amplifier electrical performance as can be devised. Obviously, one must not forget that the final result will be judged in the ear of the listener, so that, when all the purely instrumental tests have been completed, and the results judged to be satisfactory, the equipment should also be assessed for sound quality, and the opinions in this context of as many interested parties as possible should be canvassed.

Listening trials are difficult to set up, and hard to purge of any inadvertent bias in the way equipment is chosen or the tests are carried out. Human beings are also notoriously prone to believe that their preconceived views will prove to be correct.

The tests must therefore be carried out on a double blind basis, when neither the listening panel, nor the persons selecting one or other of the items under test, know what piece of hardware is being tested.

If there is judged to be any significant difference in the perceived sound quality, as between different pieces of hardware which are apparently identical in their measured performance, the type and scope of the electrical tests which have been made must be considered carefully to see if any likely performance factor has been left unmeasured, or not given adequate weight in the balance of residual imperfections which exist in all real-life designs.

A further complicating factor arises because some people have been shown to be surprisingly sensitive to apparently insignificant differences in performance, or to the presence of apparently trifling electrical defects -- not always the same ones -- so, since there are bound to be some residual defects in the performance of any piece of hardware, each listener is likely to have his or her own opinion of which of these sounds best, or which gives the most accurate reproduction of the original sound - if this comparison is possible.

The most that the engineer can do, in this respect, is to try to discover where these performance differences arise, or to help decide the best ways of getting the most generally acceptable performance.

It’s simple to specify the electrical performance which should be sought. This is that, for a signal waveform which does not contain any frequency components which fall outside the audio frequency spectrum - which may be defined as 10Hz-20kHz -- there should be no measurable differences, except in amplitude, between the waveform present at the input to the amplifier or other circuit layout (which must be identical to the waveform from the signal source before the amplifier or other circuit is connected to it) and that present across the circuit output to the load when the load is connected to it.

In order to achieve this objective, the following requirements must be met.

-- The constant amplitude (+0.5dB) bandwidth of the circuit, under load, and at all required gain and output amplitude levels, should be at least 20Hz-20kHz.

-- The gain and signal to noise ratio of the circuit must be adequate to provide an output signal of adequate amplitude, and that the noise or other non-signal related components must be inaudible under all conditions of use.

-- Both the harmonic and intermodulation distortion components present in the output waveform, when the input signal consists of one or more pure sinusoidal waveforms within the audio frequency spectrum, should not exceed some agreed.

level. (In practice, this is very difficult to define because the tolerable magnitudes of such waveform distortion components depend on their frequency, and also, in the case of harmonic distortion, on the order (i.e. whether they are 2nd, 3rd, 4th or 5th as the case may be). Contemporary thinking is that all such distortion components should not exceed 0.02%, though, in the particular case of the 2nd harmonic, it’s probably undetectable below 0.05%.)

-- The phase linearity and electrical stability of the circuit, with any likely reactive load, should be adequate to ensure that there is no significant alteration of the form of a transient or discontinuous waveform such as a fast square or rectangular wave, provided that this would not constitute an output or input overload. There should be no tinging (superimposed spurious oscillation) and, ideally, there should also be no waveform overshoot, under square-wave testing, in which the signal should recover to the undistorted voltage level, +0.5%, within a settling time of 20us.

-- The output power delivered by the circuit into a typical load- beating in mind that this may be either higher or lower than the nominal impedance at certain parts of the audio spectrum- must be adequate for the purpose required.

-- If the circuit is driven into overload conditions, it must remain stable, the clipped waveform should be clean and free from instability, and should recover to the normal signal waveform level with the least possible delay -- certainly less than 20us.

In addition to these purely electrical specifications, which would probably be difficult to meet, even in a very high quality solid state design- and most unlikely to be satisfied in any transformer coupled system- there are a number of purely practical considerations, such as that the equipment should be efficient in its use of electrical power, that its heat dissipation should not present problems in housing the equipment, and that the design should be cost-effective, compact and reliable.

Since it’s improbable that all these performance requirements will be met, in any practical design, it’s implicit that the designer will have made certain performance compromises, in which better performance in certain respects has been traded off against a lesser degree of excellence in others. For myself, I think that the total harmonic or intermodulation distortion, which is frequently specified in the makers data sheets, is less important in determining the tonal quality of the system, provided that it’s better than 0.05%, than its behavior under transient conditions, when tested with typical (or accurately simulated) reactive loads, performance in which is seldom or never quoted. Where appropriate, in the following text, I will try to show where various benefits are obtained at the cost of some other potential drawbacks.

. ===

Prev. | Next