Types of Amplifiers

The two extremes in analog electronic amplifier design are known as single-ended and pushpull. Single-ended amplifiers have a single output tube which amplifies the signal in one phase, while pushpull amplifiers have two output tubes, one of which amplifies the signal in positive phase, while the other amplifies the signal in negative phase. Single-ended amplifiers produce harmonic distortion of all orders (for example, second harmonic distortion is present when the amplifier adds frequencies that are double the program material frequencies), while pushpull amplifiers, at least in theory, cancel out the even order distortions present in identical output tubes working in opposite phase. Opposite phase means that as one tube draws more current, the other tube draws less current and vice versa, ideally in equal amounts as the signal varies. If the total current drawn from the power source, measured over one complete cycle of amplified signal, is constant, then the amplifier stage is said to be operating in CLASS A. If the total current drawn from the power supply, as measured over one or more complete cycles of the amplified signal VARIES, then the amplifier stage is said to be operating in CLASS B. A pushpull amplifier stage can operate drawing constant current up to a certain signal level, at which further increases in signal result in an increase of total current draw from the power source. This stage is said to be operating in CLASS AB mode. A single-ended amplifier stage can operate only in class A mode, while a pushpull amplifier stage may operate in any of class A, class AB, or class B modes.

The Single-Ended Amplifier

There is a popular myth regarding single-ended amplifiers. It is that an amplifier will sound `more musical' if it produces even-order harmonic distortion along with odd-order harmonics, as opposed to the pushpull case of reduced even-order harmonics. The theory goes that pushpull , in eliminating the even harmonics, creates a sort of harmonic imbalance that is less pleasing to the ear. It is also argued that, as octaves in music are based on a doubling of the frequency of the fundamental tone, adding even-order harmonics to the music will be just like playing the music simultaneously on other octaves, which is somehow going to improve the music over the original form, which in good music is based on hundreds of years of development of musical instruments and musicianship. I wish to contribute to this discussion the revolutionary concept that the music will sound most musical if it is faithfully reproduced without any distortions at all. According to this view, both single-ended and pushpull amplifiers can be made to sound musical if the amplifier is designed according to careful considerations, which are generally unknown.

In support of the above popular argument in favour of the supposed beneficial nature of added second-order distortion is the popular myth is that this distortion does not produce intermodulation effects, which manifests itself as a blending of the instruments and a loss of transparency and dynamic range. This myth is supported by classic texts on the subject. In the 1960's, when I was experimenting with single-ended amplifiers, I noticed that as the bass drum sounded, the cymbal sound was distorted, in accordance with the second-order harmonic distortion function. As a tube conducts more current, its transconductance (or amplification) increases, and as current decreases, the transconductance falls. The bass signal was varying the tube current over a wide range, thus subjecting the cymbal signal to varying amplification, or modulation of the gain. This is a situation where one instrument modulates, or distorts, the sound of another instrument. To my mind, this sounds a lot like `intermodulation distortion' which, I suggest is exactly what it is. Consider the possibility that the first classic text author who wrote on this subject got the math right, but didn't understand the application, and that other authors simply repeated the error, assuming the original article to be correct. I have found examples of this type of propagation of incorrect information throughout my study of audio electronics. This has taught me not to blindly accept what I read (or am told), but to conduct my own experiments, and to do original thinking to find out what seems to be the real truth. My finding here is that second-order distortion causes modulation effects, which causes loss of transparency and loss of dynamic range.

The Pushpull Amplifier

In the pushpull amplifier, the signal is simultaneously amplified by two tubes, working in opposite phase to provide a single output signal to the speaker. As the current rises by a given increment in one tube, it falls by the same increment in the other tube. Thus the total current drawn by the two tubes is constant. I recently explained this to a friend who had little understanding of electronics, and she remarked: "that is a beautiful concept" - which I agree with. I have discussed this point with high-end audio designers who pride themselves on building natural sounding amplifiers, but the sense of beauty seemed to be lost on them. Not all people have a gift for recognizing beauty. To those who do, there is the possibility of seeing truth, as truth is the ultimate beauty. Technical understanding of an innovative concept always follows in the wake of discovery, which is often fuelled by a search for beauty.

An infinite variety of pushpull circuits appear in commercial and specialist amplifiers, and these vary primarily as to which of the myriad of possible flaws have been chosen by the designers, all of whom have been completely ignorant of the ultimate truth in pushpull design.This is not surprising in light of the fact that there is no publication outlining the ideal amplifier, so designers have no idea of the correct goal of their work - and this is assuming natural sound quality is the goal. The goal is usually profit, which means the amplifier must only sound `good enough' in relation to the competition, and this is subject to the entirely arbitrary results obtained with the flawed speakers available. Most designers are happy if the amplifier is electrically stable, and produces good sine waves (a pure tone of one frequency) into a large power resistor. The problem with this type of testing is that music is vastly more complicated that a single pure tone. Music consists of many energy transients, which can be mathematically broken down into spectra of simultaneously occurring sine waves at thousands of frequencies. What happens to the musical information as it passes through the loudspeaker must be understood in order to optimize the sound quality, but this has not been done correctly.

True Audio System Performance

Traditionally, measured performance of amplifiers and speakers has not correlated well with subjective performance, as experienced by listeners. Amplifiers, which measure to be perfectly free from distortion, into resistive loads, usually do not provide a satisfying listening experience when played through speakers. Improved measurement techniques are needed. I have noticed that compression of the dynamic range is usually one of the most apparent shortfalls of reproduced music as compared with live performances of music.

Designs Falling Between Single-Ended and Pushpull

A number of designers, in a search for a better single-ended amplifier, have tried such things as shunt regulation of total output stage current, and shunt loading of the output stage, with some sort of current regulation device or impedance coupling the output stage to the power supply. What these techniques all have in common is this: there are two active devices (tubes or transistors) in the output stage (instead of one device in pure single-ended), and, as one device draws more current, the other device draws less current. These designs are effectively functioning in pushpull mode, but without the considerable design advantages that come with purely symmetrical pushpull circuits. The advantage here is that no phase inverter circuit is used.These designs represent a middle ground between single-ended and pushpull.

Single-Ended Versus Pushpull in Classic Theory

The main theoretical advantage of pushpull over single-ended is the fact that even-order harmonic distortion created by any tube non-linearity is cancelled out in a balanced circuit, whereas with single-ended, all orders of harmonic distortion are fed to the loudspeaker. This would appear to give pushpull a clear advantage, but as pushpull amplifiers have not been designed according to theoretically sound principles (these are not in any book), the opportunity has been created commercially for single-ended amplifiers to compete with pushpull amplifiers in terms of sound quality.

Single-Ended Versus Pushpull in the Real World

I feel that the real reason that single-ended amplifiers are preferred by people who care about the quality of the sound is the following: No designer, until now, has understood the ideal amplifier, and much less how to build a real-world version of it. Thus every amplifier is full of flaws, so the simpler the circuit, the fewer flaws, on average, will be made. As single-ended amplifiers can be made quite simple, they have this law of `fewer likely flaws' working in their favour. I have designed, built, and listened to many pushpull amplifiers, which has given me a healthy appreciation of the vast increase in possible ways to go wrong that come with a more complex amplifier. It was partly this appreciation of amplifier design flaws that fuelled my quest for the theoretically ideal amplifier, from which all optimized real amplifiers would be distillations in their various levels of sophistication.

There is the issue of the output transformer. The output transformer couples the output tube signal (at high voltage and low current) to the speaker (at low voltage and high current). Any uncoupled inductance, combined with distributed stray capacitance, forms a low-pass filter, which selectively reduces the higher music frequencies with respect to the lower music frequencies. The phase shift and cutoff frequency of this filter will vary with the permeability of the transformer core. Silicon-steel is commonly used in output transformers as core material. The permeability of steel is low at low applied magnetic force, and is much greater at higher magnetic force levels. The magnetic force is caused by the primary winding magnetizing current, which is determined by the signal voltage across the primary winding impeded by the primary inductance, which in turn depends upon the permeability of the core which depends upon the magnetic force. This means that, as lower music frequencies vary the magnetizing force from negative through zero to positive, the core permeability and hence the characteristics of the low-pass filter will vary, which means that higher music frequencies will be modulated by larger lower music frequencies. This is known as modulation distortion, and it gives rise to loss of dynamic range and loss of transparency.

This problem is alleviated in a single-ended amplifier by magnetically biasing the core by passing the tube quiescent (zero signal) current through the transformer primary winding, and operating at all signal levels in only one magnetic polarity, thus avoiding the region of low permeability around zero magnetic force. A higher quality core material of more consistent permeability will further reduce modulation effects.

In a pushpull output transformer, the quiescent currents of the two output tubes are carried by windings of reverse magnetic polarity, thus there is no quiescent magnetization of the core. The disadvantage appears to be that modulation distortion will occur as signal currents pass the core magnetization through zero, thus causing changes in permeability and hence changes in uncoupled inductance. There is however, another side to this argument: The balanced and magnetically cancelling quiescent currents of pushpull mean that the signal extremes are centred about zero magnetization and therefore the maximum magnetization (of either positive or negative polarity) is half what it would be with single-ended of the same power output. This allows a pushpull transformer to function linearly with a much smaller core than with single-ended. A smaller core means more compact windings which have lower losses, which is an advantage. Pushpull transformers may also be constructed with core materials having near-constant permeability even at very low levels of magnetizing force.

Advanced Design Pushpull

The theoretically perfect amplifier, in its purest form is a pushpull amplifier with a number of innovations based on ideal conceptual characteristics. This is a theoretical concept that has never been explicitly published. This ideal has been conceived through many years of dedicated research and development, which has been fuelled by passion for music, and an appreciation that truth must be respected in all aspects of this quest.

The development of a real-world amplifier whose characteristics closely match those of the ideal amplifier, including the ability to produce essentially undistorted music has required dedication. Serendipity has played an important role in the development of significantly improved amplifier designs.

To use the flying machine business analogy, the Wright brothers airplane of audio has not been developed until recently. The vast majority of existing amplifiers are akin to the flying machines which could not fly. They cannot reproduce the dynamic range which is present in the recorded music, due to a gross lack of understanding of the basic principles of physics, as applied to the art of audio reproduction.

A recent prototype amplifier is reproducing the acoustice dynamic range more accurately than has heretofore been accomplished. This means that, based on subjective observation, the amplifier is enabling there to be a much smaller loss of dynamic range between the reproduced sound and the recorded sound. This is observed as musical instruments sounding more like the real instruments would sound. For example, one often will blink one's eyes upon hearing a snare drum rim-shot at close range, whereas this does not happen when listening to audio systems. This recent prototype does cause one to blink, even when the intervening music is playing at a relatively low level, as is the case when listening to live instruments in close proximity.

Copyright 2011 Duncan Scobie. All rights reserved.


The Spiritual Path
& Reproduction of
Beautiful Music
History of
Duncan Scobie
Technical Discussion
of Amplifier Design
Other Audio
Design Projects

Truth in Audio
Duncan Scobie
1156 Union Road, Victoria, BC, V8P 2J3, Canada
E-Mail: lankford@shaw.ca
URL: www.crunchers.bc.ca/scobie

([an error occurred while processing this directive])