 |
|
|
PAPERS
AKSA Kitset ~ Design Philosophy
"How can an amplifier produce superbly low measured treble distortion,
yet give the aural impression that there's sand in the tweeter?"
Martin Colloms, Stereophile, January 1998.
Overview
The comments expressed above by the noted audio critic, Martin Colloms,
highlight the dilemma faced by the audio amplifier designer. He vacillates
between low distortion and good sound, and is damned if he does, and damned
if he doesn't. Within these odd constraints, Aspen Amplifiers developed
the AKSA amplifier specifically for a sonic performance not usually available
in kitset amplifiers, particularly designs of this apparent simplicity.
The design is a novel approach with high quality components which seeks
to remove many of the sonic problems which plague Class AB solid state
(SS) amplifiers. In particular, the design seeks to emulate that most
wonderful of tube amplifier qualities; emotional engagement with the listener.
This paper describes the confusing subjectivity of the hifi golden fleece
and explains the beginnings of the AKSA and the ideas behind its evolution.
It is fair to say that the vast majority of audio amplifiers over the
last thirty years have been designed with engineering considerations in
mind, using the meter, cathode ray oscilloscope and distortion analyser.
While this has led to some very good amplifiers with impeccable specifications
and unmatched stability, it has not always produced the best possible
sound and although it would be foolish to insist that bad amplifiers measure
well or even that good amplifiers measure badly, it does seem that the
distortion specification is not the only criteria for good sonics. It
may also be that as the technology progresses and we become ever more
aware of the technology, we might well find we should be measuring other
parameters.
History
The history of high fidelity audio arguably began with the single ended
triode (SET) amplifiers used in the very early 'talking' picture theatres.
These were very musical, but they produced tiny output and required huge,
efficient horn-loaded loudspeakers. In the 1930's, transformer coupled,
class A triode push pull amps for domestic use appeared from the UK company
Brook. Limitations at that time were the recording process, and transformer
and capacitor quality.
World War II saw great developments in electronics, particularly in the
new, more efficient pentode valve. The Williamson amplifier in 1947 saw
the beginning of a quest for low distortion and high efficiency which
has continued to this day with a marketing emphasis on good specifications.
This amplifier employed a push-pull design with heavy global feedback,
using a good quality Partridge output transformer, a careful choice of
valves and a very carefully arranged feedback network. The seminal Williamson
amplifier set specification standards which were probably two decades
ahead of speaker technology at the time, but more importantly, it fixed
a technical standard in the mind of the modern audiophile, many of whom
are still alive. This amplifier is still revered today, despite using
more than 15 dB of 'unfashionable' global negative feedback. However,
since the writings of the talented Finnish designer Matti Otala were published
in the US IEEE journal in 1980, global negative feedback has fallen into
disfavour.
In 1948 Bardeen, Brattain and Shockley of Bell Laboratories invented
the transistor. The early devices were crude and fragile and offered very
low gain, but by the late-fifties the first germanium devices had evolved
into something quite useful. In 1956 Harry C. Lin of RCA devised a famous
amplifier topology which still bears his name and forms the basis of many
commercial realizations of the audio amplifier today, including the AKSA.
Early germanium transistors were not robust devices, exhibiting low gain
and adverse thermal and leakage characteristics. The first transistor
amplifiers drew heavily on valve design, with transformers frequently
used as interstage couplers and in the output stage. This approach was
soon abandoned for two reasons: expense, and direct coupling – an
easy task with 'P' and 'N' type transistors. Aggressive marketing brought
these 'new' amplifiers into many consumer living rooms, and while they
did not sound as good as the valve amplifiers they replaced, most listeners
failed to notice. Besides, they were much cheaper. By the seventies, a
huge variety of silicon transistors had become available, pioneered by
companies like RCA, Philips and Fairchild. High current 'P' devices also
came onto the market, and while 'P' and 'N' devices were not (and still
are not) electrically identical, this development paved the way for the
fully complementary design, which was enthusiastically embraced by engineers
in the late seventies and has special appeal for many reasons, including
its superior distortion characteristics.
The Classic Era
In 1972 Hitachi published a double differential, output transformerless
amplifier design in an application note. The new circuit was intended
for their new power mosfets which subsequently formed the basis of many
high power professional audio designs of today. The approach taken was
a differential development of the Lin topology first published in 1956.
Professional audio amplifiers, some developing several thousand watts,
are a special area of audio design, and they pose a number of important
challenges. Hifi designers frequently look to the benchmark designs of
Crown Amcron, MacIntosh, Phase Linear and SAE when they ponder reliability,
performance and cost.
Single or multiple pairs of power transistors in a push pull amplifier
sit between the low impedance power rails and a loudspeaker voice coil.
The audio signal, via the control circuitry, modulates this power supply
energy into the speaker. All that stand between the loudspeaker and a
beefy, potentially destructive power supply are several large transistors,
so the desiigns must guarantee complete reliability and iron-fisted control
to avoid incinerating the loudspeaker. There must never be any appreciable
DC voltage on the output – offset control is not a trivial matter,
and has baffled many designers over the years. This fact points up two
quite different requirements of the feedback mechanism in a direct coupled
amplifier; feedback must control both the DC offset of the amplifier AND
the AC gain of the amplifier. This is a tall order, and has repercussions
for good sonics.
During the classic seventies, there was both a high demand for the new
technology amplifiers and a genuine desire to improve the 'specification',
whatever that meant. An unease had grown among many audiophiles that this
new technology was not delivering the sound quality of their old valve
amplifiers, and there was a ready willingness to accept learned theories
explaining just why this was so. Matti Otala's important 'Slew Rate' article,
published in a prestigious US engineering journal in 1980, attempted to
explain this conundrum by postulating the existence of Transient Intermodulation
Distortion (TIM).
This type of distortion is created when the output voltage is required
to change faster than the amplifier can cope; it is also called 'slew
rate' distortion, and produces a complex, high order spray of harmonics
which the feedback mechanism is too slow to remove. The result of Otala's
seminal work has been a ready acceptance of his ideas and a marked improvement
in commercial amplifiers over the last twenty years to a slew rate parameter
of 25 volts per microsecond or better.
During the seventies and eighties, memorable designers and their products
frequently appeared in the market. Examples were Peter Walker (Quad),
Bart Locanthi (JBL), Bob Carver (Sunfire and Phase Linear), Jim Bongiorno
(Great American Sound, 'Ampzilla'), the Crown Amcron, David Hafler, Susumu
Tanaka, Erno Borbely and John Linsley-Hood. Linsley-Hood (who passed
away recently in March 2004) worked in the nuclear industry in the UK
and wrote extensively for Wireless World (now Electronics World). He is
noteworthy because he effortlessly stood astride both valve and solid
state technologies. At regular intervals these audio luminaries introduced
new concepts of note: local rather than global feedback, fully complementary
circuitry, dual differential input stages, cross coupled drivers, current
mirrors, cascoding and sliding bias regimes; and these are all topologies
which have been and continue to be used over the last 15 years.
These individuals and their companies slowly advanced modern amplifier
design, and most at one time or another produced important designs which
still sound good, even by today's standards.
The early valve amplifiers were often characterised by high harmonic distortion,
and the celebrated Williamson valve amplifier largely overcame this problem
with global negative feedback. At the time, and with distortion figures
of the order of 1%, attempts to reduce distortion seemed utterly reasonable.
The fact that the spectral distribution of this distortion might be important
passed unnoticed until the nineties, when the different orders of distortion,
global negative feedback and the sonics of the amplifier began to be investigated.
And there were other issues too, such as phase shift and slew rate, both
of which continue to attract attention.
Notwithstanding more than forty years of research into solid state amplification,
there still remain areas of circuit design which offer promise and yet
remain unexplored; error-correction feedforward, nested feedback
loops, single ended push pull, and hybrid circuits, to name four. Recently,
we have witnessed the advent of digital amplifiers, two examples being
the Bel Canto and the acclaimed Halcro. This are rapidly developing but
face a different set of psychoacoustic challenges. Furthermore, at all
but the extreme high end the developments of the last thirty years have
always been focused upon the engineering, and particularly upon achieving
ultra-low distortion specifications. Sonics have remained an unintended
but disappointing last, reflecting a strict engineering approach where
measurement dominates design. Digital technology perpetuates this
quest. However, there is now strong evidence the sonics are becoming important
to close the retail sale, and it is even possible that a new standard,
based on psychoacoustic considerations, is taking root in the minds of
the consumer.
Measurement v. Subjectivism
In 1993, a fascinating series of articles on audio amplifiers was written
by an English designer, Douglas Self, and published in the UK by the monthly
magazine Wireless World. The series ran for several months, and
identified the 'seven great distortion mechanisms' of amplifiers, analysing
each in turn with competent mathematical insight. This article was
widely read around the world, and clearly dragged the debate back into
the objectivist camp. Amplifier design had drawn increasing criticism
from the subjectivists over the previous decade, 'golden ears' who continued
to complain of sterile, mechanical sound – and Self felt that this
situation required redressing.
Based almost wholly on objectivism and rigorous engineering, Self's work
is an extremely informative engineering text on contemporary amplifier
design, and is strongly recommended for the interested reader. It
covers all the engineering and mathematical issues in detail, and even
offers a little humour; but nowhere does it mention actually listening
to the amplifier to subjectively assess the sonics. Indeed, Self
lambasts the subjectivists roundly, insisting that all the modern advances
in audio are unequivocally the preserve of measurement and mathematics.
This highlights one of the paradoxes of modern amplifier design;
the relentless dominance of measured parameters and a dearth of designers
actually prepared to tinker and listen.
Sound engineering – with sonics largely ignored – has remained
pre-eminent, in an area characterised by subjectivity. No-one has
ever suggested that a good amplifier 'makes the listener weep', and yet
the ability to convey emotion must surely rate as the definitive test
of a good hifi. On the contrary, these products have always been
described in the specifications and oscillographs, an area utterly alien
and of dubious relevance to the majority of consumers who buy a hifi to
listen to music and derive emotional as well as intellectual enjoyment.
However, a major problem for the objectivists has been the galling knowledge
that a lot of very good amplifiers measure very badly, particularly in
the single ended tube world. A reasonable defence of this argument
has always been that bad amplifiers measure badly, too – and this
apparent contradiction has been the inspiration for the development of
the AKSA kitset amplifier.
The AKSA
During the five years commencing in 1993, Printed Electronics' design
arm, Aspen Amplifiers, began a series of subjective experiments in an
attempt to uncover the design factors which make an amplifier sound 'good'.
It was always appreciated that this was a highly subjective judgment,
however, and to that end several sets of ears were asked to judge for
consensus. Since this was an empirical rather than an engineering
exercise, it led, some would say inevitably, to valve amplifiers.
Of particular interest was the Single Ended tube (SET) amplifier, which
Aspen noted sounded wonderful, particularly at low volume on vocal music.
The emotion conveyed by the SET was palpable. And yet it measured
very badly, with generous lashings of 2nd and somewhat lower (around 10-20dB)
3rd harmonic distortion. Clearly there was something beguiling about the
sound of this antiquated valve topology from the twenties, something worth
investigatingfurther.
A two stage, zero feedback hybrid amplifier of 28 watts was designed and
built over a three year period using an octal triode valve as voltage
amplifier and two pairs of single ended output power mosfets. It
sounded wonderful, and measured badly, with distortion approaching 1%
- an unthinkable level by modern standards. A second hybrid amplifier
in Class AB was developed to a completely different design philosophy
which produced four times the power. It too sounded marvellous,
and measured equally badly. Notable in both cases was an integrated
triode, which produced measurable and high levels of second and third
harmonic distortion. It seemed to add a warmth and humanity to the
sonics which was difficult to ignore.
Could it be that by concentrating on the total harmonic distortion of
an amplifier, we were in fact missing something? Might this distortion
phenomenon be investigated more carefully, and perhaps turned to advantage?
Why was it that after so many decades of development, audio amplifiers
of elegant design and sound engineering still sounded sterile, sibilant,
flat and mechanical? In truth, most gave no passion to a musical
performance, and fewer still made the listener weep.
The work of John Linsley-Hood, still designing and now in his mid-seventies,
led Aspen to a closer examination of the distortion mechanisms and the
subjective aspects of careful listening. Linsley-Hood opined that the
higher order, odd harmonics were particularly objectionable in listening
sessions, and that the fatiguing sound of many solid state amplifiers
might be related to this type of distortion, manifesting as a curious
unease during a long listen. In the late seventies, Matti Otala
had proposed that these high order distortion components were created
by interactions within the global negative feedback loop of a solid state
amplifier.
These higher order harmonics had long been identified, but the significant
issue was the very low level of this distortion. Often audiophiles
would comment on the hard, metallic sound of many solid state amplifiers,
and this description seemed somehow related to the notion of listener
fatigue. Measuring very low distortion in modern amplifiers is exceedingly
difficult, even with single tones; 0.002% is close to the limit,
and yet it was not unreasonable to suggest that the sorts of subjective
differences noted by audiophiles were showing up despite these very low
levels, and might also involve phase shifts within the musical content
which were the result of time delays dependent upon frequency, essentially
a slew phenomenon.
It was also noted that few audiophiles actually listen to single tones,
and that since this was a particularly easy signal for most amplifiers
to process, there might be something more to it where complex musical
waveforms were involved. Logically then, the measurement process
might itself be masking the problem.
Curiously the Single Ended valve amplifier often measured a total distortion
of 2% or even higher, but when examined by order, this distortion seemed
almost entirely made up of second and third harmonics, with virtually
no higher level, odd harmonics at all. This was a powerful clue.
Aspen concluded that perhaps the hifi holy grail was less about reducing
distortion to vanishingly low levels than about removing, or perhaps masking,
the adverse effects of high order distortion. The low order distortion
of the single ended amplifier might explain why the zero feedback, valve
amplifier sounded so full and rich; and perhaps also explained the
lean, almost surgical sound of the low distortion, negative feedback,
solid state amplifier. Of course, this explanation is anathema to
the objectivist, who logically strives to faithfully reproduce the signal
as it was originally laid down. But if high order distortion can
be somehow eliminated or reduced, it may be possible for a solid state
amplifier to sound benign, perhaps even something like a tube amplifier.
If we were to place our faith in the ear, then it might prove a most accurate
measuring instrument!
This later proved an acute observation, and was the primary focus of the
AKSA.
A New Approach
The thoughts expressed here lay dormant for some years, but eventually
crystallised a different approach to audio design. Aspen had noted in
1997 that a low distortion, zero feedback, solid state, single ended output
stage was close to transparent, and when preceded by a valve voltage amplifier,
sounded similar in ambience and sound stage to a top quality valve amplifier
– but with much more drive and arguably better resolution.
Was there a way of negating Otala's ruinous effects of global feedback?
Could the designer inexpensively simulate the valve using entirely solid
state devices? After all, triodes are very expensive and valve power
supplies are killers, both literally and financially!
In late 1999 ago Aspen examined an inexpensive kitset SS amplifier rated
at 40 watts. The owner asked if there might be a way to improve
the sound to the level of his $3000 15W Sugden Class A push pull amplifier.
While the kitset amplifier sounded passable, it suffered most of the usual
solid state problems; a shallow sound stage and a sharp sibilance.
Surprisingly, there was an element of warmth, a quality of tube amplifiers.
This attracted immediate attention; why? So with the special appeal
of the Class AB amplifier's high efficiency as a design target, the investigation
began.
The simplicity of the original amplifier was exceptional. It was
even simpler than the original Lin amplifier design proposed back in 1956.
One of its features was a simple resistive feed to the input differential
pair; while this seemed simple, when it was replaced with a 'proper' current
source the warmth vanished. It seemed that a resistive feed conferred
a 'warmth' to the sound owing to second and third harmonic distortion
– just like a valve! However, further development was required to
solve the other sonic shortcomings. It became apparent that there
were a number of key areas of potential improvement.
Six Design Rules
It emerged that there were essentially five critical design rules to a
greatly improved solid state, push-pull, Class AB amplifier. They were
subtle, but simple to implement. They are:
* Prevent Interstage Crosstalk – Properly decouple the supply rail
for the low current stages.
* Foster Voltage Amplifier Linearity – Operate the voltage amplifier
at constant current, but avoid the bland CCS.
* Minimise amplitude/phase intermodulations – Split DC offset and
AC feedback control.
* Eliminate Switching Transients – Implement accurate and progressive
charge suckout on the output stage drivers.
* Specify the highest quality (silver mica) and lowest value lag compensation
capacitor.
* Choose Semiconductors with care, focusing on speed and current linearity.
Subjectives
Comparisons with truly high end amplifiers revealed that the AKSA design
using carefully chosen semiconductor devices gave astonishing synergy.
The realism, dynamics, sound stage and transparency of the sound was impressive,
and several listeners assessed the sound as 'valve-like'. Some of
the clear advantages it held over a 35W valve amplifier with which it
was compared were resolution and layering; it was possible to identify
each and every sound in the performing space clearly, and no one sound
ever seemed to interfere with another, no matter how high the level.
A surprising quality was the sense of space in a room; this was definitely
a tube quality. Since initial listening tests this quality has been
confirmed, and over a period it was felt that the goal had been reached.
The amplifier was duly subjected to exhaustive trials, the production
prototypes progressively refined, and finally, after almost a year of
development, marketed across the Internet as an inexpensive kitset.
Reports from constructors are describing this amplifier as something very
special, regularly besting Class A design costing orders of magnitude
more.
Summary
Further testing in carefully structured listening environments has shown
that the AKSA is a remarkable amplifier with extraordinary, unexpected
qualities. It is unlike any solid state amplifier ever tested by
one audiophile we know, an accomplished speaker designer, who feels it
outperforms almost every tube amplifier he has ever heard. Since
those early days, the amplifier has undergone further refinement, particularly
in the area of board design, and the final product is both easy to construct
and very durable. With reasonable care the AKSA will give many years
of long, pleasurable service, and hold its head high in the company of
some of the best amplifiers available in this 21st century world.
I invite you now to review what others say who have built and auditioned
the AKSA amplifier. Their impressions may be read here.
Hugh Dean
© Copyright Hugh R. Dean 2000, 2001
All rights reserved
Last modified 2nd September 2001
|
 |
