What sort of power supply is recommended for the AKSA?
The design of power supply we recommend can be found on the web pages
and in the accompanying kitset documentation. We can supply some but not
all of the components because of cost, weight, shipping and space considerations,
particularly when in most countries it is very easy to obtain a suitable
The AKSA power supply is a full dual mono. Put another way, this means
two transformers, two rectifiers, two sets of filter capacitors (four
for the 55W and eight for the 100W AKSA), all giving two independent positive
and negative supplies for the two separate amplifier modules. This is
electrically the same as monoblock construction, and can either be constructed
in two quite separate enclosures, or the one enclosure.
Excellent dynamics and realistic, subjective 'slam' from a good solid-state
amplifier largely relate to the ability of the power supply to resist
voltage sag whilst delivering high, sustained current to a highly reactive
load. Voltage sag' is minimised by using a high current transformer
with good regulation, quiet rectification, and an adequate, low impedance
reservoir capacitor with low self-inductance and thus low ESR at high
We therefore recommend toroidal transformers with good regulation - 10%
or better. Filter capacitors can be almost any good quality Japanese,
Taiwanese, German, Swedish, US or French 50VW types. 10,000uF is the maximum
recommended for each rail of a single channel, no more; 4,700uF is adequate.
A single stereo supply may use 15,000uF per rail. We have used Elna, Cerafines,
Fischer-Tausch, Sprague, Rifa, and Nichicons. There are some sonic differences,
but most of the sonic characteristics of the supply derive from the transformer
and particularly the rectifiers, which should be General Semiconductor/Philips,
types as discussed elsewhere in this FAQ, for the best sound.
We have found that overly large toroidal transformers do not actually
improve this amplifier, rather, they appear to slow it down. We have found
that two completely independent supplies prevent one channel from interfering
with another, and this has far-reaching implications sonically, especially
for imaging. Once you have heard the difference, you'd never use a single,
larger transformer for both channels. Mounting all components in the one
box is fine, and appears to cause no problems.
Transformer regulation should not be better than about 5%; more regulation
sharply increases primary and secondary inductance, which adversely affects
the sonics of the amp. For similar reasons, the capacitance should not
exceed 10,000uF per rail per channel. We realise this is in surprising
disagreement with many commercial solid-state amps. They seem to benefit
from very large power supplies, but the reasons are related to the nature
of the feedback overload function and the operation of the voltage amplifier,
which in the AKSA amps is super-fast; we use 100MHz transistors in this
role. We emphasise that the amp was not merely developed along the usual
engineering grounds, with CROs and meters. The basic AKSA design was set
up and studied on the bench, then tweaked iteratively and carefully for
sonic effect over hundreds of hours of listening.
The engineering is mandatory, of course, but there is much, much more
to competent amplifier design; and ceasing development with good specs
alone is a false dawn. In auditioning with the AKSA we have ranked alongside
Leak tube amps, Marantz PM66 Signature Series, Sugden A21 Class
A amps, and Audiom Silver Knight SE 300B amps. It is utterly astonishing
how good its sonics really are; these are all exceptionally good amplifiers.
It is worth noting that all transformers do not sound the same, and a
smaller transformer, while conferring slightly less bass control, offers
better midrange and top end. We attribute this to the lower charge currents,
which produce charge currents in the filter capacitors, particularly with
ordinary, regular quality bridge rectifiers. It is evident from our work
in this area that switching noise, although at RF frequencies, is somehow
perceived by the ear - particularly when the associated hash is eradicated
altogether by the use of ultra-fast, soft-recovery diodes.
The physics of transformer dimensioning in audio is not particularly well
understood as it relates to sonics. Suffice to say that the transformer,
rectifiers and filter capacitors comprise a system, together with the
amplifier, whose primary function is to modulate a current through a low
impedance voice coil in exact replica of a tiny alternating voltage applied
at the input of the amplifier. We have found that a larger transformer
slows the sound, while a smaller transformer lacks bass punch and impact.
A middle course is recommended; two 160VA transformers and 4 x 4,700uF
for the 55W stereo AKSA, and two 300VA transformers and 8 x 4,700uF for
the 100W AKSA. You may use other transformer sizes and capacitor values,
but in our experience the sonics will not be optimised.
We do not recommend a shunt or series regulated supply.
The reason is that the speed of the system then depends as much upon the
active circuitry in the supply as in the amplifier, and it really is prudent
to avoid introducing more active elements into the system. A high quality,
unregulated system is therefore cheaper and sounds best. We auditioned
a good quality series regulated supply, but while it sounded different,
it did not necessarily sound superior and was considered unworthy of the
extra time and expense.
In any event, the performance of the AKSA is so good it rivals most tube
systems anyway, and one of the strong drawcards of solid-state is the
low cost, low maintenance feature, which a regulated supply partly offsets.
You can, of course, choose to build your AKSA stereo amplifier with just
one supply. But if you do this, you will destroy the marvellous imaging
and depth of the AKSA, depriving yourself of one of lifes little
You recommend using UG10DCT, BYQ28E or SBYV28-200 diodes for the power
supply. Why is this, and how is the TO-220 device connected?
These diodes are ultra fast, soft recovery diodes, with two diodes incorporated
into each TO220AB package. They are difficult and expensive to source,
but their rectification produces DC which is almost as clean as batteries.
This absence of switching hash is crucial to the perceived detail and
clarity of the AKSA amplifiers, and has only been possible at affordable
cost in the last decade. Most rectifier hash is subliminal, since it is
manifest at very high frequencies, but it finds its way through the output
devices and to the speaker, creating a low level excitation of the voice
coil which raises the subliminal noise floor and militates against high
resolution at audio frequencies.
The diodes are strongly recommended, and are sold as an option.
I will have separate rectifiers for each channel. Can I use one transformer
(Avel Lindberg D4051) for three channels, if the amp is used at moderate
You may implement the power supply this way, but the result will be a
partially collapsed sound stage.
Will my 28-0-28Vac transformer produce too high a rail voltage for the
Mains voltages vary throughout the world, and if, when used with your
local mains supply, this gives you a DC voltage at the amplifier rails
of 41V or more, it may NOT be suitable. The 55W AKSA may not be operated
over 41V rails since there is a danger, particularly with 4R loads, that
the output devices will be driven into failure.
Could you help me understand the trade-offs regarding transformers.
Is there any reason why I wouldn't just maximise the transformers and
use two 330VA Avel Lindbergs or 300VA Plitrons? Is this
overkill when running totally separate power supplies?
There seems to be an optimum size transformer, and for the 55W AKSA this
is 160VA per channel. The larger 100W amplifier is best with 225VA
to 300VA transformers. A larger transformer, and a larger filter
capacitor than 4,700µF, (2 x 4,700uF per rail on the 100W AKSA)
seems to perceptively slow the amplifier; it becomes a little slower
off the mark, and lacks the attack for which it has become renowned.
We don't recommend a single power supply for both channels; we strongly
recommend a pair of separate, 160VA transformers, and the regulation of
these transformers is not really the big issue. It seems to me that
a smaller transformer seems better able to respond quicker to transient
demands than a larger transformer; of course, the steady state situation
is certainly different and does depend on the regulation, almost by definition.
I am going to use two Plitron transformers (077316211--300VA). Are these
too large? Will my midrange suffer? If so, which Plitrons do you
recommend for the sweetest mids and highs?
We haven't ourselves trialled the Plitron; in Australia we have our own
manufacturers, and a lot of imported toroidals as well. Actually, the
larger toroids are not quite as good as the smaller, 160VA (or thereabouts).
The power supply considerations are essentially conflicting; you
want a transformer with good regulation, so the voltage sag on heavy signals
will be quickly restored. This means a larger transformer, like
a 300VA. However, many of the current demands are rather transient,
very quick; so the inductance of the primary and secondary is important,
since it prevents current flowing rapidly. A smaller transformer
might have worse regulation, but this is essentially a constant current
parameter; the transient voltage regulation is arguably something better.
We are using 160VA transformers with amorphous cores, and they perform
outrageously well. The bass and image depth is really so much more remarkable
than ONE 300VA quality toroid. I am inclined to think the best sound
doesn't really matter too much in terms of transformer; what is
important is the fact that the supplies are separate. The filter
capacitors do make a difference, but the onboard, PCB capacitors make
a difference too, and you'd probably be better playing with them; (they
are much cheaper than altering the main filter capacitors).
How would improving power supply regulation by using say 300VA toroidal
transformers and larger filtering capacitors suit the AKSA?
We have found that overly large toroidal transformers do not actually
improve this amplifier, rather, they appear to slow it down. Two
transformers give much superior stereo image, particularly depth, and
once you have heard the difference, you'd never use a single power supply
for both channels. Mounting in the one box is fine, and causes no
Transformer regulation should not be better than about 5%; more
regulation sharply increases primary and secondary inductance, which adversely
affects the sonics of the amp. For similar reasons, the capacitance
should not exceed 10,000u F per rail per channel. We realise
this is in surprising disagreement with many commercial solid-state amps.
They seem to benefit from very large power supplies, but the reasons are
related to the nature of the feedback overload function and the operation
of the voltage amplifier, which in the AKSA amps is super-fast;
we use 100MHz transistors in this role. We emphasise that the amp
was not merely developed along the usual engineering grounds, with CROs
and meters. The basic AKSA design was set up and studied on the
bench, then tweaked iteratively and very carefully for sonic effect over
hundreds of hours of listening. The engineering is mandatory, of
course, but there is much, much more; and ceasing development with
good specs alone is a false dawn. In auditioning with the AKSA we
have ranked with Leak tube amps, Marantz PM66 Signature Series,
Sugden A21 Class A amps, and Audiom Silver Knight SE 300B amps.
These are all very good amplifiers; it is utterly astonishing how good
the sonics of the AKSA really are.
What brands of power supply filter capacitors do you recommend?
Every reservoir capacitor sounds just a little different, with the chief
differences appearing in the bass/midrange and in what I term 'engagement'
(holding the listener's emotional attention). We have trialed;
* Elna Cerafines. Very, very good with ethereal, detailed
mid and top end but a little thin on bass which slowly improves over time
and gets to be very impressive eventually - the Cerafines are marvellous,
but they do take some weeks to come up to speed. At first they are
thin and lack bass, then after some weeks they are very good, with very
nice but subdued, musical bass. Very expensive, however, and not
cost-effective compared to other, less esoteric capacitors.
* Nichicon KMHs. A good all rounder with very strong bass and good
midrange and top end.
* EVOX-RIFA. Excellent all rounder with strong bass, very good mids
and high and slightly warm).
* Sprague. Fantastic bass, moderate mid and slightly indistinct
This is not too many samples, but shows clearly that there are differences.
It also demonstrates that bypassing is important, and to some extent you
can tailor the sound with different capacitors on the PCB at the supply
rails. Use 100u F Black Gate or Hatano EXR or Cerafines here
(for C10 and C12), and you should alter the sound again with just any
old capacitor in the reservoir.
For dual mono supplies, it is not advisable to go higher than 4,700u F
per channel per rail. Too much seems to 'slow' the amp, and the
same is true for very large toroidal transformers. Best size is
160VA per channel. Too little reservoir capacitance seems to speed
it up to the point where it becomes skittish and etched, but generally
hum is not an issue.
I'm trying to decide between two approaches to the power supply.
I could build the power supply into each block, keeping the power "local"
and leads fairly short. Alternatively, I could build a central power
supply, (with two toroidal transformers and rectifier bridges), and run
umbilicals (five feet long) to the amps. This approach keeps
the power supply noise and heat far away from the amps, but I wonder
whether the umbilical approach might significantly slow the AKSAs.
Or will the added inductance further smooth the rectified DC? If
I built the central power supply, I would probably put the filter capacitors
in the amp enclosures. Is this correct? How about the rectifiers?
I'd probably put those in the power supply box. What's a good structure
for the umbilical? 10 or 12 gauge stranded copper twisted together
with a good shield?
Both methods work well. I am more inclined to recommend using the
short wire approach; combining the power supply in the same box
as the single channel it supplies. Joining the two should be a 15A
auto cable of high purity, stranded copper; or your speaker cable
of choice. I have used wires between power supply and board of up
to five feet with no problem. This places very heavy demands on
the 100µF bypass capacitors on the board, but that's fine and these
capacitors are of suitably high quality. The free passage of current
between the power supply system, (transformer, rectifier bridges and capacitors),
to the amplifiers greatly affects the bass and to some extend the midrange
of the sonics, so the bypass capacitors are very important with long wires,
and the potential bass with short wires is marginally superior, but scarcely
audible. What is audible is the use of dual power supplies, without
interaction of any sort, and to this end even using separate power cords
to the primaries of the transformers is a good idea. There is an
audible difference there.
If you choose toroidal transformers there is no hum issue with the AKSA,
but this is not so with standard E-I or double-C core transformers, which
have a substantial EM field surrounding them which finds its way by induction
into the low level input circuitry of the AKSA. Nonetheless, put
a metal barrier between the toroid and the PCB, mu-metal is best, but
not necessary; two pieces of tin plated iron, about 0.5mm thick,
separated by 5mm, is recommended. This scotches any EMI field, even
when the toroid and the PCB are very close together.
In your power supply instructions, you state that C1 must be "X2".
What does "X2" mean? What brand do you recommend for that?
X2 simply means 275Vac mains rated; any brand will do
the job and has virtually no effect on the sonics.
The 100 Ohm resistor in series with C1 in the mains primary circuit
should have what Wattage rating, and what tolerance?
0.5 Watt is fine, 5%. Quality here is not specifically important.
I have some air cored chokes that could be used in a pi filter. Could
they be used with perhaps a couple of high quality 1,000u F capacitors
to any benefit?
Use of chokes here is not recommended. Due to Lenz's law, any change
in current through an inductor is resisted by back EMF, which then detracts
from the supply voltage. While an inductor works fine on a Class
A tube or transistor amp, it is not at all satisfactory on a Class AB
amp because of the widely varying current draw. In operation, the
amp cuts out on heavy passages, the music dies, then a second later, it
comes back when the amp again recovers. It sounds quite spooky,
and just plain doesn't work.
Do you have any hum issues with the AKSA?
Hum is always possible when building audio amplifiers, and can arise
from many sources. It could be an earth loop on the input, an electrostatic
field around the transformer, (particularly if it is a C-core or E-I device),
or a poor connection in the signal wiring. That said, the AKSA is very
quiet indeed provided suitable care is taken to shield the transformer
and to set up the earthing correctly. Inaudible levels of hum with even
100dB/Watt/metre loudspeakers are readily achievable.
The first issue is star earthing. The power transformer centre tap, the
chassis earth, both filter capacitor ground connections and the PCB earth
terminal should all come together at one point the star point (earth).
Dual, separate supplies also use one star point, connected both to chassis
and to mains earth. To this end, using the optional power supply pcb offered
from this website ensures you get it right, and we do recommend it strongly.
With no signal applied, and if the hum is only apparent when the transformer
is on, switch off, leaving the amp running on capacitor reserve for a
few seconds. If the hum disappears, the problem is the transformer field,
and it may be necessary to isolate the transformer electrically from the
case, separate it by some distance (say at least 6, or 15cm) and
hook it to earth via one wire to star earth. In extreme cases, many designers
choose to mount the power supply remotely from the amplifier, with connection
via an umbilical cord, although this measure is rare with the AKSA, and
in our experience never happens if toroidal transformers are used.
If the star earthing is correct, but the hum disappears when you disconnect
one of the input RCAs to a source, the problem is almost certainly
induced hum from a troublesome transformer field. Once again, the solution
is to isolate the transformer electrically from the case, separate it
by some distance, (say at least 6, or 15cm), and hook it to star
earth via one wire. These problems generally do not occur with toroidal
transformers, since they have no 'frame' connection to chassis and all
magnetic circuits are contained within the body of the transformer. Serious
hum cases may call for a metal shield around the transformer the
best material is mu-metal. However, the toroidal transformer recommended
for the AKSA power supply is intrinsically quiet at low current, where
hum problems are most apparent. In practice, AKSA constructors who use
the recommended toroidal transformers have few problems with hum.
Could I use circuit breakers instead of fuses for the power supply?
Jon Risch recommends these vs. fuses for his AC line filters. There
is a 3 Amp breaker in the Potter & Brumfield W28 series.
A Potter & Brumfield W28 Series thermal circuit breaker rather than
a fuse is perfectly acceptable. A breaker offers less total resistance
at high current draws, and the ability to be reset after a surge. The
W28 Series is compact and fits in the same hole as a panel-mount fuse
holder. They are available from Newark Electronics as follows:
5 Amps Newark part# 75F010 Type W28XQ1A-5 $1.69
10 Amps Newark part# 75F013 Type W28XQ1A-5 $1.69
15 Amps Newark part# 75F015 Type W28XQ1A-5 $1.69
Can the power supply be built in the same box as the two amps?
Yes, and this is recommended. The amp is utterly quiet built this way,
with the Star Earth configuration implemented on the custom printed circuit
board we supply as an option. The design of power supply we recommend
can be found on the web pages and in the accompanying kitset documentation.
Physical layout is uncritical. We can supply diodes, filter capacitors
and a custom printed circuit board; but not the transformers, because
of cost, weight, shipping and space considerations. In most countries
it is very easy to obtain a suitable transformer, and extensive recommendations
of suitable toroidal transformers are made in the documentation.