 |
|
|
AKSA
Why is the AKSA so good?
The AKSA amplifier is so good because it sounds absolutely wonderful.
It is smooth, grainless, musical, detailed, punchy, and utterly clear.
It is designed specifically to sound like a tube amplifier – warm,
revealing, and involving. There has been an ongoing obsession with transferring
emotion in the design of this amplifier, and the pursuit has consumed
many years. With a gain of 32dB and an input impedance of 47kohms it was
set up to match the input parameters and gain of many tube designs, and
was conceived from the outset as a pseudo-tube amp; it really is a kind
of substitute, the sonics are that good.
The precise reasons the AKSA sounds so good derive from three separate
areas. They are topology, particularly in the area of the feedback network,
voltage amplifier and output stage; component choice; and pcb layout.
All three factors create a synergy to which the builder must be rigidly
adhere.
However, the magic ingredients of good sonics are only partly understood
by most designers, and in almost every case, taken on faith by the consumer.
Often the promise is marketed with impressive specifications, but this
is not a guarantee. Even if you successfully audition an amp in a hifi
salon, you never really know just how good it is until you listen to it
in your home for several days. Even then, the definitive test is an exact
A/B test, one on one, with identical sources, program material, speakers,
cables and room.
Furthermore, there seems to be little relationship between measured specifications
and perceived sonic performance; there is definitely something quite Zen
about all this. It makes the selection of hifi gear something like choosing
the decor of a room; unfortunately, you don't really know how it all fits
together until it is complete. This makes the technology a marketing dream,
since selling is merely a matter of convincing people - often on points
like appearance (for example the elegant B&O gear!) and specifications
(Doug Self's measured, objective approach). Perhaps this explains why
there are so many hifi vendors out there, all surviving despite some very
bad sounds, and why the Japanese mass-produced products, which offer a
dazzling array of features but very ordinary sound, continue to sell so
well.
In both models of the AKSA, distortion does not exceed 0.2% at any level.
This is, in our view, almost a coincidence, and we place no emphasis on
this misleading specification. There are many detailed design reasons
for the outstanding sonic performance of this amplifier, not the least
being that the usual vanishingly small but highly objectionable distortions
of a solid-state amplifier have been eliminated, but they also derive
from extraordinarily low levels of intermodulation and a harmonic spectrum
which is not dissimilar to a vacuum tube amplifier. There are palpable
advantages over a tube amplifier, too; chiefly resolution, and amazing
bass drive, a deficiency of many tube amps produced by the compromises
in transformer design.
Is the AKSA design mature?
Yes, the AKSA has undergone four discrete evolutionary steps and is now
fully mature. Future pcb designs may differ in minor ways; but no further
topology or circuit changes are planned at this point. The last evolution
was Version 2.0, essentially a cosmetic makeover, which was released in
early March 2002.
The AKSA has been exhaustively tested under adverse conditions and with
a variety of loudspeakers to ensure complete reliability and stability.
There is little to be gained from further development, since the design
is now considered optimal. Some component changes to audiophile quality
may confer subtle sonic improvement, and a premium audiophile upgrade
kit, the Nirvana, was offered to AKSA constructors in April 2002. The
changes relate primarily to signal carrying capacitors and some redimensioning
of certain resistors; the component choices have been determined empirically
over exhaustive R&D and are very critical to sonics, as one would
expect from a refined design, where literally everything, including input
and output wiring, matters.
Does the kitset include the schematic for the entire circuit frompower
supply to speaker? Is this schematic easy to read and understand?
Yes, a schematic for both the AKSA and the power supply is included with
the kitset. We believe the schematics are easy to read and understand.
There are 13 simple steps to build the AKSA, and 7 carefully explained
steps to set up each channel and test it thoroughly. The instructions
are more than comprehensive; more than twelve thousand words in fact for
both amplifiers. Limited technical support is offered to builders by email,
but this is necessarily limited; there are hundreds of AKSA builders around
the globe and support from just one person is naturally limited.
Can the power supply be built in the same box as the two amps?
I would think that an arrangement such as Heatsink / Left Amp ... Power
Supply ... Right Amp / Heatsink with the input / output connectors out
the back would make the most sense?
This arrangement works well, although the one heatsink (supplied) can
cope with the heat from both channels. However, for exceptional
cooling in very hot climates one heatsink for each channel is excellent
insurance against overheating, and this approach is recommended in tropical
and desert climates.
Input/output connectors at the rear are a good, logical way to lay out
the amplifier and this arrangement poses no problems in practice.
The only issues we have found which really affect performance are earthing;
it has to be good, with a robust star earth to which both PCB earths must
be connected. We offer an optional pcb and electrolytic capacitors
for the power supply, and use of this pcb guarantees all the right choices
concerning supply wiring and star earthing.
What heatsink is supplied?
One heatsink 300mm (12”) x 75mm (3”) is supplied and copes easily
with the heat from both 55W AKSA channels. Two of these same heatsinks
are supplied for the 100W AKSA. The heatsink is a Conrad Engineering MF30-75-1F,
of dimension 300mm long x 75mm high, with a 50mm wide x 6mm thick cast
flange for mounting the output semiconductors, and 31 fins of 40mm length
spaced at 10mm. You can find this company here: Conrad.
This heatsink is rated at 0.37 Celsius per watt in still air and is manufactured
locally by pressure casting from premium aluminium alloy in Melbourne,
Australia. This technique is so good that Conrad Engineering is able to
form flawless 1mm thick fins in routine, daily production.
In normal operation, the heatsink temperature is of the order of 10~15
Celsius above ambient; there is more than adequate cooling reserve (Australia
is a hot country; we regularly experience temperatures in Melbourne during
summer of over 40 Celsius [104F].) The AKSA has been set up to cope with
this environment, and thus will run even cooler in colder climates such
as the northern latitudes of North America and Scandinavia.
We have found that a 0.37 degrees Celsius per Watt heatsink is quite adequate
for 2 x 55W AKSAs running on ±36V rails with free air-cooling and
a 4R load. On 8R loads, the heatsink runs even cooler. In a home entertainment
environment, one such heatsink will adequately serve the needs of three
AKSA channels.
Two ledged MF20 heatsinks are also available as two 200mm (8”) x
75mm (3”) variants, which are suitable for monoblock, two channel
construction. For builders who would like to construct the AKSAs as monoblocks,
the MF20 heatsinks are very compact and offer even more cooling than stock
is available at a nominal additional cost (US$10).
However, for exceptional cooling at high power in very hot climates with
a single heatsink, an MF30-1F-100, of dimensions 300mm (12”) length
by 100mm (4”) height, is recommended. Such a heatsink is excellent
insurance against overheating, and is recommended in climates where daily
temperatures exceed 35C (98F). High humidity should not affect cooling;
only ambient temperature will determine this.
What size in inches is the heat sink? Since I intend to build these
on separate chassis, are there two smaller heatsinks I could get from
you?
The heatsink is 11.8 inches long and 2.95 inches tall. The fins
are 1.77 inches deep, and the ledge is 1.33 inches wide. We do not
supply other heatsinks as part of the kitset, or the pre-assembled modules.
A number of our constructors have conveniently cut the heatsink in two
5.9" lengths. This works fine, but you'd need to cut it yourself.
(It's quite a job with a hand hacksaw!).
Can I use six channels of the AKSA for a home theatre system?
Each module of the 55W AKSA was dimensioned to a width of 75mm so that
three such modules could be easily accommodated on a 300mm heatsink of
0.370 Celsius per Watt. This arrangement gives six individual amplifiers
along two separate, 300mm heatsinks – an ideal configuration for
a home theatre system with more than adequate cooling capacity given the
task.
The 100W AKSA module is fitted one to a heatsink; thus two heatsinks are
required for this stereo amplifier to accommodate its far greater heat
output.
Most home theatre systems use smallish speakers with efficiencies between
85dB/Watt/metre and 90dB/Watt/metre sensitivity. However, since these
speakers completely surround the viewer, the individual continuous power
requirements are not as great as a conventional stereo arrangement, particularly
as the center channel contains most of the intelligible, non-musical information.
However, home theatre dictates that the transient response must be very
good to create the slam necessary for the high impact, sonic effects so
common in the movies. To this end, the AKSA is superbly dynamic, and very,
very quiet at no signal. In fact, several constructors have remarked that
the AKSA is the quietest amplifier at no signal they have ever experienced.
However, if using 55W AKSA amplifiers throughout we recommend you select
a sub-woofer of greater than 92dB/W/m sensitivity, since a large portion
of the power spectrum of music is contained at very low frequencies and
more efficiency than 85-90dB/watt/metre may be required. The 100W AKSA
is suitable for any subwoofer, of course. These days many systems use
a self-powered sub-woofer, which can be connected straight to the ‘Sub
out’ plug on the DVD player. This is, in fact, preferable, since
it imposes fewer constraints for the high power requirements of deep bass.
A final consideration for home theatre is channel separation.
In stereo music reproduction, the rendering of image depth is often poor
and the amplifiers and speakers need all the help they can get. In a home
theatre situation, however, the spatial information is programmed in by
virtue of six channel processing. It is also true that the visual image
on the screen tends to diminish aural perception of soundstage. Therefore
it is usually a waste of resources to use separate power supplies for
all channels, as we strongly recommend for stereo. This fortunate economy
cuts the cost of the power supply considerably since only three transformers
and rectifier bridges need be utilized for each of the channels on each
side of the system and the center channel. This approach means that when
listening to stereo sound alone with no picture, the left and right channels
still enjoy the separation of two separate, independent supplies. A separate,
discrete supply is recommended for the center channel as this channel
carries most of the vocal content and intelligibility, greatly diminished
by crosstalk, is paramount.
However, for a stereo application, we recommend a minimum of 160VA rating
for each channel. The home theatre situation is not so demanding in terms
of continuous power demand, so a single toroidal transformer for two channels
would be more than adequately rated at 275-330VA. The secondary should
be rated at 25-0-25Vac to generate the recommended ±36V rails.
A single 120VA is suggested for the center channel. We recommend 50VW
or 63VW capacitors summing to 15,000_F for each ±36V rail and 4,700uF
for the center channel rails. Based on our experiments with just two channels,
six amplifiers, each of 55 Watts, would be absolutely deafening - painful
even - with lowish efficiency speakers in a home theatre situation.
A single 300mm x 75mm heatsink is perfectly adequate in the home theatre
role for three channels; two such heatsinks will cool six channels providing
air is permitted to circulate freely around the fins of each heatsink,
which should therefore be mounted external to the cabinet, one on each
side. One heatsink should carry left front and rear and right rear, while
the other heatsink should carry right front and center, and sub-woofer
amp where fitted.
I would like to use the AKSA as the bass amplifier in a tri-amped, active
crossover. What are my options?
There are a number of important points to make about active crossovers.
This is not a straightforward area, unfortunately, and you must consider
your options carefully since the extra cost and complexity is appreciable.
To create a low pass filter at line level (at the front end of an amplifier)
you need additional circuitry which cuts the input level to the amplifier
sharply as the frequencies pass the specified crossover point. Let’s
choose a corner frequency of 100Hz; a simple, 1st order crossover made
up of a single resistor and capacitor would, however, only give you flat
response at 100Hz (no attenuation), 6dB down at 200Hz (very little attenuation,
you could hear it as about three quarters of the original loudness), and
12dB down at 400Hz. It would be 18dB down at 800Hz, and 24dB down at 1600Hz
(and still just about audible). Depending on your drivers, this is generally
not adequate for a domestic hifi since higher frequencies are not reproduced
well by conventional woofers. Too many higher frequencies would come through
the bass driver, very likely far beyond its engineered range, and the
quality of this sound would ruin the wonderful contribution from your
mid-range driver. It would also limit the amount of power your woofer
would handle, since it would be operating well outside its normal range
for a high proportion of this power.
To fully exploit the advantages of an active crossover, you need more
attenuation as the frequencies rise above the crossover point. 6dB/octave
is not enough, but neither, with some exceptions, is 12dB/octave. For
example, speech can be heard very clearly through a woofer with crossover
of 100Hz and a slope of 12dB/octave. You don't want this to happen, since
it affects intelligibility and adds 'mud' to the sound. Woofers are usually
poor at reproducing speech frequencies and beyond.
Now we strike a different problem. A taller slope on the crossover curve
brings considerable attenuation of the signal at all frequencies –
even within the passband; it is lossy. Let's say we use a 3rd order crossover;
this is attenuation at the rate of 18dB/octave above the crossover or
“turnover” point. If this were made up of three capacitors and
three resistors, sitting at the very front end of the amp where the input
impedance is 47k_, the attenuation up to 100Hz would be about 10 times.
Thus you would need 10V input to get 1V output, even before the signal
enters the amplifier. This complication forces you in yet another direction.
Electronic (or 'active') crossovers are widely use in pro-audio. A stadium
rock concert might use around 200,000 Watts of amplification, with multiple
amplifiers, and usually with a four way crossover system, using balanced
lines throughout and a 24dB/octave slope. And since the drivers will be
located around the sound field, with some at the back as well as the front
of the arena, signal delays must be selectively introduced to the amplification,
usually digitally, to ensure that all members of the audience hear the
sound at the same instant in time. These systems are complex, but they
really work, very well in fact, and they ensure huge amounts of power
go only to the intended drivers over their optimum, designed frequency
range. This permits very large power concentrations in these drivers,
enabling them to deliver maximum sound pressure level without damage.
It is almost impossible to discern human speech on a sub-woofer with a
24dB/octave slope at lowish volume. A 24dB/octave slope is the best option
with a pro-audio active crossover; 18dB/octave is not favoured because
it does not permit full exploitation of the high power, narrow frequency
range of professional audio drivers. No, the 24dB/octave system is the
best option because the slope is quite abrupt and the phase output can
be arranged to be linear right across the passband.
However, the attenuation of even “wanted” signals (called “insertion
loss”) problem is a serious issue with a passive 24dB/octave system.
This means they must be built with integral voltage amplifiers to maintain
signal level; i.e. so that for frequencies within the wanted passband,
1Vrms in gives 1Vrms out. This is the “active” in the “active
crossover” terminology. The gain blocks can be quality integrated
circuits (good ones like the OPA275) but they can also be done with tubes.
The tube option is quite complex and expensive, however, and inclined
to create noise unless very low noise tubes are used; many designs we
have seen use the ECC88/6DJ8. (Although an 18dB/octave design, see this
example at noted designer Steve Bench’s website).
Surprisingly, all this low level electronics has very little effect on
the integrity of the music; the real damage to the music is usually inflicted
by the power amplifier stage.
An active crossover is essentially a highly specialized pre-amplifier,
with a related level of complexity. These beasts are widely available
as constructor circuits in popular electronics magazines; Elektor published
a phase-aligned crossover in the mid-eighties and again in more recent
times. The best of them are the Linkwitz-Riley circuits, which are phase
invariant and which deliver 24dB/octave slopes individually adjustable
to suit the gain of the amplifiers and sensitivity of the drivers. The
circuit we have in mind has three outputs; woofer, mid-range and treble,
and would be ideal for most domestic purposes. It requires a ±15V
power supply, and the PCB is about 8" square - making it markedly
more difficult to construct than the AKSA, in fact.
If you should choose to go the active crossover route, you cannot effectively
do it with simple 1st or 2nd order passive crossovers in front of the
amplifiers. Unfortunately the issue is much more complex, largely because
of the nature of drivers, human hearing, and the massive attenuation of
passive filters. You really need an active crossover, it should be 24dB/octave,
and it is a complex piece of equipment which requires careful assembly.
Almost all the active crossovers we have seen suggest connecting the driver
directly to the amplifier and running it as sole load. This should deliver
optimal performance, but it doesn’t in fact, and the reasons are
related to the nature of drivers and their voice coils, which are inductive.
Most amplifiers, although tube amps are an exception, do not enjoy driving
reactive loads. This is because voltages kicking back to the amp output
impress themselves across the output transistors, and can easily cause
instabilities, and in extreme cases, oscillation and destruction. Amps
like to see very small reactions from their drivers, and really need protection.
Well designed passive crossovers afford this protection to the output
stages of transistor amplifiers, and driving the speaker without such
protection carries sonic penalties.
Now we begin to see why passive crossovers are put into the speakers,
rather than into the front ends of amplifiers....
In Australia a retail electronics chain is now importing a sub-woofer
from Taiwan which is available with a built-in amplifier. These are often
called ‘plate amplifiers’. We understand there is a crossover
also available, which can be bought separately. These products are in
Europe and the US, we believe, so check out some of the higher end electronics
shops. It could save you a lot of hunting around for components and boards.
As an aside, and casting no aspersions on the other choices of amplifiers
constructors might make in active crossover systems, we have compared
the 55W AKSA to a refurbished Leak Stereo 30 and found the bass superior,
the mid-range a very close match, and the treble clearer and more detailed.
This comparison was made with a full range system operating into a bookshelf
speaker with 6dB/octave crossover and Morel MWP142 woofers in D'Appolito
configuration. While the active crossover is the esoteric, even purist
approach for domestic systems, very high levels of performance can be
achieved with correctly designed, conventional speaker crossovers. It
is true that the very high powers of pro- audio mandate use of electronic
crossovers, but their complexities make them a problematic and expensive
choice for home systems.
The secret is to work out where the damage is done to the music, and fix
the problem. In general, the damage is not done in the passive crossovers
used in conventional domestic speakers as long as the crossovers are 1st
or 3rd order, the design is carefully crafted, and the component quality
is high. There is probably more to be gained in the domestic setting by
correct choice of amplifier and cables than there is by use of a three
way, 24dB/octave crossover and multiamp implementations.
Of course, last but not least, the path we advocate is less complex and
a lot cheaper. But again, we would emphasize that a properly set up three
way active system can sound spectacular, delivering a quality of dynamics
and a clarity difficult to match with any system. But the best implementations
are very high power, unsuited to the domestic setting unless your listening
area has the dimensions of a concert hall...........
How does the AKSA perform in bridge mode? What output can you expect?
We have not used AKSAs in bridge mode. With few exceptions, (the Jeff
Rowland amps using LM3886 ICs and the mosfet powered Californian Muse),
we do not favor the bridge mode for high end amplification. There are
several reasons;
Ý The bridge requires additional active circuitry at the input, and a
phase inverter, which increases distortion (if balanced line inputs are
not used).
Ý The source impedance of a bridge amp is twice the individual amp output
impedances, and since each amp is generally of a smaller configuration
than a single, larger amp, the resulting source impedance can be up to
four times as high as one larger amplifier. This affects damping factor,
and may affect the drive of the amplifier, particularly into difficult
loads.
Ý Two bridged amps give double the voltage at the load, and this doubles
the current requirement through each stage, greatly increasing the cost
of the (combined) output stages. Since the output stage of each amp needs
to be as robust as a single, larger amplifier to cope with this current,
this forces the designer back to a single, larger amp with reduced complexity
but heavier output stage.
However, it's not all bad, since we have anticipated this particular requirement
and have developed a 100W AKSA with the same sonic signature as the 55W
AKSA. In time we hope to create an even more powerful amplifier, but certain
protection issues loom large once we move over 100W. Some time back we
realized that for the sophisticated US, European and Australian markets
we really needed more powerful amps in our product lineup; but you have
to walk before you can run, and achieving this has not been easy. To date
we have been successful, happily; the 100W AKSA has absolutely phenomenal
bass control, out of all proportion to its size; it sounds like a 250W
amp. This amplifier has been on sale since November 2001.
How has the gain of the AKSA been set?
The gain of the AKSAs, both models, is set by the ratio of two critically
important resistors in the feedback network. For both amplifiers, this
gain is 38.3, or 32dB. These resistors fix the gain across the audio band,
and have been selected to so that the AKSA may be driven to full power
with all known CD and DVD players in the marketplace, without the usual
addition of a preamplifier with gain.
To ensure stability with all known speaker loads, the gain outside the
audio band is restricted with three different mechanisms which jointly
conspire to make the amplifiers unconditionally stable. Gain is pulled
back below unity by 450KHz, well outside the audio band, to meet this
requirement.
Because of these delicate adjustments, it is not recommended to alter
the gain of the AKSA without first consulting the designer. It can be
done, but it must be done with considerable care.
How is output stage bias level set in the AKSA?
A small pcb-mounted, 25 turn trimpot is adjusted with a fine screwdriver.
The adjusting screw is turned fully CCW at installation to set the trimpot
to maximum resistance, the amplifier is powered up for the first time,
and the offset voltage checked at the output. If it is below 30mV, the
output stage is correctly activated, and two DMM probes are attached to
the emitters of two complementary output devices. The trimpot is then
turned CW until the voltage across two complementary emitters is 50mV,
and checked after the amplifier has warmed over a period of 15 minutes,
readjusting if necessary.
The trimpot and associated bias network is trimmed so that the minimum
idling current is around 3mA, and the maximum adjustment is around 150mA.
This ensures that under any adjustment circumstances the bias level in
the output stage can never be destructively high, allowing inexperienced
constructors to breathe easy, knowing the output stage of the AKSA cannot
self destruct during the vulnerable time when bias is set.
Can the AKSA modules work with a mix of speaker driver impedances, in
my intended active crossover driven system?
Yes, unequivocally. The AKSA is very tolerant of different speaker loads,
and ideal for driving individual drivers in active systems.
Is this amp ever in Class A, and for how many Watts?
Yes, but not many. Both 55W and 100W AKSAs are Class AB push-pull amplifiers,
with bias setting of 58mA per complementary pair. This means it is in
Class A for only the first 930mV of output voltage, peak to peak. This
corresponds to 13.5mW, which you could certainly hear, but it is hardly
deep Class AB!
Class A amplifiers are extraordinarily inefficient. Even 1A of quiescent
current on the AKSA would still only yield 16Vpp in Class A into 8R, which
corresponds to 4W. Power in excess of this figure is delivered in Class
B. The dissipation for a single output pair at this bias level would still
be 36W per device, which is right on the margin for a continuous rated
150W bipolar transistor, even with good cooling! Some years back Aspen
designed a Class A amplifier called the Glass Harmony, and each channel
dissipates 150W, spread over four MOSFETs. It is single ended, not push-pull.
It uses a tube front end, and is very expensive to build. It does sound,
however, wonderful - organic, like a musical instrument. But it is not
a practical amplifier, and very heavy, and 99% of audiophiles would be
unprepared to pay the costs of purchase and ownership.
There is also the issue of conservation; this is a profligate consumer
of energy. Nonetheless, it is close to the best amp we have ever heard,
and was used right throughout as the benchmark for the development of
the AKSA, which is very, very close and arguably more dynamic. I can tell
you that if the Glass Harmony is 10/10, then the AKSA Nirvana is at 9.5,
and ranks with some very good solid-state Class AB amp, and even a few
Class A’s, as well, notably the Sugden A21, which is 25W per channel
and a superb amplifier.
How many stages are there in the AKSA topology and are they direct coupled?
There are four stages, in a conventional, but highly refined Bailey topology.
They are:
Ý Differential input stage
Ý Voltage amplifier or "VAS"
Ý Driver stage
Ý Output stage.
All stages are direct coupled, as is conventional in PP SS amplifier design.
There is a quality 0.47_F input capacitor to block any DC at the input
signal. This is a standard precaution to ensure the amp's output offset
voltage is always zero at the speaker terminals, since in DC coupled amplifiers
any potential presented to the input will be replicated at low impedance
at the output. Failure to fit this capacitor, C1 on the schematic, almost
invariably results in destruction of the amplifier or the loudspeakers
or both at some point; it must be fitted.
Which output transistors do you use in the AKSA?
We began with the Toshiba 2SC3281 and 2SA1302. These are now obsolete,
and we now use the replacements, which are 2SC5200 and 2SA1943.
These devices, a complementary pair, are identical electrically to the
originals.
The output transistors used in the AKSA are 200V 12A 125W flat pack plastic
devices, selected because their beta is utterly constant from Ic (collector
current) of 0.1A to 7A. This is a breathtaking performance from
modern semiconductors and leaves MOSFET devices in the dust. Far,
far superior for linearity, but the price is measured in moderate lack
of robustness. Some of the less linear devices are much tougher,
but in engineering there is always a price.
What is the output impedance of the 100W AKSA, and what is the effect
of damping factor on bass?
The output impedance of the 100W AKSA is 45 milliohms at 1 KHz, approximately
half that of the 55W AKSA. The topology of the 100W version is very
close to the 55W AKSA, although there are many detail changes to component
values. While the feedback factor is very similar, the source impedance
of the larger output stage is inherently halved because of the additional
pair of output devices. Thus the 100W AKSA has a damping factor of 90,
and double that on an 8 Ohm load. Beyond a damping factor of about
20, there would appear to be other factors affecting the speed and quality
of the bass; it seems to be much more complex than originally thought.
Rest assured, however, that the 100W AKSA's control of bass is phenomenal.
Would the Onsemi MJL3281A and MJL1302A be a drop in match, to replace
a Toshiba 2SC3281/2SA1302 original pair?
Yes, these replacement semiconductors work fine, but have been reported
as being very slightly harder in sound. In practice, I rather doubt most
would notice, however.
What resistor type changes would peak the AKSA’s performance?
R1, 47 kOhm, could be replaced with a Kiwame, or Holco/Vishay. R3, R4,
R9, R15, R5, R6 are also in this category; all keep the same values.
Note that all values remain unchanged. Resistor values don't change, even
with better quality components. Emitter resistors, R16 and R18, could
be replaced with non-inductive ones; there are a variety of versions
here; take a look at the Michael Percy .pdf catalogue.
Further improvements, very subtle, can be gained by using tantalum resistors,
(“Warning, Will Robinson, Warning: extreme financial danger - these
are expensive!”), in the position of R1(47 kOhm), R3(68 kOhm), R4(3.3
kOhm), R15(180 Ohms) and R5/R6 (the bootstrap resistors). Finally,
resistors which give some benefit, very slight, are the emitter resistors.
If you use a thick film or non-inductive wirewound resistor in this role
it does sound slightly better, but it's hard to hear.
I want to try changing the feedback capacitor, C2, to a polypropylene.
Does the value need to remain the same?
This capacitor operates in shunt mode as the AC ground path on the base
of a hefty voltage divider. Its influence on the sonics is thus much less
- arguably by a factor roughly equal to the gain - than the series signal
application on the input. However, a minuscule improvement is available
here, but you need not use such a high value. I would suggest a 63V Ashcroft
or Arcotronics film of 4.7É F. It is bulky, but could be mounted
under the PCB. These are good sounding capacitors.
C1 according to your "cut down instructions" on your website
is a 10u F. The value is changed to 0.47u F?
Yes, true. The electrolytic needs a considerably larger value to ensure
it delivers the bass. But a film capacitor can be smaller.
We should also mention that the assembly instructions on the website relate
to an earlier, less refined version of the AKSA and that the design has
moved on since those days. However, to protect our intellectual property
we choose to leave this earlier version on the web to illustrate general
principles and show the assembly methodology.
Would it offer a small sonic improvement to change the Evox Rifa 470nF
MMK capacitor on the input to a metallised polypropylene capacitor, such
as the Arco or Musicap brand?
Yes, certainly, although the benefits are not quite what you might expect,
particularly when balanced against cost and size. The sound of the
AKSA is so astonishing, and so dynamic, that the effect of this capacitor
is not as great as it might be in other designs. You can replace
it with a quality film capacitor such as the Hovland, or RTX, but these
are rather large and space in this location is limited. This will
bring improvement, though the sheer bulk of these capacitors compromises
the aesthetics of the circuit board layout. The feedback capacitor
is an interesting one, too, because we discovered in trials that a quality
film or tantalum here made virtually no difference. Its influence
appears to be diminished by its shunt function; it is not a series,
signal capacitor like the input capacitor. The Evox-Rifa is inexpensive
and very good sonically. They are made in Finland, presumably for
the telecoms industry, where intelligibility is paramount and has strong
implications for component choice in the hifi industry.
You mention possible component upgrades but note that the improvements
are subtle and pricey. What are these component changes? Is there a 'premium
parts upgrade'? I’m looking for the 'edge'.
The AKSA amplifiers can be further enhanced with an exquisite upgrade
we call ‘The Nirvana’. A dimensional change in two resistors
along with replacement of some very carefully selected capacitors will
bring very worthwhile enhancement, the so-called ‘edge’, elevating
an outstanding amplifier into the realm of ultra-fi, true high end. For
those requiring the very best, a stratospheric performance which would
best most high end retail equipment, the Nirvana Upgrade is unquestionably
the appropriate option.
The Nirvana Upgrade has been available from early April 2002. A full upgrade
kit, including all components and detailed installation instructions,
is priced at $US78 for the 55W stereo AKSA and $US102 for the 100W stereo
AKSA.
At Aspen Amplifiers we feel that the Nirvana is the ultimate AKSA experience,
and if you feel the need for the ultimate, then this is for you…….
I read the Douglas Self "Distortion In Power Amplifiers".
How do you regard Doug’s analysis and approach?
I admire Doug Self's analysis; it is artful, clever, and lands the debate
fair and square back at the objectivist camp. But I am left with one question,
humbly put; “Does he critically and subjectively listen to his amplifier
designs?” This is so important…..
It has to be said equally that Doug is keen on math. However, beyond a
few simple Ohms law calculations and the odd capacitor calculation for
3dB rolloff and slew rate, there is little apparent relationship between
the high order exponential calculations and the sonics. For example, he
favours the current mirror on the differential pair because of its balancing
qualities and its doubling of the transconductance into the VAS. His calculations
ignore the fact that in careful, subjective testing there is no comparison
in clarity and timbre between the current mirror and a simple, asymmetric
resistive loading to the VAS as implemented in the AKSA. The higher transconductance
has considerable influence on harmonic distortion, frequency response
and Zout because of the beneficial effects of increased negative feedback.
But subjectively, where it counts, the resistive loading without elaborate
balancing circuitry seems to sound much better and is resoundingly more
natural to boot. As a general rule, I would say that the fewer active
devices the signal passes through, or is influenced by, the better. The
truth is that active devices are instrinsically non-linear, and as is
so often the catchcry in audio design, linearity rules.
These observations have been made before by others. A French study by
Lavardin on memory effects in semiconductors intimates that current sources,
and particularly current mirrors, may exacerbate the effects of signal
memory, whereby the signal just passed changes the DC conditions and therefore
the temperature of the die, and thus influences the present signal in
transit. This is a fruitful area of amplifier design which bears much
more scrutiny by the audiophile press. After reading Lavardin’s work
recently I was convinced that many of the AKSA design features conspire
to serendipidously eliminate many of the effects of semiconductor memory.
This memory effect is related to the charge carrier aspects of rapidly
varying temperatures, and underlies the very poor parasitic qualities
of a semiconductor chip.
|
 |
