Power amplifiers: exotics, circuit design, what matters for sound quality
In my younger years I designed many amplifiers, and developed my own style and design criteria. The amplifiers evolved from standard circuits copied from other or taken from application notes, into my own very custom and quite unique circuits, based on ideas around designing for a specific distortion pattern and keeping the signal path as short and simple as possible, and avoiding global feedback.
The A25 class-A amplifier is the most exotic of those. After the 25W, I developed a topology with separate V and I stages with no global feedback, intended to use standard cheap transistors, rather than the exotic FET and ring-emitter used in the A25.
I will present those designs, and look into some other designs from other manufacturers that have unusual solutions.
The A25 class-A power amplifier:
This is a unique, one-of-a-kind build of something I now in retrospect can call electronic art.
A single-gain stage, no-feedback, AND low distortion class-A power amplifier with enough power to make it usable.
The design criteria:
- All V gain in one single gain stage
- The simplest possible signal path
- Fully symmetrical (well, of course..)
- No global feedback
- Carefully tuned distortion profile
- 20W or more / 8R output power
- Ability to drive ANY load
The result was a circuit not seen before or after:
The circuit
All gain is in the T7..T10 cascode FET symmetrical gain stage, that functions as both input stage and V gain stage. This provides all the gain necessary, around 30x/28dB, controlled with local feedback. It also provides control of the distortion profile, high bandwidth, high input impedance. The selection of both BJT T7, T10 and FET T8, T9 is critical.
T15, T16 is just isolators and current amplification driver stage, the MOSFETS were later replaced by BF757/760 BJTs.
R19 and R20 is biased to provide the voltage step required to make the topology work.
T18-T23 is the current amplifier output stage, I chose a circuit with a quite high and tight local feedback loop, where the output devices drives the load as high impedance current sources and the low output impedance is kept in control by the driver BJTs T18, T19.
All the transistors you see around those gain stages are there just to create the perfect working conditions for the transistors that does the signal gain.
In addition, there is a quite simple voltage regulator, discrete of course, for the voltage gain stage, and a voltage doubler to provide the higher voltage supply for this.
DC offset is stabilized through a DC servo, also not shown in the schematic, this servo is a op-amp feeding a thermal feedback to the current mirrors T5-T6, T11-T12. So no electrical coupling from the servo back in to the amplifier.
Input FETs are Toshiba FETs, the voltage gain BJTs are ROHM devices.
Output transistors are ring-emitter, something that was new back then, they provided much better linearity and bandwidth and gain compared to what was available before. Today, this amplifier is even easier to make, as there are even better and more powerful devices readily available.
The schematic shows no zobel, this was later added, a standard inductor-resistor is serial and a RC decoupling after.
Power supply is quite standard, with generously dimensioned toroid transformers, a rectifier bridge and 2-stage capacitor battery with plenty capacity. The voltage double solution to provide the high voltage for the input stage is special, but of course makes no difference for the signal processing compared to any other solution to get the voltages needed, it is just a very practical way to achieve this.
Mechanically, it was made around a very thick and quite large solid wood frame, with cooling tunnels and a fan that runs at slow speed.
So this is an amplifier that really is special, this circuit does not look like anything else. But does it work. I will get into that..
A25: Specs and pictures
Before I get into the detailed measurements and how that relates to sound, we look at specs and some pictures.
It is a 25W amplifier. That means, output power is at least 25W into 8 ohm load, on each channel, and some to spare. Just because, that's how we specced amplifiers back then, and actually still does, for the more exotic and nice ones.
Measured output power before clipping (each channel, also note it is a completely separate dual monoblock with no shared components other than the wooden chassis frame and the ac power inlet):
30W/8R
60W/4R
120W/2R
If specced like some new class-D amps, it would be a 60-w not a 25W. That is wrong, because the 8R rating is what tells its voltage output level, and then the rest is just about current capacity.
Idle power dissipation: 150W for 2 channels.
Yes, it gets hot, this is no "green power" and "save the klimaten" nonsense, it is pure joy. BUT - we then understand why a more powerful, say 250W, simply is not sustainable, it wolud have idle consumption around 1.5Kw and that is very unfortunate in so many ways, just imagine a heater in full steam in your listening room in the summer, and it really gets very difficult to get rid of all the heat and not overheat the amplifier.
And yet the bias is set rather low, it runs pure class-A only into 8R load.
Now you get a better understanding for why it is only 25W.
Found some pictures of the chassis, not great, but you can see how it looks:
Why this matters - even if the difference in sound sometimes can be difficult, if not impossible, to hear
What most people missed about my amplifier test, was that it did show there is indeed audible differences between amplifiers, it just happened to be impossible to hear it in the music samples in that special scenario.
And a difference between two amplifiers is still there, if it can be verified by measurements, even if it is difficult to hear.
What is important here is to understand which properties of an amplifier are negative for sound quality, and which can enhance and give character to the sound.
There are 3 properties of an amplifier that can affect sound quality; frequency response, noise and distortion.
Of those, only distortion is of interest, because there should be no audible noise in a high-end audio amplifier, and flat frequency response is not difficult to achieve. Unless it is a tube amplifier, or a class-D, but even then, any deviation can be recovered with no loss to the original signal. Distortion, which is caused by nonlinearities, can not be recovered.
Distortion can be very complex. In circuits with many stages and feedback loops, the resulting signal is destroyed permanently by added distortion components, and if the input signal is complex like music, there will be not only harmonics, which are tones added with higher frequency than the original signal, but also difference components that we call intermodulation.
The characteristics of the distortion is by far more important than the distortion level in numbers. High order distortion which adds tones at much higher frequency, is much more audible than 2. or 3. order distortion. A 2. harmonic of say 2% will be inaudible, while a 8. order of 0.01% may actually be possible to hear.
This is why audio designers obsess over crossover distortion, because it causes lots of high order distortion.
Circuits with very high open-loop gain and narrow open-loop bandwidth also suffer from high order distortion, and the modulation components are very complex and difficult to get rid off.
Then there are distortions that can create a sound with a specific character, without being destructive to the sound quality, it just adds a bit of flavor.
A signal transfer function that creates very simple difference products, and very low high-order harmonic distortion, is what is desired. And that is the reason for the circuit choices made in the A25 amplifier.
I wanted to have full control of the distortion profile, and I wanted a simple profile that creates only the first order difference products, and no high order distortion. A transfer function that satisfies this criteria is a single FET stage, where the voltage across the main amplifying transistor is kept constant. And no global feedback.
When the amplifier was designed, I modeled and calculated the distortion components using FFT on the transfer function. Today, this is very simple to do in spice.
Example of measurement for the C15 amplifier, showing this desired distortion pattern:
Example lt-spice simulation of the C15 amplifier:
A25 measurements - the ones that matter:
Distortion from the A25 amplifier, 440hz sine wave input:
We see only the 2. and 3. h is visible, and even those low order components are quite low. The distortion falls off with increasing harmonics, and this pattern hold up to clipping level.
But more important, is what happens when more complex signals enter the amplifier. The 19K+20K imd measurement is a good indicator for separating the good form the lesser circuits, and here is where the A25 and similar designs are different, the high order products are attenuated or simply do not exist, and look at the low order 1KHz component showing up, and no 2K or higher can be measured:
At lower level, this imd chart now equals the distortion limits of this measurement equipment:
Exotic semiconductors:
What transistors are used in the A25, and also my preamplifier? Exotic Toshiba FET's, those that are no longer available, of course.
This article caught my attention today, as the devices pictured looked very familiar, an indeed, those are the exact devices used in the A25:
https://www.headphonesty.com/2025/08/audio-legend-reveals-black-market-obsolete-parts/
Can the circuit be redesigned to use other, new devices? Yes, well, maybe.
Clone copies - are they legit, and is the original some sort of a scam (because a clone costs a fraction of the original) ?
Copies of the Swiss DartZeel amplifiers are popular, claiming they are the same in sound as a very expensive amplifier at a fraction of the cost. The price of a clone is in fact so low, it would likely cost me more in components alone to build this amplifier.
2 questions arise; are the clones legit, in that do they provide the same sound, and, was the original a overpriced scam.
Not a scam
I claim the original DartZeel is no scam. Part from the fact that the original and the clone does not have the exact same components, the price of a product is a choice of the manufacturer, the buyer is made aware of the price before buying, the buyer can choose whether to buy or not.
Because the circuits that can be found on the net show several important design choices and some that are unique, such as the current mode input stage, no global feedback, and a special bias circuit for the output stage allowing for class AB operation with close to or equal to class A distortion characteristics.
To design and implement such an amplifier, is sort of engineering art.
Clone may also be legit
The clone circuit I found is not an exact copy, rather an implementation of a circuit similar to the original. The clone can be a very good amplifier, and this can be verified by technical measurement analysis.
If the components are different, it is not a copy, and it will not perform exactly like the original, but it can still be a good amplifier.
Places where significant savings in parts can be made, are power supply and cooling. Large power supplies cost a lot of money, large cooling structures cost and are also heavy. Save some here, and the cost goes down, but the amplifier will not be equally good for driving low impedance loads at high power.
But a clone is not a DartZeel. It is not made by, or in cooperation with the DartZeel designer, and thus it does not have the story of the DartZeel attached to it, in the same way the real DartZeel amplifer has. This means the value is not the same, regardless of pure technical quality.
Would I buy a DartZeel clone? No, I would build my own design.