@goleksiak
I’d like to address this in a more exhaustive way.
The event was CAF22. I guess it was the first CAF after the COVID shutdown, so it was 4 years ago. It was during the presentation on Saturday afternoon at the professional panel forum organized by Gary Gill that several speakers were invited to present new technologies and products. We had a 60min slot with Mytek (Michael) and another speaker who unfortunately could not make it, the title of the paper/presentation was “New GanFET amplifier technology with triode-like sound”
It was 60 minutes in total, with roughly 45 minutes for the presentation and an open Q&A session of about 15 minutes.
I was the first to present (then it was Michael). During my presentation, someone in the audience, I guess it was @armyslowrdr, asked: “What do you think of Bruno Putzeys and his designs?”
I stopped the presentation and addressed the question politely, replying that while I never had the pleasure of meeting Bruno Putzeys in person, I was, of course, aware of his reputation in the field, and, regarding his designs, that AGD strictly adheres to a very simple rule, because of our company's ethical professional standards, not to comment on competitors’ products or designs. Then I tried to restart the presentation, and again, at the first step of my speech, I was interrupted by @armyslowrdr, who asked the same question again. I stopped and very calmly repeated the previous answer. When I tried to restart the presentation, @armyslowrdr expressed his frustration with me over my reticence to answer his question. While I tried to explain what the subject of the presentation was and that he could then ask more questions at the end during the Q&A section, Michael intervened and tried to re-explain, maybe with a more elegant British aplomb, the scope of the presentation, asking @armyslowrdr to allow me to finish it.
At the end of our presentations, we opened the floor to Q&A, we had many attendees and received many questions, except any more from @armyslowrdr.
And that is what happened. Now I find it quite peculiar to see again, after 4 years, that this event is still something so painful for @armyslowrdr that he feels compelled to bring it up in this and other forums on every occasion he has and spill droplets of his anger for what it clearly seems to be an unbearably heavy chip on his shoulder. I never had the pleasure of knowing the name of @armyslowrdr, nor of meeting him again in person and addressing his criticism of my answer (or the lack of a satisfying one from his point of view). To help him out, I have now asked Claude to do it for us (remember AGD does not comment on competitors’ products or design).
“The two companies are solving the same problem — distortion in a Class-D output stage — but from opposite ends. Purifi corrects the distortion; AGD tries to avoid generating it in the first place.
Purifi's principle: crush distortion with feedback
The conceptual breakthrough behind Eigentakt was Putzeys and Risbo identifying a distortion source most Class-D designers had underestimated: hysteresis distortion (they call it "iron distortion") in the output filter's inductor. The ferromagnetic core has a memory effect — the flux at a given point on the B-H curve depends on the path taken to reach it — and this nonlinearity is heard as a grainy texture in the sound.
Their fix wasn't an exotic, perfectly linear core (no such material really exists). Instead, the principle is what they call EigenTakt Error Correction (EEC), a "full output feedback" architecture: a correction circuit that encloses the entire power stage, including the output demodulation filter, inside the loop. The loop runs at extreme gain — more than 75 dB across the whole audio band, corresponding to roughly a 110 MHz gain-bandwidth product. So when the inductor produces a hysteretic artifact, the loop suppresses it with over 75 dB of compensation. Notably, Purifi achieves this with ordinary silicon MOSFETs at conventional switching frequencies; as one summary put it, the secret sauce is the very high loop gain and a carefully designed stable feedback network. The payoff is THD+N around 0.0002% or lower — sometimes below the test analyzer's own residual, around -140 dB.
AGD's principle: prevent distortion with switching speed (and a great supply)
Alberto Guerra — who worked on the team that developed GaN technology at International Rectifier — builds around Gallium Nitride MOSFETs in both the output stage and the power supply. GaN switches far faster than silicon, which lets AGD run an unusually high carrier frequency: where many quality amplifiers operate between 200 and 400 kHz, AGD's run at about 800 kHz or higher, making the PWM comparison much more precise. More broadly, GaN's speed reduces distortion and enables a very simple output filter. Guerra also leans heavily on supporting elements: he argues roughly 50% of any amplifier's performance comes from the power supply, and on extremely short signal paths (millimeters of connection rather than feet) and careful high-frequency PCB layout.
The essential contrast
Purifi accepts that the output stage and its inductor will distort, then wraps enormous, carefully stabilized negative feedback around the whole thing to erase that distortion — a correction philosophy, with feedback as the hero. AGD instead tries to generate a cleaner waveform at the source through very fast GaN switching and a robust supply, relying comparatively less on aggressive global feedback — a prevention philosophy that sits closer to the low-feedback tradition some audiophiles prefer sonically.
That difference is exactly why Purifi tends to post the lower raw THD+N numbers: maximizing loop gain is the most direct route to minimizing measured distortion, and Purifi pushes that lever harder than almost anyone. AGD's pitch is less about winning the measurement chart and more about the character of fast switching plus minimal feedback.”
I hope the AI’s answer will satisfy @armyslowrdr; if not, he can visit us at CAF26 this year.