The origin of class D Amplifiers

This is the first part of a Tech Talk trilogy on Class D amplification with Bruno Putzeys — co-founder of Purifi Audio and the engineer behind Hypex‘s UcD and Ncore amplifier modules. Before turning to where Class D stands today, in the current landscape of Class D amplifiers, this conversation traces the origin of Class D amplifiers back to where it actually began — and the answer turns out to be nearly a century earlier than most listeners would guess.

The origin of Class D amplifiers: a 1930s patent, not a modern invention

Class D amplification is often assumed to be a relatively recent technology, but Putzeys traces the origin of Class D amplifiers to a documented example from 1932, to a patent by Burnice D. Bedford — a figure regarded as something of a hero in power electronics generally. That patent, in essence, describes pulse-width modulation: a circuit using two power triodes driven into saturation by the sum of a triangle wave and an audio signal, with the valves doubling as both comparator and power stage. Every classic building block of what Putzeys calls the “schoolbook” Class D amplifier — oscillator, comparator, power stage, low-pass filter — is already present in that single patent drawing. What convinces him Bedford had actually built and tested the circuit, rather than just sketched a theoretical idea, are small details in the accompanying oscilloscope traces: switching overshoot and residual ripple that only show up on a real, working circuit.

Born for radio, not loudspeakers

That original 1930s Class D circuit wasn’t built to drive a loudspeaker at all — its most common early application was in radio transmitters. An amplitude-modulated transmitter is essentially a large amplifier driven by a constant carrier, with the amplitude modulation applied by varying that amplifier’s power-supply voltage using a Class D stage carrying the low-frequency audio signal. The scale involved was enormous by today’s standards: currents of around 40 amps at supply voltages as high as 70,000 volts, for the biggest transmitters of the era. PWM techniques from this period quickly spread into motor control and welding applications too — just not, for another three decades, into audio.

The first transistor-based attempts for loudspeakers

The jump to transistor-based Class D amplifiers built specifically for loudspeakers happened roughly thirty years later. Putzeys points to an early self-oscillating Class D kit sold by Sinclair as one of the first widely available examples, alongside John Ulrich — founder of Infinity, who later continued building amplifiers under the Spectron brand — and Brian Atwood, best known for a 1984 paper analyzing Class D distortion mechanisms in detail and for amplifiers sold under the Peavey brand for live sound and pro-audio use.

Why closing the feedback loop was the real obstacle

The deeper problem holding Class D back for decades, according to Putzeys, wasn’t building the basic schoolbook amplifier — that part was straightforward. It was closing a feedback loop around it. The schoolbook topology, with its separate oscillator, turns out to be fundamentally difficult to stabilize with feedback: it might behave reasonably with one particular loudspeaker connected, then become unstable the moment the load changed or was disconnected entirely. Feedback itself wasn’t really the problem, even though it got blamed for one. Feedback had been used successfully in valve amplifiers throughout the 1960s without controversy; it only acquired its bad reputation in the 1970s, when early transistor amplifiers had a genuine flaw in their input stage — the part meant to subtract the feedback signal from the input signal. Matti Otala identified that input-stage flaw in his original research, but the distinction between “the input stage caused this” and “feedback caused this” never fully filtered through into popular audiophile understanding, and feedback carried the blame for another two decades.

Why “Class D” got confused with “digital”

Compounding the technical difficulty was a naming accident. Class D simply followed Class A, B, and C alphabetically, but because the output stage switches in a binary, on-off fashion, people started associating it with digital circuitry — to the point that some languages never adopted the term “Class D” at all. German technical usage calls it “digital Verstärker,” and Japanese usage similarly translates to “digital amplifier.” That association made a very specific, purely-analog design problem sound like something that could be solved with a fully digital approach: convert a digital audio signal straight into a PWM signal using only digital logic, then amplify it with a pair of MOSFETs and a filter. It looked deceptively simple — simple enough that Putzeys recalls a steady stream of companies pitching exactly this kind of open-loop digital-input design through the 1990s — but the switching transitions themselves remain stubbornly analog, with timing errors around 1% of the switching cycle even on fast modern output stages, and that reality got swept under the carpet in favor of chasing an ever-more-refined digital PWM modulator.

The accidental breakthrough, and the 1973 patent nobody noticed

The eventual way forward came from a different design philosophy entirely: self-oscillating amplifiers, which don’t need a separate oscillator at all — they oscillate by design, through a feedback loop engineered to be unstable in precisely the right way, while also processing the audio signal on average. Putzeys credits an early example to Ice Power in the 1990s, where a doctoral student working under Karsten Nielsen on a conventional schoolbook-with-feedback design accidentally left the oscillator switched off — and the resulting amplifier measured better than the intended design, a genuine eureka moment that helped establish one of the first mass-market self-oscillating Class D amplifiers. Putzeys’ own version of that breakthrough, arriving in 2000 or 2001, was realizing the output filter itself could be folded directly into the self-oscillating feedback loop — the idea that became UcD. He is often credited with inventing the global-loop self-oscillating amplifier outright, a credit he has consistently pushed back on: an engineer he declines to name later unearthed a 1973 patent by an inventor named Sturgeon describing an almost identical circuit — a comparator, output stage, low-pass filter, and feedback loop, no clock, fully load-stable regardless of what’s connected. Putzeys treats it as a case of an idea invented once by someone ahead of his time, then reinvented decades later once the surrounding technology and understanding had caught up.

What might have happened sooner

Putzeys is candid that if Sturgeon’s 1973 approach had been picked up and developed at the time, the entire detour through open-loop digital-input designs in the 1980s and 1990s might never have happened, and Class D could have reached its current, broadly accepted maturity two decades earlier than it actually did. That it didn’t is, in his telling, simply how invention often works — a good idea arriving before the surrounding industry is ready to recognize it, then having to be rediscovered once the timing is finally right.