The Blameless Amplifier?

AC Coupling, Single-Rail Supply, and Stereo Reality

Excerpts from a book that was never written, and may never be.

When I set out to design this amplifier, my goal was simple in theory but devilishly complex in practice: to create a system that worked beautifully, reliably, and safely — what Doug Self calls a “blameless amplifier.” Yet in pursuit of that goal, I had to confront ideas most audio engineers never face: asymmetry, charge flow, transformer behaviour, and the unforgiving nature of shared grounds.

The AC-Coupled Output Stage

At the heart of the amplifier was the output stage. AC coupling is often dismissed as old-fashioned, but it is the secret to safety and predictability.

The Positive Excursion — the Easy Part

Before tackling the subtleties of the negative excursion, it is worth stating plainly what happens on the positive half cycle, because this is where intuition still works.

When the upper output transistor conducts, the output node voltage rises above the quiescent mid-point of approximately ½Vcc. The output capacitor already carries a standing charge established at switch-on, and this charge ensures that as the output node moves positive, the loudspeaker terminal follows it in the same direction.

The speaker is therefore driven positive in the most straightforward way imaginable: energy is drawn directly from the DC supply rail, passed through the conducting upper transistor, through the loudspeaker, and returned via ground. Nothing mysterious happens here. This is conventional power amplification, and it behaves exactly as most engineers expect.

As the signal decays back toward zero, the output node voltage falls toward ½Vcc. At the instant it reaches that point, the upper transistor ceases to conduct. From here on, the behaviour of the circuit is no longer symmetric — and this is where intuition begins to fail.

Relic observation: On the bench, the positive excursion never raised questions. Current drawn from the rail looked exactly like current drawn from any single-supply amplifier. The anomalies — the EMI, grounding effects, fuse behaviour, and distortion mechanisms — only revealed themselves once the signal crossed the mid-point and entered the negative half cycle.

The Negative Excursion — the Hard Part

During large negative excursions the output node is driven to ground, placing the speaker-side of the output capacitor at approximately –½Vcc. Conventional current therefore flows from ground through the loudspeaker to the capacitor, in full accordance with the current-direction law. This current represents a reduction in the charge stored on the output capacitor. No energy is supplied by the power rail during this interval; the loudspeaker is driven entirely by the collapse of the capacitor’s electric field. Kirchhoff’s laws are satisfied not by a circulating current, but by charge conservation.

The discharge of the output capacitor during negative excursions can also be demonstrated directly from first principles. The current flowing through the loudspeaker is determined by the instantaneous voltage across it, which in the limiting case approaches ½Vcc divided by the speaker impedance. Since charge is related to current and time by Q = IT, the passage of this current necessarily reduces the stored charge on the output capacitor. The absolute change in charge is small compared with the capacitor’s initial switch-on charge, but it must exist. Without it, no current could flow and the amplifier would not function. In this respect, the current through the lower output transistor during negative excursions is no different in nature from the current through the upper output transistor during positive excursions — in both cases energy is delivered to the load in full accordance with the laws of physics.

Relic example: I once replaced the output capacitor with a battery at ½Vcc to model negative excursions — in simulation, of course. Watching the lower transistor conduct, I could “see” that current must flow from ground through the load to the battery, reducing its charge. The conceptual clarity of that experiment reinforced why the output capacitor must lose some charge to make the system work — physics cannot be cheated.

The Asymmetry Revealed — But What Happens Next?

The story does not end with the negative excursion.

During the negative half cycle, energy has been delivered to the loudspeaker by discharging the output capacitor. That discharge, however small relative to the initial switch-on charge, is real and unavoidable. Charge has been removed from the capacitor, and the laws of physics require that it be replenished.

On the following positive excursion, therefore, the DC rail must do more than simply supply the loudspeaker. It must also replace the charge lost from the output capacitor. This means that the current drawn from the positive supply rail during that positive half cycle is greater than would otherwise be required by the signal alone.

In other words, the positive excursion carries a double burden:

1. Supplying energy to the loudspeaker.
2. Recharging the output capacitor by the exact amount of energy previously removed.

The first makes the supply current waveform intrinsically asymmetric, even if the output voltage waveform is perfectly symmetrical. And the second adds a slope to the half-wave asymmetry. So, the resulting distortion isn't nice and predictable.

Harmonic Consequences — and Why It Gets Messy

This asymmetry in current draw does not primarily generate second harmonic distortion at the output, because the voltage waveform at the loudspeaker remains symmetric — although the second harmonic can dominate measured spectra due to supply and ground modulation rather than direct waveform distortion. Instead, the distortion manifests indirectly, through modulation of supply impedance, ground reference movement, and interaction with nonlinear elements upstream.

The dominant artefacts also tend to be odd-order — particularly third harmonic — because the correction (the recharging of the capacitor) occurs once per cycle but is phase-locked to only one half of the waveform. However, it would be misleading to suggest a single clean harmonic mechanism.

In practice, what appears is a constellation of distortion products:

• Small third-harmonic components due to asymmetric current loading.
• Higher-order artefacts caused by sharp current pulses interacting with wiring inductance and resistance.
• Intermodulation products arising from the interaction between audio-frequency currents and the rectifier–transformer charging pulses.

This is why the distortion has a characteristic texture rather than a single identifiable tone — hence the appearance of “hedgehog” pattern on an FFT display.

Why This Matters to the Power Supply and Grounding

Once this behaviour is understood, several design choices become inevitable rather than optional:

• Any impedance in the positive rail will convert these current pulses into voltage modulation.
• Any coupling between the positive rail and ground will inject that modulation into the signal reference.
• Any shared ground path (especially in stereo) will convert asymmetric current draw into audible distortion in the other channel.

This is the deeper reason why single-rail, AC-coupled amplifiers demand a fundamentally different grounding and supply philosophy from dual-rail designs. The asymmetry is not a flaw — it is the price paid for safety and simplicity — but it must be acknowledged and managed.

Relic observation: Once I stopped trying to “symmetrise” the supply and instead focused on preventing the asymmetry from contaminating the ground reference, measured distortion fell and subjective clarity improved — even though the current waveforms themselves became no prettier.

The Single-Rail Supply and Transformer Choice

Single-rail supplies are asymmetric by definition. The positive rail carries all the audio current; the negative rail — or virtual ground — carries none. This creates a waveform on the positive rail that is a series of pulses, which complicates fusing, grounding, and decoupling.

The choice of transformer becomes critical. Toroidal transformers, highly inductive and compact, are unforgiving under asymmetrical loading, prone to saturation, and reflect secondary disturbances back to the primary. Frame transformers, by contrast, offer split-bobbin isolation, wider saturation latitude, and reduced mutual inductance. Using a frame transformer allowed me to protect the amplifier with a single, quick-blow fuse, a simplicity impossible with a toroid.

Relic example: I tested a gapped toroidal transformer and saw measurable improvements in THD and noise. But once mechanical hum was accounted for, the frame transformer’s wider saturation latitude and isolation proved far more practical. It also allowed me to protect the amplifier with one fuse, rather than juggling anti-surge types (but I did add "HT fuses" just in case).

The Trap of Local Decoupling

Conventional wisdom says: decouple everywhere. In this amplifier, it would have done more harm than good. Audio-frequency decoupling from the positive rail to ground injects the signal-dependent waveform into the ground system, increasing noise. By leaving only high-frequency decoupling in place, the amplifier avoided this contamination while maintaining stability. The lesson: what works in dual-rail amplifiers does not necessarily apply to single-rail, AC-coupled designs.

Relic example: In one prototype, I observed that adding a 10 μF capacitor from positive rail to ground worsened THD measurably, even though the numbers said “better decoupling.” Removing it restored sonic clarity. This confirmed that local decoupling in a single-rail amplifier could inject the very asymmetry it was supposed to control.

Grounding and Stereo Loop Distortion

Asymmetry is further complicated in stereo. Left and right channels sharing a ground create unavoidable voltage differences across the shared return path. Even a milliohm of wire carrying brief current pulses produces voltages comparable to a quarter of a magnetic pickup’s output — enough to generate audible distortion.

My solution: separate the grounds. Input socket grounds tied to the chassis meant stereo loop distortion could not escape to "infect" upstream signal sources. Each channel’s amplifier ground ran individually to the bridge rectifier star point. The output grounds were kept separate as well, and transformer wiring was arranged to prevent induction loops.

Relic example: During early wiring, power amplifier grounds ran round the back of the centre-mounted transformer. The resulting “inverted induction loop” massively increased stereo distortion. Correcting the wiring restored clarity immediately. It was a visceral demonstration that wiring geometry, not component specs, ruled the sound.

Relic example: I also considered separating speaker grounds. In practice, the asymmetry of AC coupling made it unnecessary, but I left the option in as a safety margin — another layer of “blameless” design thinking.

The result: stereo loop distortion was minimised, but never fully eliminated. How could it ever be? Only monoblocks with independent power supplies could escape this limitation entirely.

Protection and the Philosophy of Safety

Safety is why AC coupling matters. DC-coupled outputs leave the loudspeaker directly exposed to faults; a failed transistor, offset drift, or relay misfire can destroy a £2000 speaker in seconds. AC coupling blocks DC, confines fault energy to transient events, and allows fuses to act predictably. Even the asymmetrical current waveform on the positive rail is benign: fuse heating may seem erratic, but the output capacitor ensures the speaker is protected.

Relic example: I once observed a small DC relay designed to protect my amplifier stick closed on a medium-current rail. In a DC system, this could have killed a speaker. With AC coupling, the same scenario simply caused a minor transient — nothing more. Physics, not luck, saved the day.

In short: AC coupling does not compromise performance — it constrains failure modes to the audible and harmless, letting the listener, and the engineer, sleep at night.

Lessons Learned

Despite three years of R&D, the stereo amplifier never reached the market. The ceiling imposed by stereo loop distortion, the practicalities of casework, and the dominance of monoblocks in perception meant that perfection on paper could not translate to superiority in the living room.

Yet no time was wasted. The knowledge gained — the careful analysis of asymmetry, charge flow, transformer selection, grounding strategy, and protection philosophy — forms a complete understanding of what makes an amplifier truly “blameless.” These lessons are rarer than any commercial success, and infinitely more instructive.

Closing Thoughts

Engineering is not only about building. It is about reasoning to the edge of what is possible and documenting what is discovered. The “Blameless Amplifier” lives as a relic of thought, a record of principles that survive even if the product does not. It is, in every sense, a demonstration that careful thought, patience, and respect for the laws of physics are worth far more than commercial triumphs.