Legendary Hi-Fi Brand Killed Every Class D Amp It Ever Made After It Failed One Critical Test

The amps measured well and reviewers loved them, but that wasn’t enough.

When a hi-fi brand moves to Class D, the choice usually sticks, but Rotel made a rarer call.

The Japanese manufacturer shipped B&O ICEpower Class D amplifiers for roughly a dozen years starting around 2007, then discontinued every one. And when it relaunched the Michi flagship line in 2019, every amp was Class AB.

At Montreal Audiofest 2026, the company’s North American distributor finally explained why.

More Meat on the Bone

Rotel’s clearest public explanation came during Stereophile’s walkthrough of the Michi S5 at Montreal Audiofest 2026. Robert Schryer asked Sheldon Ginn, president of Kevro International and Rotel’s North American distributor, about the amplifier topology behind the S5, a 500-watt Class AB amplifier that retails for 10 grand in Canada.

Ginn’s answer made clear that Rotel’s return to Class AB was deliberate.

His phrasing still matters here, too, as he was not presenting himself as the engineer who ran Rotel’s internal listening tests. Instead, his “I think” and “they believe” framing makes clear he was relaying Rotel’s engineering position as the company’s distribution partner, not offering firsthand test data.

But even with that caveat, the reversal he described was not casual. Rotel had spent more than a decade selling serious Class D power amplifiers before returning fully to Class AB.

What “Stability” Actually Means

Ginn’s most technically loaded claim was not “muscle” or “meat on the bone.” It was stability.

In amplifier engineering, stability can mean an amplifier’s resistance to oscillation under difficult loads. But it can also describe something listeners are more likely to notice, which is whether an amplifier keeps the same tonal balance when it is connected to real loudspeakers instead of a fixed test resistor.

Ginn appeared to be reaching for this second meaning when he talked about maintaining control across the full frequency range.

Class AB amplifiers usually behave more predictably when a speaker’s impedance changes. The speaker is connected more directly to the amplifier’s output devices, without the same kind of reconstruction filter sitting between the output stage and the speaker.

In simpler terms, the amplifier is less likely to change its tonal balance just because it is driving one speaker instead of another.

Meanwhile, Class D amplifiers’ output begins as a high-frequency switching signal. A low-pass filter then reconstructs that signal into analog audio before it reaches the speaker. And in older Class D designs, this filter could become one of the places where speaker matching started to matter, because the filter was not working into a simple, unchanging load.

But real loudspeakers are more complicated than that. They are not fixed 8-ohm resistors and their impedance can drop below 4 ohms in the bass and rise above 30 ohms at resonant peaks.

When those swings meet a Class D amplifier’s output filter, the amplifier’s response curve can shift depending on the speaker attached to it.

The result can be frequency response that sags or peaks depending on the speaker. Damping factor can also degrade at higher frequencies because the filter’s inductor adds series impedance in the signal path.

This is where Rotel’s specific Class D history matters.

The B&O ICEpower modules in Rotel’s Class D amplifiers used an older feedback architecture that predated the Ncore-class approach of placing the output filter inside the amplifier’s feedback loop. That older architecture could not fully correct load-induced shifts after the filter.

So whatever Rotel’s engineers heard in their internal comparisons, the physics of those specific modules gave the company a real technical basis for preferring Class AB.

The Problem Class D Already Solved

Rotel’s ICEpower-era concerns were not baseless, as they belonged to a specific moment in Class D development when some switching amplifiers were more exposed to the speaker-load problems Ginn seemed to be describing. But modern Class D has moved past much of that weakness.

In 2011, Bruno Putzeys’ Hypex Ncore architecture moved the feedback loop beyond the output filter, allowing the amplifier to correct interactions older designs could not fully see.

Then in 2019, Purifi’s Eigentakt pushed the idea further and made load-invariant frequency response one of its explicit selling points. And NAD shipped the Masters M33 in 2020 as the first integrated amplifier built around Eigentakt.

The market followed for a reason. NAD made Purifi Eigentakt central to the Masters M33, while Marantz used Hypex Ncore in the PM-10 and Jeff Rowland built the 825 and 925 monoblocks around the same family of switching modules.

So instead of using Class D as a cost-saving shortcut these high-end manufacturers treat modern Class D as a serious reference platform.

How engineers view it

The stronger defense of modern Class D comes from engineers, especially Bruno Putzeys, whose work helped reshape the category through Hypex Ncore and later Purifi Eigentakt.

His argument has never been that switching amplifiers are automatically superior, and otherwise focuses on how Class D becomes excellent when the engineering solves the topology’s built-in problems.

“State-of-the-art Class-D amplifiers are good, not because they’re Class-D, but in spite of it.” Putzeys said.

According to him. Output filters, switching behavior, feedback topology, electromagnetic noise, and speaker-load interaction all have to be controlled.

And while older Class D designs could sound or measure differently depending on the loudspeaker attached to them because the output filter and the speaker’s impedance curve were still interacting in ways the amplifier could not fully correct, modern module designs attacked that weakness directly.

Other industry voices have reached similar conclusions:

• Steven Stone, writing for The Absolute Sound, called the belief that Class D is inherently inferior to Class AB antiquated.
• Greg Stidsen, NAD’s longtime director of technology, has argued that the best Class D designs already compete with or outperform most Class AB and Class A amplifiers on objective measures.

So the fair reading is not that Rotel was wrong to abandon its older Class D amplifiers. It’s that Rotel had a plausible technical reason to prefer Class AB during the ICEpower era, but modern Class D has weakened that technical case. Something else helps explain why Rotel stayed the course.

The Structural Moat

If modern Class D gave Rotel a credible path back into switching amplification, Class AB preserved a different kind of advantage. It kept the company competing around something it already knows how to build in-house.

Rotel has leaned on that capability since 1961, and the Michi S5 turns it into a product statement. Each channel gets its own 2,200 VA toroidal transformer, potted in an individual epoxy-filled enclosure, feeding a capacitor bank totaling 188,000 uF.

The result is 500 watts of Class AB power into 8 ohms and more than 800 watts into 4 ohms. That much power supply hardware gives physical weight to the “muscle” Ginn described.

On the other hand, modern Class D changes the competitive equation. A brand can build a high-performance amplifier around a module from Hypex, Purifi, ICEpower, or another specialist without needing the same transformer-heavy design language.

In that market, Rotel’s six-decade manufacturing advantage becomes easier for competitors to bypass.

By staying with Class AB, Rotel competes on ground it can still make meaningfully its own. The company can point to its transformers, power supply design, factory history, and overbuilt physical execution in a way that still resonates with buyers who want high-power linear amplifiers.

Ginn’s “old school” description, then, does not read like an apology. It reads like a competitive identity. Rotel may have had technical reasons to leave Class D behind, but its continued loyalty to Class AB also protects the kind of manufacturing story few rivals can copy.