Expanding the Class D Conversation: How Would You Characterize Their Differences?

Go to youtube and type in iiwi reviews.  Should be the most recent review.

Class D is merely a circuit topology for power delivery.  Wouldn't all the rest of the circuits complications be filters and such the manufacturer used to tackle various issues.  I'd imagined the character of the sound has more to do with the component choices, and additional circuit subsections than the fact that one is class a or d, and the circuit choices that tend to follow those designs.

I have a simple, Schiit Magni 3, which is a class D product, which i used as a headphone amplifier for years.  Then I moved to their Asgard 3, which is A/AB, and the soundstage was clearer and wider on the Asgard, but the low end was less prominent, which didn't make me happy, so it was a trade-off.  But I dont know that it would be fair to attribute this to class topology.  I've seen more difference in a sound by changing decoupling capacitors in a pre-amplifier. 

@logistics 

In my experience, limited, I hear discussions about the sonic character of Class A and Class D. I don’t hear discussions about the character of Class A/B. You’re right that they’re topologies and there’s a lot of other factors involved, so it’s an interesting question why people talk this way.

Let me propose some ideas and I welcome your pushback. If AB’s whole design tradition is built around correcting for a topology-native problem — crossover notch distortion — then AB implementations across manufacturers are all solving roughly the same problem the same way, which would explain convergence toward a "neutral" or generic sound.

Class D and Class A each have their own different topology-native constraint, distinct from AB’s and from each other’s. Class A runs output devices at constant high bias across the full signal cycle, which means most of the power dissipates as heat rather than reaching the speaker — that’s the shared problem A implementations are built around, and it tends to push designs toward low feedback, since the devices are already fairly linear and there’s less nonlinearity to correct for. That’s also usually the explanation given for A’s reputation of low-order (2nd harmonic) distortion dominating, versus the higher-order products associated with AB’s crossover region.

Class D’s shared constraint is different: every implementation needs an output filter to reconstruct the analog signal from the switching waveform, and that filter interacts with the speaker’s impedance curve in ways that are somewhat inherent to the approach.

So the claim isn’t that D and A lack a corrective project the way AB has one — it’s that each has its own corrective project, different from AB’s, which could be what pulls D implementations toward one kind of family resemblance and A implementations toward another, while AB implementations converge toward a third, more "neutral" cluster.

If that’s right, the interesting question becomes: what’s the actual mechanism in each case — the filter/impedance interaction for D, the low-feedback/harmonic profile for A — and how much does it survive across specific component and circuit choices, versus getting swamped by them the way you’re describing with your decoupling cap example?

@hilde45 You know much more about this than me, but you touched on another area that i think creates a lot of the confusion.  There is also, the very different sound of Class A tube amplification vs transistor-based Class A, including OpAmp based solid state Class A.  I have a feeling many folks have mistakenly walked away from discussions of vacuum tube Class A designs and their sonic characteristics, thinking this applies to solid state as well, when it's quite possible that much of the nuances they heard were a tube amp trait rather than Class A.

This is basic observation but Class D Amplifiers are in a Constant on state. I take this to be constantly oscillating. I have a JBL with Class D amps and listening to it over time, it seems to be unable to resolve small mechanical variations, such as being unplugged and replugged without change. This would mean an employment of Square Wave, so my take on it is that Class D amps vary because the rate of Oscillation in comparison to the basic design parameter creates a Square Wave function .. similar to a very weak black hole. Because the two items are not correlated,  this results in potentially large variation in outcome or if you like 'Cohesion'. I have not studied the B&O variation of Class D. But I am sure that you will find that the designer compensated for this eventuality, well at least I hope he did. I wrote a work on the effect of Economic Design with the Human condition. It makes for a nasty read. These sorts of errors happen all the time as most people are not sufficiently aware and because of development software are incapable of changing it.