What has made Class D amps better?

Hey Ragan, just curious as to what you work with. Totally agree btw. I have friends who’ve had monitors with class D amps running for well over a decade. They almost never turn them off, and they work day in and day out!

Not sure what you're saying here.  

Class D amps have two parts.  One is the SMPS portion, which is similar to any other power supply but the power output is scaled for the needs of the second stage, the PWM drive.  This creates a series of pulses with widths related to the amount of power in the audio signal.  The PWM frequency is much higher than audio frequency and can be analogous to the sampling frequency of a DAC.  These pulses are passed through a low-pass filter to average out their powers and smooth it into a sinusoidal wave, also very similar to how a reconstruction filter works on the output of a current-mode DAC. 

So if there is little or no audio, the pulses are very narrow which means the current consumption is very small, smaller than idle currents of an A/B amp.    MOSFETs are very efficient if you switch them quickly.  Their ON state is very low resistance so that they do not waste power in heat like BJTs do.  Their OFF state is low leakage, so they don't waste heat there either.  Only in the transition period do they become quasi-resistive and if you can switch with fast enough speed relative to the audio then you minimize their waste heat during that period as well.

Overall, that's why a Class-D with similar powers to an A/B amp can run completely fanless and heatsink-less while the A/B will need significant cooling.  It's "ON" but it's also "OFF" during that time too while an A is truly always on and an A/B is lower current for some period of the cycle but still pulling more current during it's idle times than a Class D.

One big drawback for Class-D is that as you approach their saturation limits (both the FET pulsewidth and output inductor current) you start getting drastic increases in distortion.  One should always overspec a Class-D amp's power by some amount for this reason.  

Its awesome to have an expert who has built these designs, thanks Svart! I'm learning a lot.

I think its interesting that Class D only started taking over when SMD and pick/place took over. And the same time some places started doing multi-layer PCB's where the circuit is hard to figure out. Size of layout shrank, and cost shrinks, next its the heatsink where they cut way back since Class D runs less hot. I just read that an amp for each channel in a consumer device is down to a 10cm square, so they can cram multi-channel surround easily into designs.

From a high performance perspective, why are they using off the shelf plate amps? Why not fully discrete (except for ICs) through hole repairable amps if Class D truly holds the performance edge?

(What I mean is.. shouldn’t we have discreet class D amps? Is there a plate amp PCB out there that high end manufacturers could reach to instead of the prefab stuff?)

Well, I'm no "expert" but the technology is similar to stuff I've done that I understand what it's doing in more detail than others.

SMD probably took over because of the amount of parts needed to support the design architecture and the intended physical size.  I'm sure you could do a through-hole Class D amp but it would be huge!  

Let's face it, it's a budgetary concern too.  SMD is easily stuffed by machines.  Through-hole isn't.  When you do find through-hole machines in the industry, the cost per-part is much higher to the point where it can actually be cheaper to do through-hole by hand.

On the other hand, there's also an IC vs. discrete performance disparity too.  ICs can do things like standardize timings that would see tolerance issues outside of the IC in the discrete world.  There's a situation when switching your FETs where you want one FET turning OFF at a known instance in time while the next FET is turning ON.  You need these timings to be VERY close to each other for efficiency's sake, but NOT overlapping as they would essentially "fight" each other in a condition known as "shoot-through".  While you could certainly design your discrete circuit to have a known timing, the tolerance between discrete parts could cause the timing window to be much, much wider (or even narrower) than you anticipate.  Trace lengths need to be tuned, parts need to be selected, extra testing needs to be done on each unit at production.  All things that add unnecessary costs to a design for very little payoff for the producing company.  As was brought up before, there's few folks out there equipped and willing to even try repairing something of this complexity (although it can be done!) so designing for serviceability is rarely considered.  Companies also don't want folks messing with stuff like this because a poor repair can cause more damage to the product and to reputations...

Anyway, I don't think that Class-D has a "performance" edge in terms of fidelity.  Apples and oranges in that regard.  What it does have is a performance edge in terms of heat dissipation and size for the fidelity.    As you've noted, a multi-hundred watt Class-D amp can be TINY and not need significant heat control and still have enough fidelity to compete with others.  That's where Class-D really shines.