Could Class D really be that good?

@atmasphere Thank you for the very helpful information. I recalled someone asking / commenting about the musicality and I have examined all the ASR measurements but could not find one appropriate to be an indicator.  After a brief search I was only able to identify two gears (see below) that ASR give the THD vs frequency chart. I am not sure what the thought process was behind making this chart obsolete.

The THD vs frequency chart was soon replaced by the following load-dependent THD charts by frequencies.  As you may know, the load-dependent THD is usually tested using 1kHz tone in the industry but was later criticized because it ignores the adverse effect of higher frequencies over 1kHz on the overall THD measure.  So, including Erin, the reviewers start tests using multi-tones.  As shown in the following chart, the 15kHz frequency has big jumps in THD levels, while the better class D module like Purifi demonstrates a much more controlled increase in THD at these high audible frequencies.   

The phenomenon of even-order harmonic distortion (HD) contributing to a "pleasant" or "musical" sound in tube amps, while odd-order HD in transistor designs may sound "edgier," is a well-known concept among audiophiles. However, the idea that modern class D modules can also produce a "sweetened" or "musical" sound despite lowering total harmonic distortion (THD) might seem contradictory.  Can you provide some actual test done for those class D circuits demonstrating how they accomplish what you have described in the last paragraph?

@lanx0003 Sure! Just look at the harmonic spectra of the Purifi module. You'll see a dominant second.

Keep in mind that in a class D amp, the things that cause traditional solid state amps to sound harsh and bright might not exist. For example, in our class D the primary non-linearity is caused by the deadtime used in the output section (something that all class D amps have). In our circuit this results in lower ordered harmonics.

When feedback is applied to most tube and solid state A or AB designs, it is traditionally applied to a non-linear input to the amp (the cathode of an input tube or base/gate of a differential pair at the input of a solid state amp). The feedback signal is thus distorted and so does not do its job properly. This causes higher ordered harmonics and IMD. This has been documented by both Norman Crowhurst (technical writer for Sams publications) and 20 years later by Peter Baxandall.

In most self-oscillating class D amps this technique is not used. So they often lack the higher ordered harmonic generation common to 99% of traditional solid state designs.

Since the ear uses harmonics to tell the difference between sounds, you can see that class D amps are capable of very different distortion as opposed to traditional solid state. So I don't see it as counter-intuitive that a class D amp can have a lot more in common with the sound of a tube amp than regular solid state amps. 

Of course, class D designs vary a lot in their sound so YMMV. IME they can vary more than the sound of tube amps, which can very quite a lot! The point here is don't write off class D because you may have heard one that sucked. As with class A or AB, the sound quality is a function of design and execution rather than the class of operation. 

@atmasphere Just look at the harmonic spectra of the Purifi module. You’ll see a dominant second.

These are freq. spretrum FFT plots for THD and IMD from the published Purifi 1ET400A data sheet. For IMD, all spikes are beyond the audible freq. range so not a concern there. But, in the THD plot, where is the ’dominant’ 2nd HD? All odd- and even- order HD are almost inaudible.

@lanx0003 

I was looking at a different article. In this one we see a dominant 3rd (my memory playing up; I still seem to recall seeing a dominant 2nd) which is treated by the ear in much the same way as the second and is the only odd ordered harmonic for which this is so.

A dominant 3rd suggests cubic non-linearity; if so the circuit is inherently lower distortion than when a 2nd is dominant (which suggests a quadratic non-linearity). Our OTLs, being fully differential and balanced, have a dominant 3rd.

@atmasphere  I was there initially but just like asr said "... The third harmonic hovers around -130 dB which again, is almost at the limit of what we can measure..."   It is inaudible either at that noise floor.

I am sure that, with such a low noise floor the Purifi / Hypex class D modules provides, the sound from it is transparent bar none.  I also concur with you that the resultant sound traits hinges on how the modules is implemented.  But I don't think the sound will be sweetened approaching to tube amp unless specifically with a filter or other component.  Maybe that is why VTV adds the input tube buffer.  Just my two cents.