OK, I investigated the cause of excess distortion at higher power. I had noticed that with buffer setting distortion had gone down which should not have been the case. I noted on the graph that my dummy load connection may need revisiting. Indeed that was the case. I use screw terminal locking banana plugs. The screws sometimes come loose. I checked and they were tight. But under instrumentation I could see that wiggling the wiring caused good few dB change at higher power. So I tightened them to death and result was much improved performance:
Due to lower distortion, we could eek out some more power at 1%THD so I updated the max and peak measurements as well:
As I expected, the impact here is lower due to much less current travelling through the wiring terminal. Still, we get closer to state of the art. As an aside, I verified that the Purifi measurements were with high gain (re-ran that measurement as part of testing the wiring).
Finally, the lowered distortion exists at all frequencies:
View attachment 257942
The buffered results would likely get a bit better as well. I will see if I can re-test those.
These results show the Hypex NCx500 to be even more remarkable than my original testing showed!