Hopefully, we're done with this unpleasantness. I will follow up here with updates on existing business, and start a new thread if and when there is an A5 project.
I hooked up the A3 again last night and my impressions are the same as before. The A3 adds warmth and body to the sound at the expense of detail. The A4 is colorful but more revealing. I do not believe in magic. I have to assume the low feedback of the A3 gives it a higher output impedance and a thicker distortion spectrum. I have experienced in the past that reducing feedback widens the soundstage and fattens the sound. The A3 has about 18 dB of feedback, IIRC, enough to introduce some feedback-related artifacts but not enough necessarily to suppress them entirely. SPICE (and Stereophile measurements) show the HPA-1 distortion rises smoothly with output level. Some articles claim this is a psychoacoustic cue that the ear interprets as lifelike. Evidently, it's a matter of fine-tuning to get a pleasing balance.
With low feedback, the circuit is going to be more sensitive to component choices and power supply interactions. I can't characterize it with formulas, but it makes intuitive sense. The sound of the HPA-1 is said to be highly parts-dependent. The A3 uses different parts so it's no surprise if it fails to capture the HPA-1 magic. It could also be a matter of personal taste. I cannot dismiss the possibility that the reputation of the HPA-1 is partly based on hype. I don't have much faith in audio reviewers. The ear is very adaptable. I won't go so far as to impugn anyone's taste, but I will say this: Some people prefer ketchup on their hotdogs, some prefer mustard, and some like them plain.
I noted earlier that the HPA-1 has low feedback because the driver stage is heavily loaded. I confirmed in SPICE that it's easy to tweak the A3 to have the same gain-bandwidth profile as the A4. This is an A3 Bode plot with the two driver load resistors changed from 4.7K to 200K, giving 100K in parallel, like the A4. I doubled the feedback network resistance to bring the loop gain back down to 40 dB and lowered the compensation capacitor accordingly.
The theory is this may make the A3 sound more like the A4. To my ear, that would be an improvement. We will have to see. Assuming a good result, it may be that the same tuning would not sound good in a production HPA-1. Different amps, different tradeoffs.
We can learn something from these experiments even if we are not trying to replicate the HPA-1 exactly. There's no need to avoid deviating from the established path. If this were my design and other people were tweaking it, I'd say, "Good for you. You're on your own, but please let me know what you find." I am incredibly mindful of the infinite possibilities audio design presents.
These projects are my escape from the grim reality of modern life. Do I prattle too much? Maybe. No one forces you to read it.
I will order these parts today and report on the results next week.