Interesting article on amp distortion by Boyk and Sussman

[bcarso]

Beautiful job Larry.

Boyk and many other feedback-averse folks have been aware of the Baxandall article or at least its gist, and a modified version of the curves at the end have been reproduced by Linsley-Hood, among others. Peter's conclusion was use lots of feedback and get over it :razz: , whereas the more aurically aural tend to fuss about dogs chasing their tails etc. :roll:

I remember the prominent fun-spoiler Peter Aczel weighing in with his truncated version of the Baxandall view in his Audio Critic mag.

The Boyk-Sussman paper has extended the inquiry to IM distortion, maybe for the first time (although there is very little new under the sun, so it's quite possible there are predecessors).

 

[bcarso]

[quote author="Larrchild"]

The choice of their sim frequency ratios, although done to avoid fortuitous cancellations/augmentations, is not too far-fetched in its relevance to music, interestingly, being the just intonation interval of a minor 6th.

Yeah but the inversion of that is a Major-Third, lol. This all means something![/quote]

This is sort of a musical pessimist/optimist glass half-empty or half-full thing. Reminds me of what a music prof. once said about the melodically ascending minor sixth being the most frequent opening interval in music of the early 20th century---a time when people, and especially artists, were generally in a pretty foul mood.

 

[SonsOfThunder]

[quote author="Larrchild"]Notice how the addition of local feedback improves the near-in components while creating additional higher-order spectra.[/quote]I hope you guys will forgive me reviving this thread, but I fail to understand something here...
How can you come to the above conclusion with the included photo? What I see is that you have reduced the broadband crap by introducing FB as evidenced by the "yellow line data all below the blue line data" (oversimplification, but clear enough). Yes, the upper harmonics are now more distinct, but I fail to see how this is "additional higher order spectra". Has there been an offset of the data? Please tell me if I am off here.

Audio Precision will do FFT analysis (thru post processing) and that can be educational if a little slow. I have often watched the distortion product on a scope, but I had never thought about connecting a regular spectrum analyzer to the AP's output of the distortion product to watch things happen real time. This ought to create some fun at work next week!

Peace!
Charlie

 

[bcarso]

I don't think the upper spuriae are there in the blue trace at all to speak of---it's not just that they are more distinct in the lower trace. There's not an offset per se, but I believe Larr has adjusted the levels so that the two input tones produce a comparable output. Thus the relative noise is lower since the gain is lower, and as noted the first-order harmonic distortion and IM products are reduced, but the forest of higher-order IM has emerged.

What would be interesting would be to use a lot of signal averaging to try to bring up the possible energy already there before feedback. Or synchronous detection techniques or other narrowband filtering could be used at each frequency of interest, and adjacent to them, to estimate the noise contribution. I'm sure with the bipolars you will see something, but not as much as in the feedback case.

I did a bunch of sims for other circuits, some appreciably more complex than the Boyk/Sussman ones, and I see the same results qualitatively. The biggest likely discrepancy with real-world results is probably with the JFETs, because my simulator still assumes that even short-channel ones behave in accordance with the popular square law. And I know the SK170 departs from this conformance, based on some experiments I did a while ago. Barrie Gilbert (of the eponymous variable gain cell) has also mentioned in passing that modern FETs are often modeled inappropriately.

I've been using IM sims now to validate designs, paying special attention to the 11th and 13 harmonics of an implied fundamental at 200Hz, that is, using equal amplitude input signals at 600Hz and 1kHz. The results are quite fascinating.

 

[Larrchild]

The grounded-emitter class a stage (No Feedback) won't take much drive signal, but my noise floor on the FFT has improved since I went 24 bit. I will repost soon.

Post-processing, synchronous detection, yeah.