Thanks, Don, for a response that is evenhanded and civil - but more than that, for a response that is useful. I think your post is useful for two reasons:
- You identify and start to discuss a clear, specific mechanism by which distortion could be more or less easily heard (because of differential masking of various orders of harmonics). This is of course central to what Ralph discusses in the linked comment, and it's quite reasonable and interesting as a topic.
- Your comment also helps us focus on the problem with the OP's apparent belief, and with the "opinion" that Ralph seems to be promoting here: even if certain harmonics are more easily detected than others, that doesn't matter if all orders of harmonic distortion products produced by an amp are well below the threshold of human audibility. In @amirm 's amp tests, we routinely see that 2nd and 3rd order harmonics are at the highest levels (usually by far), and that higher-order harmonics - the "dangerous" ones - are almost always at much lower levels, even in amps deemed "competent" or mediocre.
It's the super-common problem in audiophile culture: someone - in this case Ralph and our OP - focuses on a mechanism or a phenomenon and acts like if the mechanism or phenomenon is
real, then it must automatically be
audible. The part where the magnitude or significance of the mechanism/phenomenon is evaluated compared to human hearing ability gets overlooked. This has been the formula for the sales pitch for almost every piece of audio snake oil, although folks can of course also employ this formula in good faith or at least without any specific agenda of selling something.
When discussing the audibility of distortion, we have to keep in mind that the distortion products are not a separate signal. If they were, and as low as many amps are capable of, they would be easily masked. Distortion is different in that its a
modifier. IOW it
changes the sound of musical instruments by adding harmonics. Harmonics define that sound of a musical instrument, in case their role isn't clear... ask any luthier about that.
I like and respect Ralph as a designer. But I think he's wrong here. Not so much that higher order distortion is more audible, but that human ears are super-sensitive.
For the record, I
did not say the ear is super-sensitive! To many things it is not. What I
did say is that its keenly sensitive to the higher orders.
One could use Pkane's Distort software to see how much distortion becomes audible. Or to see if lots of 2nd and some 3rd mask other distortion to give a smooth sound.
A 3rd harmonic in loudspeakers is well-known to mask higher ordered harmonics. I'd be interested in a study that shows that somehow this phenomena does not happen in amplifiers as well
Has a version of Distort been issued that allows one to model rising distortion with frequency? The author mentioned on this site that he had a version that could do that but didn't release it because the idea was 'too arcane' (paraphrasing). Since this is a common problem with most amplifiers employing feedback (which is most amplifiers in general) an audibility test that does not include this behavior is oranges and apples.
If you've not read it, Bruno Putzeys has a great article on feedback:
https://linearaudio.net/sites/linearaudio.net/files/volume1bp.pdf
If you read this article you'll see how important this aspect of distortion can be. We see that, as frequency rises, when the circuit meets its Gain Bandwidth Product limit, feedback decreases on a 6dB slope so distortion rises on a similar slope. As frequency continues to increase, its not uncommon for the distortion to rise at an even faster rate than that (because the feedback can fall off on an even faster rate depending on circuit variables).
If we take the example of 1KHz as a turnover point, we can see that at 7KHz the distortion can be quite a lot higher than the THD otherwise suggests! Our ears, as pointed out earlier, will assign a tonality to this and 7KHz is at the upper end of our ear's most sensitive range. In a nut shell if Distort does not model this you can't rely on the result.
If distortion rises with frequency, I think you'll find that the THD figure that an amplifier might have in practice isn't really accurate. IMO it really isn't a useful spec.