tuga
Legal Alien
The mechanism of nonlinearities transferring energy from inaudible frequencies to audible frequencies is a real one but the magnitudes are too small not unlike a lot of other audiophile nonsense concerning DACs, cables, amplifiers, and the like. Resonance is primarily a linear motion although perhaps a tad less so near a peak (inertia and stiffness are strongly linear but they cancel at a peak leaving the weak damping force to control the motion and this can be more nonlinear although not greatly so). In order for a part of the small nonlinear part of the motion to alias and become audible the resonance needs to be strongly driven. There may be plenty of energy in a frequency response test signal but there is little energy in music at ultrasonic frequencies particularly on CDs. The numbers are too small. This shouldn't be a surprise given how many smart people design and use metal tweeters. And if it was a real problem then the first thing people would do to reduce the effect is low pass the signal. Is this standard advice when using speakers with metal tweeters?
I understand your point regarding the bandwidth of CD and of most mics. SACD and vinyl have considerable ultrasonic noise though, and JA mentioned this when measuring such speakers.
And why has (hard-domed) tweeter design focused on pushing the break up higher, away from the audible range?
What about noise-shaping and tweeters with a resonance peak that's 10dB above the audible range like the BnWs I mentioned previously?
B&W Silver Signature, individual quasi-anechoic responses of woofer,
and tweeter on tweeter axis at 45" and corrected for microphone response,
with nearfield woofer and port responses below 400Hz and 1kHz, respectively.
Read more at https://www.stereophile.com/content/bw-john-bowers-silver-signature-loudspeaker-measurements-part-2