Originally Posted by arnyk
What have you said that is inconsistent with the article mentioned in the OP?
The article makes no mention at all about ultrasonic failure of the tweeter other than thermal. No mention whatsover...In fact, the article goes out of the way to show in essence that ultrasonics cannot do so as clipping doesn't contain enough energy, same slant you've taken..
Both of ya's is incorrect. That is inconsistent with real failures of real tweeters.. I have repaired about a dozen of those ev tweeters used in the old K horn set, T-50's or T-500's IIRC( EDIT: it was the EV T35...jn
). Also, 2 Gauss-Cetec puppies, can't recall the model, but they had about a 5 inch diameter body with heatsinks on back. For more current tweets, the selenium D205TI's, I've repaired 6 of them for ultrasonic fatigue failure, and 2 for total thermal burnout.
Originally Posted by arnyk
How does one reliably determine the frequency of the signal that caused fracturing or overheating?
It's not overheating, it is fracture caused by cyclic fatigue failure from resonance.
For product which is hand assembled, the resonance frequency will depend on how the lead is dressed, it's length, how long it's in the gap flux, it's guage.. So even within one product line, I believe it would not be possible to analytically determine the frequency or frequencies (depending on the mode of the resonance). Across product lines, I would guess virtually impossible. It may be possible to get a range, but that doesn't help in the field.
To prevent it, address the failure mechanism. That would be the unsupported use of vc lead in/out wires which climb into the magnetic flux. I recommended to Selenium that they lay the wire against the diaphram and up the diaphram to vc former fillet, then bond it using epoxy or the polyimide the vc is assembled with.
edit: if you look at the pictures the author provided, note the lead in/out is held to the vc with masking tape. Not a well engineered solution, but for that particular vc assembly, there is probably no chance of resonance because of the coil inductance in the slot.
ps. I have worked on speakers where u/s oscillation toasted the VC's as well as destroyed the crossovers. Somebody had used a multiconductor snake to carry low level from the sources to the electronics, then used some of the twisted pairs to bring amplifier output back to the stage. Since all the pairs were in the same jacket, and they all had the same twist pitch, there was significant coupling between twisted pairs. Obviously, the person who installed it believed shielded twisted pairs cannot communicate magnetically.. Definitely incorrect when they have the same pitch.