Mike,
Some years back now, when I was developing amplifiers pre-AKSA days, I looked very closely at this problem.
I tried the usual protection system on the voltage amplifier drive; a transistor which cuts base drive when signal excursion is too high. Self uses this approach; others use shorting transistors on the output stage.
I found that both these regimes, which comprise about 80% of the protection racket, conferred considerable sonic coloration. I persevered for some time, about six months, but could not remove the effect. Basically, it compressed the sound, making it sound less dynamic and adding 'constriction' which was quite apparent.
As late as the end of last year when I was developing the ultimate form of the AKSA (which was unsuccessful, incidentally, and which I codenamed Utopia) I tried again with an output stage variant developed by a Californian audio designer, an elderly friend of a pal of mine called Robert Jones, a brilliant ophthalmologist in Newport. This system too showed early promise, but required very careful calibration, and eventually because of the variability of the semiconductors used I decided to shelve it for the moment. I still think it has promise, but haven't the time just now to complete the design, which adds considerably to the complexity and introduces another 12 or so components.
Then I reconsidered the problem. Protection circuits are interesting; they protect themselves first, and the circuit they are designed to protect, last. They should be designed to take account of the SOAR curves of the output devices, and this makes them quite complex. There is constant debate about their effectiveness in the designer world, but I don't like them, and so I investigated the failure mode of the AKSA.
First up, the tweeter and midrange are capacitively coupled. DC cannot damage them; only oscillation. To my knowledge, no AKSA has ever destroyed a tweeter or midrange by oscillation, and from my own work on the CRO and various loads, I know that this kind of failure is so rare as to be almost unthinkable. That leaves the woofer.
The woofer can be destroyed by DC, and in fact during development of the N+ I did destroy a woofer in just this way. Not nice, but entirely my fault; I went beyond the limits, the amp went into DC, and the voice coil was deep fried. I keep the driver to show people how the speaker is designed; it's a Vifa M18 used in the AKSonics, a beautiful driver.
I found that in almost every case, one device fails before the other, but the 100R fuse resistors maintain a working potential to the offset control circuits in the amplifier even after the fuse is broken. This enables the other, good output device to hold offset to safe levels; in the meantime the sound is terrible, the fuse resistor begins overheating, smoke and smell is created and the amp is switched off in short order - after all, people are listening to it, so ordinarily realize immediately.
On a commercial product, lack of protection is more of a problem, particularly if used with very expensive speakers. However, when low cost, low parts count and DIY is concerned, if the amplifier fails in a reasonably benign fashion, protection may be eschewed.
Those who favor active systems will realize that the tweeter and midrange are indeed DC connected, and thus vulnerable where woofers are more robust. In this case, protection is perhaps a good idea, but a blocking cap is normally used. Incidentally, don't be squeamish about a blocking cap; there is one on the input, one in the feedback loop, and one relaying the music to the top end of the output stage. As if that were not enough, there are also blocking caps in the power supply, which actually serve as earth return for the music signal at the speaker, so the amp is redolent in capacitors passing audio signals, and thus one more, a quality metallized polypropylene, should do no damage providing the quality is up to the mark.
Lastly, a good protection system is indeed the Rod Elliott circuit, which senses DC at the output and then switches off a relay, disconnecting the speaker. However, this can introduce degradation in the music signal, as the contacts of a relay deteriorate over time, and are clearly audible when corroded. Nonetheless, if you use gold/rhodium plated relay switch contacts there should be no discernible problem within about five years.
Hope this puts it into the 'AKSA' perspective,
Cheers,
Hugh
