Surely it "works", it just doesn't produce the traditional distortion tone we have come accustomed to because traditional distortion circuits we have learned to love do not distort the signal at multiple separate bands. And traditional distortion actually has some fair amount of IMD in it though "on paper" all those unrelated harmonics of IMD must indeed seem very "unmusical".
I have several times warned about pitfalls of trying to apply high fidelity design principles to things that are essentially signal processors and meant to mangle the signal in ways that would make a hifi apologist loose his night's sleep.
Want another example?
Soft clipping. We can use a CMOS linear amp to very softly compress the signal peaks. But in practice the common complaint is lack of headroom and "blurry" sound without enough of "bite", which is the exact result of distortion creeping in too early, no matter how gradually and "softly". And then when you look at just about every CMOS linear amp distortion circuit they actually use them in cascades to increase headroom and make clipping harder. And only then it starts to sound more like those tube distortions we have learned to love, which don't clip even remotedly as "softly" as a CMOS linear amp.
I like your explanation of CMOS linear amps as used in distortion circuits, Teemuk. They can sound warm, but sometimes a little bit compressed and ''blurry''.. with not enough definition and detail in the sound. I guess it comes down to the approach the designer takes, there are a lot of ''derivative'' distortion pedals based on the early CMOS based circuits. I have seen a variation of CMOS circuit also incorporate reverse-parallel diode pairs across the NFB to get more ''bite'' out of them.
On the last JFET based booster I built, there's a switch selectable diode limiting in the NFB, two diode drops adds just a bit of grit and bite to the sound. selecting single diode drop gets more distortion, but also a more noticable "compression" or more accurately in this case "hard limiting". That harder limiting doesn't sound excessive though, the gain stage pot only goes from 0 dB to about +22 dB, so the diodes aren't chopping off huge chunks of the signal. The other advantage is that the noise floor stays very low, even at full gain this thing is really quiet when you're not playing.
Excessive noise is one of the pitfalls of using multiple high gain stages, that's true for tubes as well as the solid state and even DSP. CMOS linear amps and Mosfets can get pretty noisy if you're not careful. Sometimes people complain about hiss and noise using high gain distortion pedals or the distortion channel of solid state amps. Most decent op amps are pretty quiet, but if it's amplifying the signal by a factor of 50 or more and then chopping off most of it, that noise floor is coming right up with the signal.
Excessive noise is one of the pitfalls of using multiple high gain stages, that's true for tubes as well as the solid state and even DSP. CMOS linear amps and Mosfets can get pretty noisy if you're not careful. Sometimes people complain about hiss and noise using high gain distortion pedals or the distortion channel of solid state amps. Most decent op amps are pretty quiet, but if it's amplifying the signal by a factor of 50 or more and then chopping off most of it, that noise floor is coming right up with the signal.