The sound of speakers is distortion.

I think they may get paid by the word

Oh, you don't need a Unicorn speaker. Any violin, guitar, singer or even yes cowbell will do. That's your gold standard: the whole recording/playback chain recreates the original event.

Speakers being the last link in the chain, so many opportunities to mess up the sound long before it gets to the speaker, even if the speaker is perfect its hard to tell. Which by the way slays the old "spend the most on the speaker" canard. Yet still it seems to be true its the easiest one to spot flaws in, so maybe there's a reason its still around even though it can be proven not to work very well.

The thing about speakers, there's way too much focus on sins of omission. We measure frequency and say see, bass is rolled off, not good. Or the highs, same thing. When in reality it turns out to be very easy to enjoy living with such a speaker- if it does a bunch of other things well. Sins of commission though, if the box was poorly designed so it adds a woody hooty character to everything, most people won't stand that at all no matter how perfectly flat the response.

What this all means of course is speakers are no different than anything else. You just go and listen and get the one that sounds the best. That is after all why we buy this stuff, to listen to it. If there's anything naive it would have to be the idea there's any other way of going about it.

Speakers are the only component that cannot escape the scrutiny of the measuring microphone. If it measures better, it will sound better! Time/phase, dispersion, frequency response, power handling are all well-understood (and easily measured) qualities. The days of "cut-and-try" of speaker design are long past! 

Two answers to this:

Yes, an ideal speaker would simply move the air as instructed by the amplifier. Additions such as harmonic distortion and subtractions such as low frequency rolloff detract from that goal.

No, it's impossible to model an ideal speaker, for reasons of room interactions and the limitations of stereo. Which is to say that even if we could engineer a speaker that did exactly what we want within the laws of physics, there would still be choices to be made with respect to dispersion, polar pattern, and frequency response.

For example, it might seem desirable to make a speaker that directs sound only at the listener's ears, eliminating room interaction. But such a speaker would sound awful with conventional two channel stereo, since the room reverb is needed to provide a sense of ambiance; in an anechoic chamber, stereo sounds like a slit between the two speakers.

Similarly, it's long been known that flat response sounds terrible with two channel stereo (but not multichannel). A speaker's response has to slope down if it is to sound real.

So we do know many of the things that matter, but in other cases, the solutions aren't entirely simple or clear.

I appreciate your position, roberjerman. But speaking from long experience and from many conversations with other designers, there are no simple measurements that represent how a speaker sounds. Much of this problem is from the microphone not hearing as we do.

If a designer hears something wrong, then only many indirect measurements can be made, with many specific listening sessions, to zoom in on what might be wrong.

If a designer uses physics to predict something might be wrong, he must imagine how that problem should sound on music and then how it will measure, how it can be measured, using what input signal, etc. But most designers do not rely on physics, which is why so many different designs are on the market.

Elizabeth, I've enjoyed some of your posts, thanks. What you write is true enough given how different most speakers sound. My experience and many others has been that, when a speaker's design gets better in the time-domain, not just the frequency domain, all listeners' opinions begin to converge, agreeing the sound is 'correct' on more and more instruments and voices.

This would be a speaker design becoming more and more time-coherent, more free of cabinet reflections and diffractions, of internal resonances, cone-breakups, and has a simpler crossover with fewer crappy parts. Those are not revealed in the usual Stereophile measurements.

And you are right- one must listen to what is most important to you. However, sort of to your point of some unified basis, I've always recommended we listen first to the voice range and work outwards from there. This makes sense when we agree we all know how voices are supposed to sound, that voices can be our reference 'base'.

Best regards,
Roy
Green Mountain Audio

Good points, josh358

Some great points here. I used to think that a distortion-free speaker would theoretically let the listener hear precisely what the mastering engineer heard on his studio monitors.  But, clearly there are playback systems that will make that recording sound far better and more lifelike than anything the mastering engineer heard when he created the recording.  Should we call this ‘good’ distortion? Of course not.

On the other hand many of us have heard speaker systems that blow us away on first listen, but we do not own these speakers, because we suspect that they would produce fatigue with extended listening. Perhaps this is something that we could label ‘bad’ distortion, or ‘mixed bag’ distortion...not!

Hence my conclusion that the idea of distortion as applied to loudspeakers is simplistic and naive.

Nelson Pass made a gizmo that does exactly that (adds a "little sloppy"  "even ordered" distortion to the amp). He sent amps to people with an undescribed knob and they all liked the "little sloppy" knob turned up, so he gave a pile of the gizmos away at a Burning Amp thing. What a guy.