The closest approach: what amplification?

Unsound asked Bill about high output impedance... obviously I'm not Bill, but I'll take a shot.

If the amplifier has a very high output impedance (very low damping factor), to a certain extent the loudspeaker's impedance curve will "modulate" the amplifier's wattage output; that is, the amp will put out more wattage where the impedance curve is high and less where the impedance curve is low. This is the opposite of the behavior of an amplifier with a low output impedance, where the wattage goes up when the impedance goes down and vice versa.

Putting on my speaker designer hat, in my opinion the answer is to design speakers with as smooth an impedance curve as is practical. They will thus have virtually the same tonal balance with either type of amplifier, and one consequence is that such a speaker will allow a more apples-to-apples comparison of various amps. In practice it's usually not possible to avoid some impedance peaking in the bass region, and it might not even be desirable to do so; the increased wattage output of a high output impedance amp in the bass region can be leveraged to extend the bass deeper than it otherwise would have gone.

Having built such speakers for several years, and having had the opportunity to hear them on a fair number of amplifiers, in my opinion high quality OTL and high quality SET amps in general do sound the best. For most of the past few years my personal amps have been OTLs.

If the speakers were designed for contstant-voltage amps (most are) and have significant impedance peaks or dips above the bass region (most do), then I would still look for an amp with little or no global negative feedback.

All that being said, let me give an analogy from the world of amateur speaker building (my thing for twenty-something years). Many's the amateur speaker builder who sets out with high hopes intent on combining the best woofer with the best midrange and the best tweeter in the best enclosure using the best crossover. Almost never does the result live up to the promise of all these bests. The reason is, our enthusiastic amateur designer friend isn't taking a wholistic systems approach, based on the best outcome, and working backwards from there to assemble those components that will give this best outcome. Turning now to the question of what is the "best" amplifier, I would say look at the amplifier+speakers+room as a single system and go from there.

Duke
dealer/manufacturer

Unsound, an across-the-board raising of the impedance curve (a la Speltz autoformer) will usually make a speaker more high-output-impedance-amp friendly, if needed. Redesigning the crossover to smooth out the impedance curve is seldom practical, but in some cases the addition of an external filter to tame an impedance peak (often present in the crossover region) can be beneficial. I would estimate that taking the impedance curve into account in the crossover design phase can more than double the workload, as the designer is trying to simultaneously optimize both impedance and frequency response, and often that which helps one hinders the other.

I like Samujohn's comment:

"I strive to recreate not the most mathematically complete model, but rather use a bit of trickery to recreate a bit of the emotional experience that draws me to music in the first place."

In my opinion the goal is to recreate the perception of the original (or fabricated) event, and focusing on perception is different from focusing on recreating the waveforms. For example, very low percentages of high, odd-order distortion are both audible and objectionable, while very high percentages of second harmonic distortion are inaudible or barely audible, and are not objectionable. I've seen data that indicates 30% second harmonic distortion is statistially inaudible, but it looks bad on paper if the yardstick we're measuring with is THD. A perception-based yardstick has been proposed, but has not found acceptance unfortunately.

Unsound, obviously I should have made it clear which part of my response was about things an owner of existing speakers could do, and which referred to something that would have to be done during the design stage.

In my opinion using an autoformer is something a speaker owner would do to make his speakers more compatible with OTL amps.

If we're just looking at the design stage, then yes it is easier to smooth the impedance curve by lowering it. But if the end goal is compatibility with OTL amps, for example, lowering the impedance curve can be counter-productive. Some of the design choices that lead to a medium to high, and smooth, impedance curve need to be made before the crossover design stage. For example, I can't expect to build a 2.5-way system using two 8-ohm 6" woofers and end up with an OTL-friendly design.