How do autotransformers affect sound?

Okay. My understanding is that the output transformers in tube amps contributes to the difficulty they have driving speakers that have wild swings in impedance curves and phase (802D3).  Power/voltage paradigm stuff.

Even though we make OTLs, the statement above is false. Output transformers **enables** tube amps to drive speakers like the 802D.
The 802 has a set of dual woofers.  B&W claims a 'nominal 8 ohm load' but in reality if you plan to use a tube amp, you use the 4 ohm tap and even B&W will tell you to do that. The reason is the woofer array is 4 ohms and that's where the most power from the amp will go. At that point, the higher impedance of the midrange and tweeter is really pretty effortless. A tube amp acting as a voltage source will have no worries on the 802D, provided it has enough power to do the job in the first place. The feedback in the amp will sort out any frequency response issues!

My question to you is do they sound better or worse on a good solid state amp, like a Pass Labs ect ect ? I've tried it and and the answer is definitely worse.

My understanding is Steve McCormick disagrees with you. IME, the ZERO needs about 45 minutes to warm up (I would not be surprised if this is part of the warmup phenomena many tube amps have) and you have to play with the taps and where the ZERO sits in the connection between the amp and speaker (I set them up with my normal speaker cable going to the ZEROs and very short cables, perhaps only 6" long, between the ZERO box and the speaker's terminals).

Its a fact that when you ask a transistor amp to make more current (drive a lower impedance) there is an associated degradation that has to do with a capacitance multiplication that is occurring within the output devices themselves. This causes the amplifier to take on a harsher character, something that transistors already have a reputation for.

IOW if you were to drive a 16 ohm load as opposed to a 4 ohm load with a transistor amp, all other things being equal the amp would sound better on 16 ohms.

Steve McCormick noted this recently in using a set of ZEROs (which are an outboard autoformer) to drive a 4 ohm load with his amps, even though his amps have never had any trouble driving 4 ohms directly. I interviewed 3 other manufacturers at CES about this and I found that quite independently of each other, there was a consensus on this point.

I suspect that MacIntosh sorted out this fact decades ago.

The turns ratio of the Mac autoformers, like the ZERO and our Z-Music autoformer before it, are all very low. The result is low distributed capacitance and lower than normal inductances, resulting in bandwidth that is likely to exceed that of the amplifier its being used with: up to 2MHz in the case of the ZERO.

I was not aware of the setup that Mac used in the old days, thanks to Kirkus for the history. Of course, back in the 50s and into the 70s, semiconductors simply did not have the capabilities that modern devices do so the autoformer approach is a good one.

The capacitive effect I mentioned earlier has some Miller effect similarities, but how they taught me in school is that in semiconductors it is much more profound. There are some devices that take advantage of this effect: variactor tuning in an FW tuner is a good example. In this case, a diode has a capacitive effect, varied by its bias voltage (and resulting current). By simply varying the bias voltage, the capacitor of the junction of the device is changed. This eliminated the need for an expensive tuning capacitor and also provided a simple access for Automatic Frequency Control (AFC). Variactor tuning is at the heart of most tuners made since the mid 80s at least- for example in nearly everyone's car stereo, unless you are playing a 60s or 70s relic :)

Semiconductor devices all exhibit this phenomena and it is one of the areas of solid state amplifier design that gives me the willies! As this is happening with all the devices in the circuit, not just the outputs, the effect can be magnified from the input of the amp to the output. I would expect at the very least that to avoid gain anomalies associated with frequency that the devices would have to be chosen carefully with good attention given to the bandwidth product of the resulting amp in the feedback loop. That is one of the marks of a good designer IMO.

I've not looked at all the class D modules out there, but the ones I have seen suggest that an autoformer would improve distortion performance for them too, IOW they show lower distortion (nearly an order of magnitude) specs into 8 as opposed to 3-4 ohms.

what you're describing is simply the tendency of certain topologies to present non-linear loads to their preceding stages. If this is the case, then how is this substantially different than a tube output stage running in class A2 or AB2 whereby grid current becomes significant for part of the cycle?

How its different is that the domain is more that of current than voltage and that there is a substantial variable capacitive issue intertwined. In a tube, once you get to the class A2/AB2 window, you are dealing with grid current but the capacitance is not really a variable. Semiconductors present you with this issue regardless of class of operation.

Eldartford, its true that solid state amps will be smoother sounding at higher impedance but the old Macs that we were talking about were solving a trickier issue of limited rail and breakdown voltages as well. The autoformers were a solution for that, to get the target performance of the amp inside the limits of the available devices. With the newer Macs this is certainly not the case. If you were to load one of those early Mac amps with a low impedance, I will guarantee that the results will not be as satisfying as with a higher impedance.

Arthur, FWIW the ZERO (zeroimpedance.com) has a very low turns ratio as it steps from 16 ohms down to 4,3 and 2 ohms. Its a nice problem solver when trying to get a tube amp like a small OTL to drive low impedance speakers. Its also very helpful with SETs, provided the speaker is otherwise fairly efficient.