How do autotransformers affect sound?

Onemug and Atmasphere are right on. The impedance is indeed low (2.1 Ohms for the latest generation) and is constant thanks to the autotransformer. This allows the transistors to remain in the most linear part of the operation region so that distortion is reduced.

The distortion contributed by transistors when presented with a highly nonlinear complex impedance (eg, a speaker) is roughly 2 orders of magnitude higher than the distortion contributed by a bilar-wound split-winding autotransformer with tight coupling. Add to that a musical signal, which is itself highly complex, and the difference only gets larger.

The only real pitfall, other than the complicated and expensive manufacturing involved, is bandwidth. You have to make sure the autotransformers have a larger bandwidth than the output stage so as to prevent any undue high frequency roll off. However, this can be overcome with excellent coupling between the windings - which is what led McIntosh to develop their "unity-coupled" implementation back in the 1950s which they still use today. Look at Bode plots of the latest amplifiers and you will see they have superlative bandwidth despite the Autoformers, higher even than many direct-coupled designs.

So yes, from a technical standpoint, McIntosh is doing the right thing if you can afford the cost and handle the weight - which in hifi are obviously non-issues.

As far as the sound of Autoformer versus direct, I agree the Autoformer amps do seem "smoother" but at this point, I think it is actually that the direct ones are "grainier." It is a two way street. I used to have a MC7200, MC7100, and MA6500 which are direct, as well as a MC2125 and MC202 which have Autoformers. Which is better really depends on the quality of the speakers' tweeters. The higher the quality tweeters you have, the more obvious the benefits of Autoformers become. As for bass, I found the MC202 to have the finest bass of all the ones I've owned.

Arthur

Kirkus - Autotransformers can indeed be bifilar wound if you use split windings, as I mentioned. All you have to do is run mutiple parallel signal runs (since it is still an autotransformer, the signal references ground). This gives you the advantage of excellent coupling between the windings so long as they are wound according to the Right Hand Rule. We do this in our lab for 3-phase applications where EMI must be extremely low (coupling and EMI are inversely related, as are coupling and leakage inductance).

The 2.1 Ohm design, as I understand it, started with the famed MC2255 which used 6 winding sections, 5 of which are connected in parallel and a common rounds it out. But it had 1 Ohm taps which later got eliminated so the newer ones have fewer sections.

I agree the Autoformers lend natural DC protection and improve stability of the output stage. As an aside, this latter can also be the demise (everything in nature is a compromise). There can't be any denying that you are adding a considerable amount of damping to the circuit when you plop a complex inductor on the output. High frequencies DO suffer, no doubt. But if done right, (making transformers is more art than science) those high frequencies will be in the 200kHz (-3dB) range for audio applications which is higher than many output topologies.

In addition, one key reason for using these Autoformers is to protect the BJTs. They have a nasty tendency of overloading with temperature so they must be carefully controlled to remain in their safe operating area. The use of a constant "load" is put into effect as the control method of choice - and I have to say it is a very elegant solution for a significant problem since you get the added benefit of even better linearity (which is a BJT strength to begin with). The only downside is that you don't get the "doubling down" of power like the direct-coupled amps.

But in the end, operating an ampilfier with exceedingly low output inductance is asking for trouble. So some inductance is necessary in any case. Impedance is all we have to keep nature reined in.

Arthur

Kirkus - The way I know "interleaving" is by interleaving different phases at symmetrical angles - like 90, 180, etc. This gives a ripple cancellation to lower the noise floor. Works great in Class D applications where the ripple is way larger than rectifier ripple in a linear amp. This interleaving is done in a transformer and wouldn't work in an autotransformer because there is only one signal path, so pefect phase comes automatically, up to the point of core saturation. So bifilar windings and interleaved phases are different mechanisms.

Incidentally, bifilar-wound autotransformers are only really applicable to solid-state output stages because the step-down impedance ratio is low - on the order of 4:1 for push-pull BJTs. In tube amps, the ratio from the plate circuit is on the order of 500 to 1 so in this case a transformer is much more feasible.

McIntosh were the first to use paralleled primaries wound together in a push-pull tube amp output transformer to achieve a big reduction of crossover distortion, turns ratio, shunt capacitance, AND leakage inductance! It's an extremely clever and elegant concept. They patented it sometime in the 1940s and called it the "unity-coupled" circuit. They then adapted this idea for use with push-pull transistor amps by running the transformer single-ended (though still paralleling winding sections) and having the return path be at ground potential, which allows the use of a nonisolated autotransformer.

Shadorne - Excellent question indeed. From a technical standpoint, the use of air coils for the inductance of crossover networks is because they have extremely high linearity to preserve phase information to a high degree (audiophile ears are very sensitive to this). So they are high bandwidth components but their drawback is that their impedance gets out of hand for large inductance values because many windings are necessary to obtain a given value (because there is no core). Crossovers are one example that satisfies the criteria of desirable low-inductance values and need for high linearity.

You can look at adding a ferrite core a way of "cheating" nature into giving you more inductance. The price you pay is in bandwidth - so you must choose the frequency range desired by carefully choosing the right ferrite material (and there are many types). The high inductance values you get are needed for compact inductors and transformers.

Now this latter one is not to be confused with the "leakage inductance" in a transformer which is what's responsible for the effective impedance the signal sees - and not transformer action. This leakage value represents the power loss of the transformer and so must be as low as possible.

But in the end, Kirkus is right that it boils down to a cost/linearity relationship because ferrites that can handle very high frequencies are quite expensive for anything more than 10s of microHenries, and inductor size isn't a design issue inside a speaker cabinet. Not to mention that the improved bandwidth of an air core is probably audible in some fashion.

Arthur