This post was written by Dave Gillespie and not I. Also it was brought to my attention by member Pio1980. It is a very concise explanation of how the Unity Coupled Stage works.
This question was asked on another AK forum, specifically as it relates to biasing, and the screen grid connection. I was asked to repost the answer I gave there on this forum, for the benefit of those here who are interested in the McIntosh design. That post follows:
"The Unity Coupled circuit is a tertiary feedback arrangement. But to be more specific, it is exactly a 50-50 arrangement. That is, whereas in a conventional arrangement, the load for either tube is presented entirely at the plate, in this case, 50% of it is at the plate connection, while 50% is at the cathode connection. These two sections together then make up the total impedance that would normally be seen at just the plate in a conventional arrangement.
Since the peak AC voltage seen at the plate under full power conditions would normally be about 350 volts in a conventional arrangement (assuming the same tubes and operating/loading conditions), that means that 175 peak volts is appearing at the cathode with the 50/50 split arrangement that the Unity Coupled design uses -- which produces some serious challenges for the driver stage, requiring the boot strapped cathode followers etc., that appear before the output stage.
But this 175 peak volts that appears at the cathode would also cause serious reductions in power output if the screen voltage were held constant. If this were done, the application of increasing power output causes the cathode voltage to increase on the tube that is being excited, which against an otherwise steady screen voltage, would produce a net reduction in screen/cathode voltage as power increased. With an effective ever increasing reduction in screen voltage with increasing power output then, maximum power output becomes significantly limited.
What to do? Attach the screen to the opposite tube's plate! Why? We know that when one tube is conducting, its cathode voltage is rising, and therefore, its plate voltage is falling. But the exact opposite is happening in the opposite tube. It's plate voltage is INCREASING (while its cathode voltage is falling). Since the cathode voltage is rising in the tube that is conducting, and with the screen of this tube connected to a source that is also rising by the same exact amount (it must since the windings are all 50/50, remember?), then it means that now, the effective screen voltage in each tube does not change, in spite of increasing power output. Quite a neat, but very necessary little trick.
As for the bias, because there is so much feedback employed in the output stage (100% if you base it on just the signal developed at the plate), then a critical setting of bias became unnecessary for Mac to achieve its stated performance specifications. They could even relax quiescent current draw, and still meet the specs because of it. With such relaxed current levels, virtually any tube could be installed without any real concern for excessive quiescent dissipation levels, since such a relaxed setting would accommodate virtually all tubes in this setting. No doubt, the bias could be tweaked for even better performance, but to achieve the published specs, it wasn't necessary.
The whole affair requires very tight coupling in the OPT between the various windings, which in part, generated the "Unity Coupled" name given to it because the windings are so closely coupled. Special bifilar and trifilar winding techniques were used to attain such tight coupling, with the results being more expensive to produce, but producing performance which speaks for itself.
I hope this helps!
Dave"