Thanks for your comments -- and the reminder of the Quad-II's use of this type of NFB as well.
The interesting thing about ARC however, is that the concept was used in virtually all of their designs. You'd think that if an afterthought, then future models would would have included a true cathode tertiary winding in the OPT, but that was never the case. Of course, the more you went that direction (more FB via a true cathode winding), the more the design would start to take on a Mac type look (loosing uniquity) -- and also suffer power output loses as screen grid voltage becomes more and more compromised, until you finally go all out with a 50% cathode winding and then use cross coupled screens to achieve a true Unity Coupled design.
Regarding the quality of OPT used, I think your comments are spot on -- as long as the secondary winding reflects the total HF characteristic of the transformer at at least the 4Ω and 16Ω taps. Several decades ago, my first experiments with this type of feedback were (unfortunately) based on the Acrosound series of transformers (TO-300 and TO-330), both of which will absolutely fail miserably if this type of FB scheme is employed with them, because with at least these specific transformers, each output tap presents a different HF response characteristic relative to the other. As a result, even when connected for
negative feedback, the output stage becomes an oscillator. On the other hand, the Dynaco series of OPTs presents an identical response characteristic at each tap, and so works perfectly when this type of connection is used. Back in the 70's, Eisenson experimented with this type of FB with the ST-70 design, producing excellent results. Ultimately then I agree with you, that it can be even a mediocre OPT that can be used to good effect -- as long as the response characteristic is identical at at least the taps relevant to this discussion. But also.......
I've never had a Fisher model 70 amplifier come across my bench so as to be able to work with that design, but it employed both UL taps and a separate cathode tertiary winding. Fisher rather quickly dropped the UL taps (maybe to eliminate any patent rights issues), and used just the cathode tertiary winding in their later model 80 and 100 "Boxcar" designs -- amplifiers I have had significant experience with. These amplifiers went through a series of build "gimmicks" throughout their production life to help promote stability, in addition to additional capacitance that the output stage required to be connected between the primary windings to maintain stable HF performance within that stage. Tests showed that the cathode winding represented 20% of the total primary winding, that ultimately provides about 9 db of NFB on each side of the push-pull connection. This effective NFB of 18 db, coupled with 15 db in the outer loop had these amplifiers operating with over 30 db of NFB, yet for all of that feedback, mid-band THD was still a measured average of 0.44% at 1 kHz, while IMD clocked in at 2.1%. 20 kHz power output suffered a 20% loss over mid-band power output, and tipped the scales at 3.3% THD when power output at that frequency was just 66.3% of mid-band power capability. Clearly, output stage/output transformer issues are at play with this design, as the rest of the design (rather typical Williamson type front end) is capable of far lower distortion performance than the unit actually delivers.
This suggests that (again), while a mediocre transformer may in fact remain stable when partially cathode coupled, and even provide good benefits at mid-band frequencies, it still takes a very high caliber transformer to produce the benefits of push-pull cathode coupling across the full audio band of 20 Hz to 20 kHz. Clearly, with the distortion results produced in the modified SA-300, this was achieved with this combination of connection, tubes, and OPT, while it is highly questionable that the purposed designed transformers of the 80 and 100 amplifiers ever did (same transformer used in both models). It is quite possible then that for all
practical purposes, the use of high Gm tubes, a quality OPT, and partial cathode coupling represents a very desirable (if not optimum?) mix of good effect (distortion reduction), good stability (as long as the secondary tap HF characteristics are all equal), and practical application (allowing the use of rather standard driver configurations), beyond which, a jump to a full Unity Coupled design may be the next
most worthwhile step to take. No doubt that Walker knew what he was doing. But it just may be that in spite of having the design capability, Johnson recognized that partial cathode coupling using the output winding was the most practical way to go -- hence, no dedicated cathode tertiary winding (cost never hindered any ARC design!). With the results produced in this modified SA-300 project, it's hard to argue with that reasoning.
Thank-you again for your comments and input on this subject. If this project and discussion proves anything, it's that just as with output transformers and vacuum tubes, the application of NFB -- and how effective the results are it produces -- are as much a product of art and creative thinking, as it is design and engineering.
Dave