>
any experience with the Williamson or Ultralinear output stage? This is really a form of negative feedback to the screen grids
You should read Williamson's papers. It looks like AudioXpress ran out of printed re-prints, but they have it
on CD at fair price.
Williamson was one of those turning points. We had just come from triode amps with little or no NBF, to pentodes that gave a LOT more output but no damping at all. Since pentodes like 807/6L6 have more gain than 50/2A3/300 triodes, they sacrificed gain in a little feedback to make pentodes tolerable as loudspeaker drivers (and also to use cheaper transformers and power filters). Williamson has some interesting comments on these two breeds of amplifiers. Many trends that have been in vogue in recent years are what he wanted to move away from. Things run in cycles, and the amps he sneered at are now hot stuff. Oh well.
Of course Williamson did not use screen grid feedback-- he used triodes (but already he was forced to use triode-strapped pentodes: big audio triode development stopped when the 807/6L6 appeared).
He did use every inch of NFB he could manage, leading to the long-running fad for high-NFB amplifers.
He didn't do it to be cheap: his output transformer is extremely large and extensively interwoven, his power supply beefy.
He also apparenlty was one of the first designers to have an oscilloscope handy, and did a lot of thinking about transient behavior. The series grid resistors and caps are picked carefully.
A well built Williamson is still a fine amplifier, and generally more accurate than the older all-Triode amplifiers, certainly as you approach (but not exceed!) full power. But it is about 120 watts of heat and 60 pounds of bulk for 14 watts in the speaker. To be stable (and Williamsons are not always stable) with as much NFB as Williamson used, the output transformer is incredibly costly.
Williamson marks the start of the trend to using large feedback to cure all problems. While Williamson did it thoughtfully, many later designers were careless. And a Williamson does show the low-level high-order haze of NFB around multiple stages. At the time, higher-level low-order "fatness" was the common problem, and the Williamson reduced that in a big way. It took time to notice the low-level high-order haze in the new designs.
UltraLinear is another matter. It puts Pentodes half-way between Pentode and Triode operation. It splits the difference between high power (Pentode) and adequate damping (Triode). Frankly, I think if good audio triodes were available, UltraLinear would be pointless. But given that the mass-market wants the most watts per buck, power pentodes are all you can get in large sizes. UL lets you use pentodes more like triodes, at the cost of a little more trouble in the transformer. It also provides a NFB loop right AT the transformer, which gives less phase-shift trouble than taking the same feedback around many stages.
