completed a testbed series on this idea and have to say the results are
astonishling good.
My testbed was a new Gemini PA-7000 DJ preamp... It's architecture
is a passive attenuator with balance control feeding a unity-gain buffer
which drives a pair of SE outputs and an XLR (balanced) line driver. The
PA-7000 got POOGE'd to the maximum with ease. Replaced the NJM4558 unity-gain buffer with an AD826. Removed the NJM4558-based cross coupled XLR driver and replaced with a small PCB hosting a pair of Analog Devices SSM2142s. The SSM's are direct coupled to the output of the AD826 buffer op-amp. So, in line stage configuration, there are no capacitors in the signal path.
SSM's driving a variety of output transformers really shine... Highs are extended, transients are crisp, all side effects of the SSM2142 havings lots of muscle to drive the input capacitance of the XLR cable plus the input capacitance of the transformer at the receiving end. The SSM chip plus remote transformer gives you galvanic isolation too, which is the real advantage to this topology.
Sound wise, its amazingly satisfying... On the receiving side, I've tried
the LL1660/S, the Sowter L3575 and the Edcor WSM10K/10K. All
performed very well in the bass (no DC offset problems) and the top end
(plenty of drive). The Sowter 3575 did have a slight sonic advantage
in detail - likely because of it's nickel core.
So, the answer is, if you need to drive an input transformer / phase splitter on a tube amp, you can do it with the AD SSM2142 (or the Burr
Brown DRV134). Further, the SSM is easy to implement, just give it
a clean +/- 15V supply and supply it with some local decoupling caps
and you are off to the races.
-- Jim