Jim has it right.
Here's the long story, backwards: I grew up with slide rules, which are logarithmically based. I had a career in acoustics, where frequencies and levels are most often on log scales. For these reasons, I am pretty good at approximate arithmetic (I'm terrible at precision work though!). In the early days of VALVE, Dan dubbed me "Brainiac" for this reason - the old archives have several articles under that moniker. At some point, we developed a design that was a step up from the ParaBee, using a plate choke that matched the TFA-2004 physically; Mike called it the BPC for Brainiac Plate Choke. We made all of two amps that way before other priorities intervened; we no longer have possession of either one. They had the BPC and TFA-2004Ni topside, with the power transformer (PGP8.1) between. The next step would have been to add DC heaters, filament chokes, and PSU chokes, with the filament and PSU chokes under the audio iron. That never happened, and when we went to the Paramount both audio devices had solder terminals (much less costly that end bells and flying leads!) and had to go under the plate. So the chassis got enough deeper to put the filament choke behind the power transformer.
Notice that in all these iterations, the iron is mounted with mutually perpendicular coil axes, and symmetrically placed. I did several experiments to confirm that this really would minimize the coupling - theory takes you a long way, but not all the way. In the ParaBee, the plate choke was in front of the power transformer and under the chassis; it did pick up a trace of hum but the AC filaments made enough more hum that this was not important.
An imaginative experimenter could come up with a huge number of other variant possibilities, if you assume you can get some of the iron in forms that are safe to put topside. :^)
