You're pretty much on base with some some slight adjustments.

The switching frequency is about ten times higher about
100KHz or so. The high voltage might be full wave
rectified, doubled or boosted by a power factor corrector
.... but 300V is close.

The step down transformer is usually an E-E core or ETD
core. The toroids are output inductors. An additional
inductor is used for 3.3V in some supplies as a magnetic
amplifier, which lops off some pulse width from the +5V
output.

It's possible to modify a PC supply and add a B+ output
but you'd have to re-wind the transformer and output
inductor(s).

Usually the outputs are cross-regulated. That is the +5V
is the regulated channel and the other outputs track due to
the transformer turns ratios (sometimes with help from
weighted-sum voltage sensing.) That's why the 5V output has
to be loaded to get any output from the +12V.

If the 6.3VDC output was most of the power it would make sense
to use the 5V channel for this, strip off the other outputs and
wind a new secondary for your B+ in their place.

You're right, the transformer and filter components would be
much, much smaller. Though you'll need room for additional
EMI filtering... even so it's still much smaller. Some newer
PC supplies spread the clock frequency (frequency modulate it
with pseudo random noise) which can help.



Understand that inside a typical computer switching power supply there
is a higher multi kilohertz voltage (350v RMS????) produced by the
switching action of the rectified 115 or 230 volt AC input?

This high freq AC through various usually toroidal step down
transforner windings is then rectified to provided the plus and minus
5, 12 volt outputs etc. Correct?

We recently modified such a power supply to get a single 12 volt DC
output at about 20 amps for a particular, amateur radio application.
In order to do so we had to load one of the 5 volt ouputs with a
couple of amps in order to get the unit to work; but that's normal.

Question: Is it possible to get at that higher AC voltage inside and
directly rectify it as a B+ supply for tube equipment? While also
possibly tiddling one or more of the 5 volt DC outputs closer to 6.3
volts for tube heaters?

Rectified 10 or 20 kilohertz wouldn't need much filtering compared to
60 or 120 DC ripple of a conventiaonl power supply? Recall building a
number of conventional 50 and 60 hertz power supplies many years ago
with heavy chokes and large capacitors. But now have a number of
slightly older ex computer power supplies of various wattages around.

Or is the idea completly off base?