Hello Bob/Jeep:

In the end I believe my only solution is to build.  But discussion before and after will assist my understanding.  On the British VAMRS web-site I found two or three very functional circuits that replaced a (not-now-common) vibrator with transistors or FETs.  So they built a - 12V to B+ PSU around a reverse-wired 240 .. 9-0-9V transformer.  That's one option.

However, some year ago I saw a hand-book on OpAmp design with a statment to the effect that a small modification to some particular circuit gave a completely different result.  "Only a dull and unimatinative mind could fail" - the book said - "to investigate the possibilities available from a few modifications".  Thinking about this - adding a couple of transistors to form an "H" bridge inverter seems to be a circuit that my dull an unimaginative mind must contemplate.

First, consider the circuit driving the main "H" bridge transistors.  Switching the transformer primary (i.e. the low voltage winding) with reverse-phase square waves suggests a problem at the moment the wave-form switches.  A short 'off' interval between each alternate power cycles would be 'useful'.  Consider the venerable CD4017 decade counter IC from the CMOS logic family.  Feeding this device with a square wave produces a sequence of pulses, each 1/10 the duration of the input signal.  If I leave outputs 0 and 5 unconnected and OR-wire output 1, 2, 3 & 4 to switch one arm of the bridge and OR-wire outputs 6, 7, 8 & 9 to switch the opposite arm then I have a convenient 'off' interval between alternate cycles.

Drew up the circuit last night.  I need 4 signal diodes for each 'OR' circuit; x8 1N4148 diodes in all.  I read that many vibrators run at about 80 .. 90Hz.  Clearly this is a frequency that a common E-I or toroidal transformer will 'accept'. 

This particular PSU is for a small receiver - specifically the 3-6MHz BC-454.  Many inverter circuits are free-running.  Birdies that 'get thru' appear at no specific frequency because the switching frequency changes with the load.  Crystals are cheap.  What if the PSU oscillator ran at a fixed frequency?  A crystal controlled switching frequency would put the 'birdies' I can't eliminate at regular intervals (frequencies) on the dial.  Annoying, certainly, but not so annoying as a birdie that appears anywhere on the dial.  With careful choice of frequency  I may even get a free 'spot' frequency generator from the PSU.  Some harmonic of the final switching frequency.  Say 50kHz.  Even if I can filter out every harmonic from the switching frequency - I get a bonus.  I'm thinking about driving the CD4017 with a CD4040 now.

At this point my power supply circuit is still simple and practical.  That's where the "Q" about highest operating frequency I can use to switch the power transformer originated.  At the moment I'm guessing it will handle 200Hz - maybe even 400 or 800Hz.  I reckon (I don't know) iron core losses will be high at 1kHz.   I know nothing about the iron in transformer cores - hence my "Q".   I'm confident some-one on this list will have experience in switching a conventional transformer core at some frequency above 50/60Hz.

Leslie.