Theoretical to practical, tests after failures after trials, success and understanding are all but relative to the fertile mind.

Sometimes, a Design Ideas submission doesn’t quite fit with the theme of that EDN section, yet still promises interest for our readers (you). Such is the case with a DC Transformer design that reader Newton Ball recently sent in. The title certainly is catchy.

Newton and I engaged in a bit of friendly debate over the circuit. It looked to me like the core of a forward converter (for you power engineers) or of an isolation amp (for you analog engineers), with explicitly driven synchronous rectification–nothing really new–just a totally generalised implementation or exposition of the circuit.

Newton countered that any capacitance on the primary would be greatly multiplied by the turns ratio (he envisions use in, say, a single-stage 48V-to-1V telecoms converter), reflecting and effectively improving the capacitor characteristics viewed from the secondary and taking advantage of CV vs. CV2 (volume vs. energy) proportionalities.

He’s done some basic testing: The circuit above was constructed, using toroidal cores for the drive and main transformers. In operation, it successfully passed substantial power in both directions.

I of course browsed the interwebs in research mode, looking at, among other things, synchronous rectifiers.

Newton continued: Transformers, AC or “DC,” provide a continuous, nearly instantaneous, link between ports and are completely reversible. This continuous linkage or referral or reflection, works for voltage or for current, in proportion to turns ratio. Near instantaneous reflection of impedance is by the square of turns ratio.

This quality is of importance in the original intended use, 48V to <1V conversion in data centres, inspired by an EDN article. It is difficult to get enough capacitance at the sub-1V level to support current surge demand. The DC transformer reflects every µF at the 48V supply input as ≥482µF at the ≤1V level. This reflection is very quick and independent of reversal (switching) rate.

So, do you think this generalisation ad absurdum of a forward converter pulls back the curtain to reveal a hitherto unconsidered attribute of the architecture?

Or is it what I like to call a “Partly-baked Idea”? Interesting to ponder, but not of practical value. Not yet anyway.

First published by EDN.