Power Converter Topology Selector
An interactive decision table for switch-mode topologies — filter by input/output ratio, isolation, and power level, then compare stress, part count, and where each topology stops making sense.
| Topology | Fit | Practical ratio | Power range | Switches | Magnetics | Switch stress |
|---|
How to read this
The ranges are guidance, not physics
Every range in the table is where a topology is normally the sensible answer, not where it stops working. Flybacks have been pushed past 250 W and full bridges have been built at 50 W; both were the wrong first choice and the right final one for reasons specific to those designs. Use the fit column to shorten the list, then argue with it.
Conversion ratio drives more than you expect
For non-isolated topologies the ratio is the duty cycle, and duty cycle runs into minimum on-time, gate-drive bootstrap refresh, and controller resolution long before it runs into theory. A 48 V to 1 V rail at 1 MHz needs a 20 ns on-time — which is why that class of design moves to a three-level, hybrid switched-capacitor, or two-stage architecture instead of pushing a plain buck.
Isolation is a requirement, not a preference
If isolation is genuinely required — safety, ground-loop breaking, level shifting across a large common-mode difference — it removes the entire non-isolated half of the table. If it is not required, adding a transformer costs efficiency, area, height, and a magnetics vendor. Decide this first.
Where the losses actually go
Topology sets the shape of the loss curve; devices and layout set its height. Once you have a shortlist, size the real thing — the buck design calculator does that for the non-isolated step-down case, including RMS currents, output capacitance, and a first-order loss breakdown.