While looking around the web for something completely unrelated, I stumbled across this awesome discrete switch-cap booster circuit by Ivan Sergeev. Now, technically, I suppose this isn’t actually a boost converter, since it doesn’t use inductive flyback — rather it’s a charge-pump stacker, but the nomenclature is secondary. It’s a cool hack, and I really like the way he’s laid the whole process out, from theory & concept through simulation and finally implementation.
A switched capacitor / charge pump boost converter works by repeatedly charging a flying capacitor to an input voltage source and then stacking it on top of the input voltage source. This is achieved in the ideal schematic above by closing alternate switch pairs SW1/SW4 and SW3/SW2. When SW1/SW4 are closed, C1 is charged to voltage source V1. When SW3/SW2 are closed, C1 is stacked on top of voltage source V1. After some start up time, the big hold-up capacitor C2 is charged to 2*V1, and the load R sees roughly 2*V1 provided that it does not pull the charge out of C2 faster than it is replenished.
My 5V regulated switched capacitor boost converter consists of two boost stages, a 74×14 based oscillator, and a simple cycle skipping regulation scheme. I implemented the top two switches (SW1/SW2) of each boost stage with Schottky diodes for their relatively low voltage drop, and the bottom two switches (SW3/SW4) with PMOS and NMOS FETs. The diodes work in their switch roles because D1/SW1 will be reverse biased in the discharge/stack phase, and D2/SW2 will generally be reverse biased in the charge phase (unless it is charging the second stage). Since the first and second stages are being operated in phase, the total boosted output is theoretically 3*Vin. The FETs are driven by the 74×244 buffer which has roughly 12mA source / 12mA sink to ensure relatively fast turn on time.
While charge pump circuits are readily available in integrated packages (the many variations of the ICL7660 spring to mind), there are good reasons for building one out of discrete components. Namely, the challenge and what you learn from it. Lucky for us, he wrote it all down so we could learn from it too. His idea of using diodes for the high side is also neat, even though it results in lower efficiency than using CMOS switches. Technology has a way of leapfrogging itself, though. Perhaps someday we’ll have integrated nano-rectifiers which are more efficient than the best FET profile, and such a design would become preferable.
He’s provided a complete schematic and BOM, along with the LTSpice sim file — check it out!
Great work, Ivan!