Ćuk’s resonant buck slashes magnetics

Article By : Michael Dunn

Dr. Slobodan Ćuk proposes an interesting hybrid charge pump/resonant buck that virtually eliminates the magnetics.

The name of Dr. Slobodan Ćuk (pronounced, roughly, chook) is familiar to most engineers as the originator of the eponymous Ćuk DC-DC converter architecture, justly famed for its low input & output ripple currents, and ability to act as a buck-boost.

So when a new converter architecture of his was recently brought to my attention, my interest was very definitely piqued.

I’ve been in communication with the good Doctor, but am still a bit unclear on the current status of his new design. A prototype appears to have been built, but details are sketchy.

The design is proffered as a resonant converter which, despite running at a fairly low frequency (e.g., 50 kHz), can get by with miniscule amounts of inductance – even just PCB traces – that resonate with large capacitances.

Dr. Slobodan Ćuk’s proposed resonant buck converter-cum-charge pump

As I’ve found the existing circuit descriptions a bit tough to follow (which doubtless says more about my skills than the circuit), here’s my own fresh take on the design:

If you ignore the inductors (replace them with shorts), we basically have a charge pump that natively runs at a 2:1 ratio:

Imagine the circuit more or less in equilibrium, with the switches as shown: The input voltage will be divided between C1 & C2. When the switches flip, C1 will be placed in parallel with C2 (via S2 & D1), transferring some of its charge to replenish C2.

With the inductors in play, each charge pump (CP) phase comprises one-half of a resonant cycle. This reduces the current spikes that may be seen in a standard CP design, and makes duty-cycle control of output voltage feasible without losing efficiency (since the inductors slow the rate of charge transfer). I imagine control circuitry would also have to implement a burst mode to keep output voltage from rising at low loads, as L2’s energy will keep getting dumped into the capacitors during the charge-transfer phase.

D1 & D2 can be actual diodes if you don’t mind their losses, but in most cases would be synchronous switches. In that case, Ćuk points out that the FET replacing D2 may need to block current when it’s open just as the diode would, but an N-channel FET with its source in place of D2’s cathode (as in one of Ćuk’s schematics) would have a body diode pointing the wrong way. Back-to-back FETs may be necessary, but then again, with the right control circuitry, I think the source could be on the left.

I confess my analytical skills are being stretched here, so if you think I’ve got it wrong, please gently explain in the comments your take on the circuit’s operation. Is this the clarion call I need to improve my simulator skills? We’ll see.

While Ćuk seems to like the idea of keeping the switching frequency low, I see no reason not to increase it to reap the usual rewards of smaller LC values and faster transient response (and pay the price of increased switching losses). What kinds of values are we talking about? Consider a few examples:

50 kHz: 1,000 µF, 10 nH

500 kHz: 22 µF, 4.6 nH

2 MHz: 6.8 µF, 1 nH

Sometimes, square roots really work in your favour.

Google around a bit if you want to read more about this Ćuk resonant buck circuit. What do you think of its potential?

Related articles:

Michael Dunn is Editor in Chief at EDN with several decades of electronic design experience in various areas.

Leave a comment