MHz isolated power supply - we are at the inflection point

Ever since SMPS (switching mode power supply) came into the world in 1960’s, several technologies appeared that I consider exciting to designers – Integrated Magnetic (1970’s), Soft Switching (1970’s), MOSFET (1970’s), and Digital Control (1970’s). Except digital control, which is aiming to make power supplies smarter, all the rest are aiming to squeeze more power out of smaller size.

From a different angle, SMPS development after 1980 has been smooth. When I say “smooth”, I actually mean – a little bit boring. Therefore, whenever an engineer tells me that he’s found something of power supply revolutionary, my immediate response is – “Man, there is nothing new under the sun.”

Figure 1: Interesting developments through the 70's, then a bit boring until now.

When I started my career as a power supply designer in 1999, I believed the industry would never stop its pace towards higher power density, and I knew I had several tools in hand to help me to race in that direction – increasing clock frequency to reduce transformer and inductor size, employing soft-switching to reduce switching loss, finding better MOSFETs to reduce conduction loss. All the books I read said that too.

Led by that belief, I attempted to build a MHz isolated power supply (whatever the application) for years, till one day, a very experienced designer told me that, based on his study, 500kHz is the limit for isolated merchant power supplies, from an efficiency and cost perspective.

It seemed to be true. For example both the LM5025A and UCC2897A can physically run up to 1MHz. They have been dominant solutions for 50~200W telecom DC-DC converters, where higher power density is always appreciated, but I’m not aware of any design with these parts that runs above 500kHz.

In AC-DC applications, even the UCC28180, the latest analog PFC controller, is designed to run below 250kHz. PSFB (Phase Shift Full Bridge) and LLC are the two most popular soft switching topologies being broadly used in AC-DC applications including server, TV, etc., but I haven’t seen one who has operated a controllers like UCC28950 or UCC25600 at above 500kHz in products.  

I’ll tell you why. The controller has never been the bottle neck. It was the FOM (Figure of Merit) of MOSFET that limited the minimum switching pulse. There are several other things that also keep the industry from moving to MHz. For example, in MHz range, core loss of magnetic components increases significantly; parasitic inductance of capacitors cannot be neglected, etc. Nevertheless, MOSFET has long been the gating factor.

A real breaking-though technology – GaN FET – appeared in the last decade. Right from the beginning, the development group I’m with, has been engaging with GaN FET producer to develop gate drivers, say LM5113 and UCC27611. With its significant reduction of FOM vs. MOSFET, GaN FET also lights up my passion for MHz or even 10MHz power supply again.

Driven by fast growing users of wireless and wired communication devices, the power density of isolated DC-DC modules reached a new level in 2013 – the 864W out of quarter brick and the industry is marching towards 1kW in 2014.  MOSFET, sub-MHz hard-switching, digital control all play significant roles in the next 1-2 years.

I have already seen isolated power supply designers showing a strong will to leap over MHz. I’m optimistic in seeing MHz products in the market in 3-4 years. To make that happen, of course, the industry needs to fully understand how to play with GaN technology. Passive component suppliers need to take aggressive move towards MHz operation. Power designers need to find the best topology to drive ultra-high frequency without losing much power on switchers. When all these are realized, the landscape of isolated power supply will be very different from what it is today.