Editor’s note: This is the third in a series of guest blogs by Jeff Nilles from Texas Instruments. Jeff will be giving us his observations of APEC 2014 as he tours the conference this week.
On Wednesday, APEC held their annual social event, this year at the home of the Dallas Cowboys, the Fort Worth AT&T stadium. The playing field was definitely big enough to accommodate all of us. They warmed us up by taking souvenir pictures of us throwing or receiving footballs. After an hour, the queue got quite long when the Dallas cheerleaders joined the photo booth. Here is a picture of me with my wife, Jayne. Definitely the most popular event that night.
The food was great at the event. True south-of-the-border shrimp tacos were my favorite. It was much better than last year’s finger food and kid’s menu at the Long Beach Aquarium. By the way, after APEC 2015 in Charlotte, we will return to Long Beach for APEC 2016.
Now let’s talk technical. I thought Wednesday morning's Intel “FIVR” Haswell presentation was a bit light. I would have liked to see more discussion of the material that was in the paper. However, Ed Button did have a good picture of the 140 MHz air-core inductors on flex used in the 16-phase switcher. I was hoping for better than 90% efficiency for such easy duty cycles, but I do not think that fine-tuning the efficiency was Intel’s concern. Most people were impressed with the 80 MHz unity gain bandwidth. But that should be easy for a 22 nm process.
A paper that caught my eye Wednesday was the technical session 8.1 “Resonant Switch-Capacitor Voltage Regulator With Ideal Transient Response,” by Mor Mordechai Peretz. This is a soft-switching, switched capacitor buck, using a 1 uF external capacitor. It does have 4 amps of output current. With a Vin range of 2X, it does require three switches, and a 150 nH inductor in series with an energy storage capacitor to force resonant. Here the capacitor is storing all the energy, with really nothing in the inductor. See the diagram below.
I think we might see more papers on this soon, because capacitors can store more energy per area than inductors. This architecture could yield the highest power density in a few years. Right now the frequency is only 250 KHz.
There were a lot of GaN papers this year. On Wednesday, HRL Labs was showing a >50 V/nSec switch fall or rise time for its experimental GaN. It even showed a 100 V/nSec scope shot, but I assume that one blew their “AlN” gate oxide away. It was not the only one to address this, so a trend has surfaced. That is the problem with driving the gate of the GaN devices. This has led to the “DrMos” concept coming back, however this time for GaN, instead of MOS eight years ago. Last year there was lots of talk about modules for GaN because of package inductance problems. Both of these issues just seem to force a new type of package or co-packaging for GaN.
I will talk a bit about the Thursday technical sessions in my next blog — even my 5 MHz GaN flyback project!