The big story of 2007 will have some serious ramifications in 2008 as companies react to Power-One’s victorious lawsuit against Artesyn. The jury ruled that Power-One’s patent claims over the digital control of point-of-load (POL) power converters were valid. It also ruled that a product proposed but never built by Artesyn had infringed on these patents.
This decision throws a monkeywrench into the plans of companies like Silicon Labs, Zilker Labs, Primarion, and even Astec, which expected to use PMBus in a role stretching from the front-end supply in the data center to the dozen or so POLs on each server board.
When asked if his company would be licensing its technology to other companies that wanted to use PMBus for POLs, Power-One executive vice president Dave Hage played his hand close to his chest, saying only that anyone who was thinking about digital control of POLs should talk to Power- One first. This seems to imply that Power-One is a lot more interested in filling sockets than in collecting licensing fees.
Good News and Bad News
That’s good news, from the broadest industry perspective, because it validates Power- One’s conviction that smart POLs have a role to play in taming the data center’s voracious appetite for watt-hours. Even more than PMBus, Power-One’s Z-bus provides handles for dynamic control of each POL. Moreover, it’s completely interfaceable with PMBus-based control further upstream in the power-distribution mesh.
On the other hand, it was nice when we had a horse race. Competition forces innovation. For example, Zilker Labs first pushed the idea that temperature data from the multiple POLs on every board in a server cabinet could provide a nearly instantaneous 3D picture of temperature gradients inside that cabinet and that this could be used to dynamically adjust cooling-fan speeds for optimum efficiency.
In any event, the jury verdict only represents the penultimate chapter in the story. There will be an appeal. Meanwhile, one industry insider notes that nobody has been watching Texas Instruments, a key player in power. It’s a member of the PMBus Consortium, and its UCD723 buck gate-driver and UCD91xx and UCD92xx controllers all target POL management via PMBus. While these parts don’t appear to be in production, they are on TI’s Web site.
It’s hard to imagine a company with a bigger patent portfolio or a more astute bunch of patent attorneys than TI, so things could still get interesting. Unlike the courtroom battles so far, the final events will likely unfold quietly, and the details may only be revealed in the back pages of annual reports and quarterly SEC 10-K filings. Yet the showdown is coming. As Yogi Berra said, it ain’t over until it’s over.
Energy harvesting could be the Next Big Thing. During a recent Silicon Valley cocktail party, an entrepreneur who sold his last company for a tidy sum was talking about a scalable wave-powered sea vessel that could maintain a fixed latitude and longitude more precisely than a tethered buoy. And that’s not the only plan in the works.
Darnell’s energy-harvesting conference last summer revealed the energy-harvesting features being considered for Boeing’s 787 Dreamliner. For example, thermoelectric generators in the composite exterior would generate power from the temperature difference at altitude between the cabin and outside air. Also, piezo pushbuttons in the passenger- seat arms would, when pushed, self-generate the power for their built-in transmitters to send the passenger’s request to light the overhead light via a wireless mesh network.
It sounds bizarre at first, but copper wiring is heavy and expensive, and a wireless mesh gets rid of a lot of it. Moreover, the scheme makes reconfiguring the airplane cabin for even more cramped seating (or for switching rapidly from passenger to cargo configurations) easier. Eliminating the connectors could even make things more reliable.
The keys to energy harvesting lie in better generators (see the figure) and in energy storage and release. Capacitors seem more efficient than batteries, but uncharged capacitors look like dead short circuits, and their output dV/dt needs to be flattened out. Dealing with that efficiently on a small scale presents design challenges.