Practical Chip Design

I don't suppose there's much debate now that we are going into a serious recession. Even the US economy can't vaporize this much wealth and soldier on as if nothing had happened. Especially the US economy, with its severely over-extended consumers, huge trade deficit, tenuous hold on manufacturing, and crumbling infrastructure, is not going to go on without changes. At the very least, one presumes, the US will see markedly lower consumer activity until consumers have paid down their unsecured personal debt, have some confidence about staying in their homes, and have some plausible scenario for health care and retirement. People are just too frightened right now to tramp out and start shopping.

So where does that leave the heavily consumer-dependent semiconductor industry? The bad news part of the answer is obvious enough to not require much elaboration—we aren't going to need to design or build parts for toys that consumers aren't going to buy. But there are good-news parts to the answer as well. One of these was nicely illustrated by a recent conversation with Kazu Yamada, vice president and general manager of the custom SoC solutions business unit at NEC Electronics America.

NEC, like pretty much everyone else in the ASIC business, has been seeing a general decline in design starts. Part of Yamada's strategy has been to find emerging segments in which ASICs have a natural advantage over their alternatives, and specifically over FPGAs. Since NEC has not focused on the extreme high-end of the SoC business, simply outrunning FPGAs on speed and density wasn't the best answer. Instead, Yamada chose to talk about applications that require very high energy efficiency.

To most people that would mean mobile consumer devices, which would leave NEC spiraling into the black hole of vanishing consumer spending. But that's not what Yamada has in mind. Rather, he sees opportunities to exploit low-power processes in—ironically enough—power generation and distribution. So why would a facility that is creating or transferring power need low-power electronics? The answer is kind of interesting.

On the generation side, Yamada is looking at photovoltaics. "Today, there is so much variation among photovoltaic panels from the same vendor that you cannot connect them in parallel—you would end up applying voltage to the weakest panels," Yamada explained. "So you have to connect them in series, and then the weakest panels limit the output."

Yamada said the solution is to integrate digital intelligence and a solid-state regulator into each panel, so that the panels have uniform output. To be a net gain for the system, this module has to have extremely high efficiency—a high-efficiency regulator and almost zero-energy digital stuff.

Another example Yamada offered was load control. In the power grid of the future, individual consumers would be able to sell surplus power from their windmills or PV panels back into the grid, and conversely, the distributor would be able to shed low-priority loads as necessary to balance the network. In this scenario the lowly Watt meter on the side of the house becomes a mini switching center, interacting through an RF link or a power-line network with the system operator's control facility on one side, and with the controllers on individual generators and appliances on the other side.

This creates a demand for custom chips that combine ARM-7-level computing with a full range of networking capabilities, but at very low power. And the scenario puts a premium on security, since the consumer is in effect giving a public network access to every electrical device in or on his house, and the electric company is performing metering and billing over the same public network. But once again efficiency is of the essence—the less energy consumed by the control electronics, the finer-grained the control you can exercise, and the greater the network-level returns. So there is plenty of scope for challenging low-power design involved.

This conversation has become more interesting in the couple of weeks since it took place. Then we were facing uncertain times. Now, we appear to be on the brink of a prolonged recession patterned on the one Japan experienced beginning in the early '90s. One of the most plausible of the solutions that have been proposed is to stimulate demand through infrastructure spending—not unlike one of Japan's countermeasures. Modernizing the archaic US power grid certainly falls in the category of infrastructure spending that would both create jobs and have a positive RoI. That would put Yamada's ideas about adding low-power intelligence to the power grid in a sweet spot, not in the accretion disk of the black hole. Maybe we can learn some things from the Japanese experience.