Making It Differently

Anyone who hasn’t bought into the IoT/E as a fundamental shift in electronics should check out what’s happening in China. All appliances and electronics being developed for the home market are now being sold as smart-ready. This is a market that comprises roughly one-fifth (19.4% to be exact) of the world’s population.

In the future, whether or not you hook up a new washing machine or TV to the Internet, these devices will have the capability of communicating basic functions such as reordering supplies, connecting with a repair center, or working through some smart hub to save energy costs and overall energy consumption.

The big and immediate savings are still on the industrial side through the Industrial Internet of Things—smart meters, vibration sensors, more efficient lighting, better power management—but the biggest design changes will happen on the consumer side.

There are five critical elements in consumer-based IoT designs:

1. Power, as defined by time between charges for a battery or power savings on wireline devices;
2. Cost of devices, and all the components within them;
3. Security, particularly as these devices are connected to the general Internet;
4. Applications—either one killer or multiple non-killer applications;
5. Time to market.

There has been a lot of very public concern surrounding battery life and cost, as evidenced by the coverage of the Apple Watch long before a release date was set. (It doesn’t go on sale until April 24, although you wouldn’t know that from all the press it has received.) There also have been plenty of warnings about security—most of which have gone unheeded and many of which are being flatly ignored. And there has been much discussion about which applications are critical.

There has been far less discussion about how to bring these devices to market in a cost-effective manner within a shrinking time frame, however, despite the fact that this may be the biggest shift of all. While slick devices like the Apple Watch or smart TVs will employ advanced power management techniques and costly, well-designed security and routing, the reality is that most of the companies developing devices in this space are far more concerned with rapid time to market with enough customization to appear different with a competitive cost structure. The target consumer here aren’t the millions of Apple customers. They’re the billions of other people, and they’re starting to have a major impact on how semiconductors are designed and manufactured.

The basic engine in most of these devices is a sensor—and most likely a fusion sensor for maximum flexibility—coupled with a processor of some sort, whether it’s a CPU, MCU, GPU or APU, and connected to some sort of memory structure. All of this needs to be assembled and packaged quickly, and it needs to be done with enough flexibility to respond to market/ECO changes as well as narrow markets that can’t support a complex SoC design. The only way to accomplish this quickly is through some sort of platform-based multi-chip module approach, whether it’s 2.5D, fan-out, or some other system-in-package. Equally important, all of this has to achieve economies of scale that haven’t been reached in this market because the focus has been on moving to the next node with an all-in-one chip.

While the mobile phone revolutionized processor and SoC design compared with the PC, it still followed the same road map set forth by Moore’s Law. The IoT is about to change that. It’s well known that 2.5D and 3D integration are more power efficient and faster, because the distances are shorter between processor and memory and I/O is wider and less resistive, but many companies have shied away from this approach because there are still no economies of scale. That’s the reason most vendors are looking at die stacking for cost-resilient, time-insensitive applications such as high-speed networking and data centers.

The reality, though, is this shift is likely to come from a market where cost and time to market are the big drivers rather than power and performance, which are a bonus but not the real impetus. The ability to swap components in and out quickly and predictably will be much more important. And that almost certainly will creep upward into designs until the two approaches meet somewhere in the middle.

Despite very logical arguments to the contrary, the IoT is taking root in ways that no one expected. And as it continues to gain ground, those changes will become more widespread and mainstream, fostering even more changes.