Will Wearables Work Well Enough?

By Ed Sperling & Ann Steffora Mutschler

The rollout of the Apple Watch later this month has reset expectations for the wearable electronics market, just as early implementations of the Pebble, Fitbit and Google Glass helped raise awareness about a new level of portability and connectivity. Early projections are for strong sales, which in turn will propel a new level of connectedness for the .

There are still some kinks to work out, however. All of these devices share a similar set of problems. There are questions about which functions should be included and how useful these devices can be in a relatively short window between battery charges. There are thermal issues to contend with, caused by leakage in both standby and active modes, which are much more noticeable in a wearable device. And then there are questions about how much it will cost to get it right, both in terms of dollars and silicon.

These are critical issues that typically are decided well in advance of products hitting the market, but in the wearables market they are being solved almost in real time and frequently after the fact. The growing buzz over the IoT and the hyper-competitive market for wearables has forced many companies to rush products to market quickly with hopes that problems can be solved in the field or in future versions.

Solving these issues isn’t going to be simple, however. It’s hard to argue with physics, economics and market realities. And unless these problems are resolved, either business expectations will have to be adjusted, functionality will have to be restricted, or architectures will have to change significantly.

The three faces of power
There is a trio of power-related issues when it comes to wearable electronics. The first involves wireless communication, which is the basis of any useful mobile IoT device.

“Whether it’s a watch or other wearables, our world will be more wirelessly connected, and RF consumes a lot of power,” said Mike Thompson, senior manager of product marketing at Synopsys. “You’ve got RF, an antenna, and a display that needs to be readable, and getting that within acceptable power limits is not a trivial exercise.”

He said that with multi-function smart watches, 25 to 30 hours of battery life is the upper limit if it’s communicating on a regular basis with the outside world. That number may be optimistic, depending upon the frequency and duration of communication between a wearable device and the outside world, the effectiveness of the antenna, the signal strength of the connection, and the amount of data being transmitted and received. Others have pegged the number closer to 12 hours, and just several hours under compute-intensive loads.

“Whatever data this thing is collecting is likely to be shared with the outside world to one of the radio protocols, which means you have to have one or more radios alive,” said Serge Leef, vice president of new ventures at Mentor Graphics. “And then, if you intend to communicate with the user in some way, there needs to be a display and that brings a whole new level of power demand.”

A second challenge is leakage current, which is difficult enough to deal with in a mobile device. As the size of the device decreases, and as proximity to the body increases, the challenges go up proportionately. There simply are fewer ways to dissipate that leakage, which shows up in the form of heat, and the problem is exacerbated by process technology.

Each successive process node in the mainstream IoT world—65/55/40nm—grows increasingly leaky in standby mode. While major efforts are underway to improve this, with new LP processes at those nodes, as well as FD-SOI at 28nm and finFETs at the most advanced nodes, there always are tradeoffs. Performance on 28nm FD-SOI isn’t as high as with 16/14nm finFETs, but it’s less expensive to manufacture because there are no finFETs and masks don’t require double patterning. At 16/14nm, devices are faster but they’re harder to design and current density is higher. Chips developed using 65/55/40nm processes are less expensive, but current leakage is higher, which requires more sophisticated power management schemes than in the past.

“There have been steady improvements,” said Gordon Cooper, international product marketing manager for NXP controllers. “Active power is better, but leakage goes up. So you don’t get the benefit from process nodes that you might expect.”

One way to deal with that is with sensor fusion, using low-power sensors whenever possible the way an ARM big.LITTLE processor shifts lesser tasks such as e-mail or Web surfing to the smaller core while reserving the large core for such compute-intensive tasks as video processing. “So in active power mode, we have one that runs at 100MHz and another at 50MHz, and you can switch between the cores,” said Cooper.

Greg Fyke, marketing director for IoT wireless products at Silicon Labs, said the heterogeneous core approach will work effectively for higher-end devices, but he noted that at the lower end the typical strategy is to have one microcontroller doing everything. “These kinds of architectural choices haven’t always been the focus, but they are coming into play,” he said.

These kinds of market nuances are just coming into focus in wearables. “We think there is a great opportunity for new wearable devices,” said Frankwell Lin, president of Andes Technology. “Some types of these will be standalone, un-tethered devices that seldom connect to the network. Other types of these, like the iWatch, will always be connected.”

Lin noted that the big difference is communications capabilities. The “always connected” types of devices will require support for a variety of communications, from NFC to WiFi and others. In many ways, ‘always connected’ makes wearable devices, from a semiconductor standpoint, a lot like low-power IoT devices.”

A third problem with power involves the batteries. So far, improvements in battery technology have been dwarfed by leaps in hardware power management and better efficiency. But work is under way at many universities around the globe, both to improve battery technology, reduce the cost of those batteries, and to more effectively charge them using energy harvesting technology. This will likely bring big changes in cars and many mobile devices, but the impact on wearable electronics could be minimal because of size and weight restrictions.

Tradeoffs and killer apps
Power extends to the software that needs to run on these devices, as well, which is potentially problematic because so far there is no clear killer app or set of applications that are considered “must-have” features.

“The key to wearables is low power, so devices that have been designed over the years for low power applications are the ones that make sense for wearables,” said Mentor’s Leef. “Now, once you add bright, colorful displays, then you are challenging the battery. The pacemaker/defibrillator guys have been making low-power devices like pacemakers that have to sit inside a patient and operate on a single battery for 12 to 15 years, so there’s nothing new about creating low-power systems. But those things do not have radios or displays. What makes wearables interesting are the sensors that are reading some kind of data in real time. There needs to be some low-level processor activity just for collection of data.”

There’s another angle to this, as well. The cost-pressure on designs is extremely high because many of these devices are being sold into consumer markets in places such as China, which is a couple years ahead of the United States and Europe in adopting the IoT with strong backing from the central government. In fact, there is a huge push by the government to make all appliances and home electronics “smart-ready.”

Being successful in this market isn’t just a matter of getting to market first, though. It requires a delicate balancing act between squeezing costs out of the design process—a combination of packaging, platform technology and standard IP—and doing that within a very short market window, with enough customization to provide differentiation, and with enough sophisticated power management built in to make the end devices useful. The question being asked now is, ‘Given these parameters, how much battery life is enough?’ No one is sure, but there is almost universal agreement that right it’s not enough and designs need to be changed to reflect that.

“The most important thing in these devices is power management,” said Charlie Zhi, president and CEO of China’s Brite Semiconductor. “If you have to charge a device every week, that’s not a problem. And if you’re not using it very frequently, it’s not that hard to create a device where the battery will last three months. But there’s also not a killer application there yet.”

And therein lies one of the big problems. What is the killer app?

“The world of ‘always-on’ sensor processing, whether it be in wearables, mobile, smart home, or IoT, is rapidly increasing with many new and unique use cases,” said Jake Alamat, marketing manager for MCUs at NXP. “What excites me the most are the applications for natural user interfaces, medical applications, and augmented reality. What will make these applications successful will largely depend on ease of use, interoperability, and the positive impact on consumers daily lives.”

Changing the formula
Behind the scenes, IP vendors are preparing for this shift and the eventual definition of a killer app—one that likely will include some biometric sensing coupled with the ability to locally process some data and connect to an external device or devices. This explains why Samsung made a big deal about its push into digital health last year, and why Apple has focused on health and fitness apps in its upcoming smart watch.

How the components inside these devices are integrated and assembled will likely be unique, along with the software that runs on them, but the actual IP blocks and memories will not be. One indication of this shift is an increase in the amount of hard IP, which is already pre-developed and characterized, and full subsystems with sensors for vision, sound and motion.

“We’ve been broadening our hard IP catalog,” said Martin Lund, senior vice president of the IP Group at Cadence. “So you can connect into the controller and connect that into the system, but the chain to interconnect the IP controller to the hard IP is very complex and it can take time.”

Along the same lines, ARM has developed a suite of microcontrollers that provide a range of options from wearables to less power-sensitive devices.

“If anybody has done a good job optimizing designs for power it’s ARM, and there’s a lot of software for ARM, and there are a lot of people who know how to design with ARM and a lot of people who know how to program for ARM,” said Mentor’s Leef. “If I were running a startup focused on wearables and I was relatively well funded, that’s the way I’d most likely go. But people who are not well funded typically would have a tendency to pick up more mature processor cores like 8051 — things that are from 25 years ago that are readily available in public domain. Those are not great as far as power is concerned but they are perfectly usable. The only things that are special about wearables is the power — the fact that you have something that’s always on and that you have communications and sensors and a display that come into play here. The 12-year battery in the pacemaker will no longer work for 12 years, it will work for one day.”

Security
One big question mark across this market is security. If wearables do indeed capture personal data, how will it be shared and with whom? And how will it be safeguarded when security requires more energy in an already energy-constrained system?

This becomes particularly important as more devices are rushed to market. Andes’ Lin said innovation and new products are critical to get this market rolling, “but of greater importance is security, battery life, functionality, and cost—the elements that will enable IoT devices to proliferate.” He said that will require big investments by semiconductor companies developing smart sensors devices, medical devices, smart appliances, touch panels, wireless charging and power management ICs, all of which need to be hack-resistant. And all of that has to be done within a very tight power budget with enough performance to satisfy end users.
“Right now in many of these devices there is no security,” said NXP’s Cooper. “When you add security, you increase the power budget. The only way to deal with this effectively in the future is to move security from software to hardware because hardware can do it with less energy.”
That requires some serious architectural planning, however, and so far the wearable market is a competitive frenzy of who’s going to be first to market. A well thought-out security model will likely come second.