The case for multifunction DSP in wearable devices -

The case for multifunction DSP in wearable devices


MADISON, Wis. — Wearable devices are suffering from a fatal flaw: batteries that die too fast. Developers blame the problem on the dearth of tailor-made wearable SoCs that could meet their requirements.

With that in mind, if they were to develop their own wearable SoCs, what should be their priority?

On a system level, system designers first and foremost need to rethink displays. Will Strauss, president of Forward Concepts, said the biggest power drain in a cellphone is the screen. Switching “to black and white to save power when color is not needed” is one idea. Other power wasters include “keeping WiFi on when you're not near a hotspot.”

On a silicon level, Linley Gwennap, principal analyst at Linley Group Inc., calls “higher power in the application processor” the main culprit in power drain. “Processors with four or eight cores, particularly big cores, are much more power hungry than the single- and dual-core processors from a couple of years ago.” Choosing a low-power processor (CPU and/or DSP) combined with wireless communications (e.g., Bluetooth, WiFi, cellular) should be a high priority in wearable SoCs, he says.

Eran Briman, vice president of marketing at Ceva, told EE Times, “Today's IoT and wearable devices are underserved by inappropriate solutions.” For the moment, we should forget about using a multicore app processor to do everything, he said. Adopting a multifunctional DSP core in combination with a small MCU/CPU is one way to design a new wearable SoC.

Last week at the Linley Mobile Conference, Ceva pitched its TeakLite-4 DSP v2 architecture with 50 new instruction sets. It now can handle Bluetooth connectivity (Classic or Low Energy). That connectivity comes on top of the same single DSP core running other functions, including audio and voice software packages, always-on user interface (UI) functions, and a full suite of sensor-fusion capabilities, according to Briman.

The basic principle behind the DSP+MCU idea is to let the DSP handle the always-listening chores. The DSP triggers an MCU to wake up when necessary, thus saving the system's overall power.

In essence, Gwennap said, Ceva's DSP platform supports “always-on functions and low-power connectivity. Those are the two capabilities that are critical in wearable devices.”

Let DSP listen all the time
Let's take a closer look at the tasks that consume most of the power during screen-off. According to Briman, these include Bluetooth Low Energy, sensor fusion, always-listening (voice) and always-watching (face, gesture) functions. If a single DSP core — running at extremely low power — can do all that, it “dramatically lowers the cost, complexity, and power consumption of chip designs for IoT, wearables, and wireless audio devices,” according to Ceva.

Moto X is the first mobile device to delegate always-on to a DSP, Briman said, instead of keeping an application processor running all the time. Moto X used Texas Instruments' C55 DSP to run low-power voice recognition. The DSP is there to wake up the application processor when it hears a voice.

Power challenge for wearables. Click here for a larger image.(Source: Ceva)

Power challenge for wearables. Click here for a larger image.
(Source: Ceva)

But seriously, how big an issue is this always-on feature in terms of draining power in mobile devices?

To read more of this external content, go to “Is always-on a power hog?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.