Remember your old “dumb” cellphone? The one that you could charge once or twice a week, depending on how much time you spent talking on calls? Now consider your smartphone; most people have to recharge theirs every day (or maybe even more often than that). Designers are caught in a dilemma; they can get longer battery life by including a larger, heavier battery, or they can keep phones small and light and needing to be charged more often. Part of this problem is that the availability of many functions are “all or nothing.” Either the display is on or it’s off. The phone is ready to respond to a button press or an incoming call, but in general, either the central processor is on and working or it’s in a sleep mode waiting for something to wake it up.
If wearable devices are going to be practical in the Health Tech applications, they are going to have to be able to run longer between charges. And a new company, Ineda, has a solution that its founders think may be able to reduce wearable device power consumption by as much as 10 times. The secret behind their Dhanush wearable processing unit (WPU) is to build a system on a chip (SoC) that has multiple processing cores. However, instead of using many cores of the same size and functions, they rely on the concept of “hierarchical computing” to include different sized cores in the same system.
This approach means that there can be a tiny, power-efficient processing core that is active all the time. It would be limited to simple tasks, such as recognizing that someone is speaking, and unlock the system if a single keyword is recognized. This then wakes up a low-power core that is able to recognize and process a limited number of simple spoken commands. If a command requires full natural language processing, then the system would wake up the full-featured application processor.
Ineda initially has designed four different versions of the Dhanush processors, with increasing processing capabilities. The most capable version, the Dhanush Advanced, supports mobile operating systems, such as Android. The company is providing samples of some of the chips this year along with a software development kit (SDK), with mass production planned for next year.
