Semiconductor IP News and Trends Blog
IP That Senses and Cares
Engineers may be nervous about user experience (UX) based hardware design trends, but the future lies in creating devices that sense and care about their owners.
This week, Semicon West has taken over all the halls at the San Francisco Moscone Center. The show focuses on microelectronics design and manufacturing, from EDA (mostly DFM/DFY) an device fabrication (wafer processing), to final manufacturing (assembly, packaging, and test).
IMEC, a large nano-electronics research organization headquartered in Leuven, Belgium, is also at Semicon West. The R&D organization partners with many major semiconductor companies, both IDM and fabless, in both microelectronic design and manufacturing.
At this year’s Semicon show, IMEC tried something different. They held one of their technology forums. This one was on smart phone design. After introductions, the forum began with a talk by Intel’s CTO, Justin Rattner.
Rattner presented many interesting factoids during his presentation. My favorite was that Twitter handles over146M tweets per day that require more than 1.4 TB of storage per day. That’s a lot of bird seed!
But the main thrust of his talk was two fold. First, smart phones are becoming as important as our vital senses. Second, user experience now drives innovation. Each of these observations has rather staggering implications.
If smart phones are becoming more sensor-centric, then they’ll need more sensing capabilities. For the world of semiconductor IP, this translates into MEMS cores for functions such as acceleration, pressure and temperature sensors – just to name a few. Naturally, the typical analog to digital signal conditioning IP will also be needed, such as amplifiers, filters, converters and the like.
The growth of sensor technology is nothing new and has been predicted for a number of years. What is new – at least to me – is the recognition that future chip designs and innovations may be focused in the area of enhancing user experience. This approach is sometimes called experience driven design.
“User experience design makes engineers nervous, since it relies on one’s perspective for what makes for a good experience. But this is now becoming a formal, qualitative experience,” explained Rattner.
This approach involves more than just getting the user interface correct. It requires a great deal of user input and feedback throughout the product development process. The key to user experience- (UX) based hardware design seems to be determining the basic level of the needed experience.
Could such user experiences be captured as semiconductor IP? Why not, especially if such experiences could be captured in algorithms? Perhaps future chip designers will need courses and training in the social sciences. Now that would be ironic.
Rattner continued his talk by noting that the next major trend beyond user driven experiences is context-aware mobile computing. He predicted that such computing would be the focus of design for the next 3 to 5 years. Context-aware mobile computing will involve both extending our traditional “hard” senses to include such things as skin temperature, heart rate, etc. Such devices will also make more use of audio signals in the environment to decide if you are in the same setting as someone else, e.g., the same side of a wall. Again, such systems will need a great deal of signal processing IP.
Hard sensing will be complemented by soft sensing, i.e., information from a user’s personal calendar, social network, browsing trends, etc. Data from both hard and soft sensing will be needed to infer the users health, state of mind and actual location. The hope is that these devices will make for a high quality experience through the use of activity fusion algorithms. (Sounds like an energy drink to me.)
Will devices that can infer so much about the user actually start caring about their owners?
What are the implications for the growth of hard and soft sensors, user experienced based design and context aware computing? Rattner observed that R&D opportunities abound:
- Low-power integrated sensors
- Always-on sensor subsystems – These sensors may not, in fact, be always on, but would be sampling data at a rate for which they appear to be always on.
- Inference algorithm acceleration – As these algorithms grow in sophistication, they may need to be accelerated on separate hardware co-processors. Intel would like that.
- Client-cloud workload partitioning
- New apps and services.
These opportunities served as a great introduction to the next set of presentations by IMEC, which I’ll cover in my next blog.