I was able to spend a third day (Wednesday) at the Embedded Systems Conference before more pressing work caught up with me. It was a very busy show with roughly 175 presentations and 230 exhibitors. Here are the primary impressions I took away:
Energy harvesting, the ability to extract energy from the ambient environment, is becoming much more of an enabling technology. This is not so much because energy harvesting techniques have improved as it is because low power electronics have progressed to the point where very small amounts of energy are now useful. Micro-controllers and RF modules can now run on micro-amps of current and extremely low duty cycles. Applications demonstrated ranged from wireless sensor meshes to wireless light switches.
Software has become the dominant area of embedded systems development. This may not be news to you but, I have had my head down in the hardware trenches for a long time. Now I find myself becoming increasingly involved with software development – though not necessarily for embedded systems. In particular, I am now learning object oriented design. So I found the vast array of software development and management tools interesting. For example I met with John Greenland of LDRA Software Technology. His company offers a full start to finish development environment to take an embedded project from requirements to code on the board. If this sort of thing had been around when I originally studied programming, perhaps I would have stuck with it.
Numonyx, a new player in the flash memory market, celebrated its first birthday March 31. They are a joint spin-off of the flash divisions of Intel and STMicroelectronics and had a strong presence at the conference. This company seems to be one of the few semiconductor companies that is still financially strong despite the faltering economy.
David Brannam, a technical marketing manager at Numonyx, gave an excellent talk on designing memories in embedded designs with enough flexibility that you can change memories, and perhaps memory vendors, at a later date if need be. He suggested selecting multiple sources up front. Besides the obvious considerations such as component footprint and pin-out, he warned to take a close look at the memory command sets. There can be subtle differences that are easy to miss if you just gloss over the datasheets. This seems like a perfect place to employ simulation models from each vendor to be sure your controller can handle the second sources.
Which takes me to my final observation. There were a number of semiconductor companies exhibiting at the conference. I tried to stop by each one and talk to them about models. At past events, this has usually been an exercise in frustration as most of the people I spoke with had a hard time understanding even the concept of a functional simulation model. This time things were different. Several of the people manning the booths knew exactly what I was talking about and volunteered that they had customers asking for models. So now I can spend the next couple of weeks trying to reach those people and see what FMF can do to help them fulfill those needs.