SMALL FORM FACTOR FORUM
COLIN MCCRACKEN & PAUL ROSENFELD
Page 1 of 1
With another Embedded World show behind us and ESC (“Design West” now) upon us, it’s clear that the momentum behind small ARM-based modules is not going to slow any time soon. Quite the opposite; it is becoming the new frontier in the off-the-shelf board space. Like the “Wild West,” however, the shootouts and standards turf wars are not going to settle into law and order anytime soon.
What is driving this trend? Before answering, it must be noted that RISC modules and SBCs targeting embedded computing have been around for more than 10 years from small to mid-size vendors—ranging from DIMM-PCs to PC/104-style CPUs to non-bus-expandable SBCs to proprietary modules. Most of the shipments have been custom StrongARM and XScale boards derived from an off-the-shelf baseline. The sudden emergence of several large vendors jumping on the RISC bandwagon signals that it’s time for the broader market to start paying attention.
The first of three trend drivers is the Tablet PC, a consumer product that provides efficient technology building blocks for embedded applications. Portable embedded computers continue to fan out into the market. By “portable” we mean systems that operate off a battery for hours at a time, rather than always being plugged into line power. Lightweight devices have been built around custom ARM designs with small batteries. Heavy devices such as forklift-mount or ruggedized soldier computers have been both standards-based and custom x86 designs for years. The explosion in consumer electronics has contributed to an enormous ARM system-on-chip (SoC) ecosystem for differentiated low-power single-chip solutions with sufficient graphics for decent-size LCDs. With a multitude of RISC-based embedded board-level product introductions each year, low-power RISC processors are furthering their penetration into the traditional x86 board space. But it’s far too soon to call this fight.
The second trend driver is the establishment of an application platform. Best positioned as the next de facto standard, call it “Embedded 2.0,” is Android running on the ARM processor architecture. Key requirements include ease of obtaining the OS, tools, APIs, and development and debug tools. In the small form factor, long-lifecycle computing market, “Embedded 1.0” was x86 running DOS, and was subsequently updated to version 1.1 running Windows and 1.2 running Linux. (Of course, real-time and network-protocol applications utilized different OSs and processors.) Late to the table but still poised to capture some of the embedded 2.0 RISC board market share is the heavyweight Windows 8. At this point it’s still to be seen how the modest performance of low-power processors will run the OS. Windows comes with its own CPU utilization, power consumption and even thermal design power (TDP) issues.
The third trend driver comprises the board-level manufacturers who simply package processor manufacturers’ silicon reference designs, firmware and software in an appropriate way for widespread adaptation to the unique requirements of hundreds of embedded applications. Since there isn’t a one-size-fits-all design, there must be enough expansion interfaces or “hooks” to add application-specific I/O. Therein lies the rub. Computer-on-module (COM) approaches have carved out a huge chunk of the SFF board space. For their RISC implementations, COM suppliers simply define what is easy and convenient for them to design and manufacture in high volume with the adaptation to a specific application as an exercise for the user. One must look past the board size and shape to understand not only how to attach I/O circuits to the minimal familiar “buses” provided, but also to realize that interoperability and future upgradeability is even more doomed than in the x86 COM space. For no two ARM SoCs are created equal. COM vendors don’t like to add cost or size in order to create a least common denominator of features. So the result is more reserved pins… “players to be named later,” which can be spoked out on a carrier board (untested) that will need to be spun again later in any event. And as usual, the sales strategy will be selling futures (multi-sourcing and upgradeability) under the guise of “future-proof” modules.
While these new modules are indeed welcome as the next trial balloons to keep innovation healthy in the embedded market, don’t think that “Embedded 2.0” won’t be littered with dot releases and upgrades in the future. As with Embedded 1.x, the software is much more mature than the hardware due to the primary (consumer) market that is driving the software and silicon. Just plan for a litany of carrier board spins in your engineering budgets for year+1, year+2 and so on.