Network dynamics, SoC impact MCU design

Turning MPUs into MCUs, however, is more than a matter of higher integration, said Kevin Klein, standard-products-marketing manager for the 32-bit Embedded Controller Division of Motorola's Semiconductor Products Sector (Austin, Texas). It requires a rethinking of the arc hitectures-either removing or masking the effects of features that were useful in traditional computing but that get in the way of full functionality in a controller environment.

Klein provides details on how system-on-chip-based network controllers are blurring the MPU/MCU boundaries in an exclusive online article in this week's In Focus section, which explores how the networked environment affects controller development.

But where Klein views such changes as a "convergence" of microprocessors and microcontrollers, contributor David Fotland, chief technology officer of Ubicom Inc. (Mountain View, Calif.), believes the dynamic is one of replacement. What is emerging is not converged 32-bit MPU/MCU designs, he said, but totally new architectures that will be specific for the deterministic, real-time control applications that must now operate in a much more open, internetwork-connected environment. He sees opportunities at the Internet edge, where the data throughput rates are modest compared to the network core, usually no more than 1 to 10 Mbits/second. Here, connected microcontrollers must deal with lower levels of the TCP/IP protocol stack, where determinism and real-time are essential.

The approach that Ubicom favors, said Fotland, is what he calls a programmable-software system-on-chip technology. The idea is to keep the basic microcontroller as lean and mean as possible and throw all of the available process technologies at driving up the clock rate rather than the chip density. The result, he said, is sufficient CPU overhead to decode in software virtually any Internet-edge protocol, while performing the other necessary control functions and while handling the incoming packets at wire speed.

Other microcontroller vendors are looking at options that leverage new system-on-chip design methodologies to allow them, with their customers, to develop the solution most appropriate to the application. More recently, there is a trend toward the use of programmable-logic and micro controller combinations, in effect yielding programmable systems-on-chip.

Philips Semiconductors (San Jose), for example, is moving to a highly flexible SoC design methodology. In their exclusive online article, Ata Khan, director of innovation for the microcontroller business at Philips, and Clive Watts, CPU product manager at ARM Ltd. (Cambridge, U.K.), explain that Philips' strategy is built around its new 0.18-micron CMOS process to ensure sufficiently high integration to build a true 32-bit microcontroller around a standard ARM7TDMI-S core. The approach has ample room for all the necessary customer-specific peripheral functions and has sufficient on-chip memory to support more than a nominal portion of the processor's 16 Mbytes of theoretical direct memory access.

Hoping to convince embedded-microcontroller designers that PLD/MCU and PLD/MPU combinations provide e ven more opportunities for reconfigurability, companies such as Altera Corp. and Xilinx Inc. are promoting system-on-programmable-circuit (SoPC) schemes.

Whatever the approach, the combination of technologies is making remote diagnostics, debugging and replacement of software code and hardware functionality a reality. This can be implemented on an incremental, catch-as-catch-can basis or in a more structured approach, writes contributor Stephen Kingsley-Jones, vice president of engineering at Blue Iguana Systems Inc. (Sunnyvale, Calif.)."It will require a sophisticated Web- and Internet- based remote-device management infrastructure to coordinate the care and feeding of literally millions of microcontrollers that in the future will all be network-connected to some degree or other," he says.