What's the state of microcontrollers? They've become more prevalent across the board. Low-end chips continue to shrink and consume less power. Developers gaining confidence in desktop PC programming are turning to 32-bit microcontrollers as their platform of choice. Likewise, high-level-language compilers dominate microcontroller development—even on smaller architectures—primarily due to the increase in flash-memory capacity.
BLURRING THE LINES
Microcontrollers tend to be grouped by their bus width, but the lines between systems are becoming extremely fuzzy. New architectures enter the mix, while old ones get extended and refined.
Texas Instruments' MSP430X architecture expands the 16-bit register set to 20 bits, providing access to up to 1 Mbyte of flash memory. This helps meet growing demands for program space while retaining compatibility.
Likewise, low-end DSPs are finding their way into MCU applications (see "Digital Signal Controllers Migrate To General-Purpose Processors" at www.elecdesign.com, Drill Deeper 11701). In many cases, a single chip of this sort can simultaneously handle general applications and signal-processing chores.
The overlap between architecture coverage is extending as well. Even low-power 32-bit processors now challenge high-end 8-bit processors for some tasks. This trend will persist over the next few years as each group of processors continues to grow, taking on the features of its neighbors.
Automotive applications still take the top spot, followed by consumer and industrial use. There is growth in all areas, with more change across sections (Fig. 1). The 8-bit products are holding their own, but dropping prices have sent volumes soaring (Fig. 2). Only 4-bit solutions show a decreasing trend. The 16- and 32-bit solutions still chase each other and are expected to exceed 8-bit revenue in a few years.
Single-chip solutions will be key to the success of most MCU lines. On-chip clocks eliminate an external crystal, and improved analog support does away with other external components. An offshoot of the single-chip trend will be its implementation as a logic or black-box design block.
Though it's still possible to find 4-bit microcontrollers, their 8-bit brethren effectively stamped them out. The 8-bit MCU is the low end-solution, and that will likely continue despite pressure from 16-bit solutions.
Look for 8-bit MCUs with even lower power requirements and lower voltages. Long-life, battery-operated applications will increase because of these new alternatives—especially those that can use a single cell or alternative power generation.
Debugging remains the big problem with small, low-pin-count (six to eight pins) chips. Yet it does help to be compatible with higher-pin-count siblings. No solution exists for this problem, though, except for small packages with more pins.
Ever-expanding 8-bit architectures are now even bumping into their 32-bit brethren. Rabbit Semiconductor's 4000 incorporates memory protection, 32-bit registers, hardware encryption, and even Ethernet interfaces. Don't expect Rabbit to be alone for long.
Communication will be crucial this year. Incorporating technologies like CAN (controller-area network) in lower-end 8-bit solutions will lead to more distributed solutions.
As for ZigBee, its holy grail is to integrate ZigBee transceivers with 8- and 16-bit processors. They reduce power and footprint. Also, they simplify programming because software can be matched to the hardware, taking advantage of any available hardware acceleration. Look for more solutions like Chipcon's CC2430.
Volume in this class continues to grow. Falling prices keep the 8-bit arena from being the processor-class money winner, but it's a good place to make a living.
Some industry watchers predicted the end of 16-bit microcontrollers due to the increased availability of 32-bit solutions and pressure from more powerful 8-bit products. The truth is far from the prediction, with thriving 16-bit families applying pressure in the other direction.
Lower power and small form factors make an 8-bit processor an attractive option. But products like Texas Instruments' 16-bit MSP430 family are moving into these arenas with small packaging and power consumption that rival most 8-bit solutions. The MSP430 isn't alone, so expect all 16-bit vendors to deliver smaller devices that keep 8-bit alternatives on their toes.
Analog support is one place where 16-bit MCUs shine, and it's one of the areas of greatest enhancement this year. An 8-bit digital-to-analog converter or analog-to-digital converter is useful, but 12- and 16-bit units are better.
The 16-bit MCUs have the register width to manipulate data without the register juggling 8-bit processors. The same is true for pulse-width-modulation (PWM) support often used with motor-control and power-control applications where 16-bit MCUs have been very strong.
The 16-bit processors also benefit greatly from the increase in on-chip flash and SRAM, enabling powerful single-chip solutions. Watch for some interesting architecture changes like Texas Instruments' MSP430X, which extends the original 16-bit registers to 20 bits. It will have some interesting implications on a 16-bit processor.
ARM-based microcontrollers from a range of vendors will continue to make a big splash, but they aren't the only option. Freescale's ColdFire is benefiting from continuous enhancements, and other alternatives are out there. So, the competition remains fierce. The 32-bit architecture leads in monetary growth in the MCU arena.
The plethora of ARM MCUs is leading to some confusion on the developer side because of the variance between support for things like bit-banging and interrupts. This inconsistency between vendors will likely persist, especially as some vendors see their implementation as a feature.
More versions of the 32-bit MCUs exist versus 8- or 16-bit chips. That's because they often come with off-chip memory versions, making it possible to use very large memories. Still, single-chip 32-bit MCUs are gaining in popularity on their smaller siblings due to easier programming and an increase in the number and power of 32-bit MCU on-chip peripherals.
The rise of commercial real-time operating systems (RTOSs) and open-source operating systems like uCLinux, now part of the main Linux development tree, continues to push memory capacities and chip feature sets to greater heights. This will drive developers to prune down applications to fit in smaller, less-costly 32-bit MCUs, as well as take advantage of the higher-end, memory-rich products. Megabyte flash memories are just the start, not the limit.
Java hardware acceleration and even native code Java execution works best on a 32-bit platform. Arm's Jazelle DBX (for Direct Bytecode eXecution) is found in a number of microcontrollers, but it seems to be an underused feature with the exception of a few key application areas.
This year may be different with the release of ARM's Cortex-A8, which uses the Jazelle RCT (Runtime Compilation Target) architecture instead of DBX. It suits Java, but RCT works equally well for other programming languages. It may well be that 2006 is the year of hardware introduction and experimentation.
Production will continue in hardware acceleration for wired networks such as Ethernet and security. Hardware encryption and on-chip key storage will provide the basis for security. Also, rights-management protocols are now moving to the fore. Some 32-bit MCUs will gain DSP-style features to handle applications like Voice over Internet Protocol (VoIP).
If you're looking for performance, there's a growing crop of 64-bit microcontrollers to choose from, including the multicore area. That's where you'll find PowerPC and MIPS cores coupled with high-performance interfaces (e.g., PCI Express, HyperTransport, Serial RapidIO). Many chips even have multiple Gigabit Ethernet interfaces on-chip.
Off-chip memory is the norm. Therefore, single-chip solutions are rare. But not all applications demand a multicore server farm. Storage systems require more performance, and 64-bit MCUs will sit between the storage devices and the main processors.
Look for more of these high-performance chips to show up to fulfill expanding storage and network processing systems. High-speed serial fabrics will have an impact on this class, providing the intelligence between powerful clusters and peripherals. High-end network edge devices also benefit from a 64-bit MCU's power. Look for dual high-speed Ethernet links and encryption support.
Finally, 64-bit solutions will compete with 32-bit MCUs in high-end multimedia applications. Hardware acceleration will be included on both sides, but 64-bit processors can still move more data than a 32-bit processor—especially if it requires floating point.