MCU landscape undergoing transformation

While volume shipments of 8-bit microcontrollers (MCUs) are currently 4x those of their 32-bit counterparts, their growth rate is the complete reverse by a huge magnitude, with 32-bit microcontroller shipments growing about 50x faster than 8-bit, according to industry estimates.

There are a number of solid business and technology trends driving this migration. Even applications that appear to be “low-end” now demand more sophisticated processing, and can benefit from the additional performance delivered by the 32-bit processor architecture. Take the classic white goods product – the humble washing machine.

Government regulations and consumer demand for better energy and water utilization from a wash cycle now require a more efficient way of controlling the motor. The design of efficient fractional horsepower motor drives, which employ techniques such as pulse width modulation, presents some interesting technical challenges that require advanced algorithms to be implemented. These challenges are well beyond the capabilities of the 8-bit MCU, and often require a full 32-bit MCU.

Future application trends
Looking into the future, a growing trend to simplify user interaction with appliances will further increase processing demands. Once again the washing machine provides a graphic example. Faced with a confusing selection of 30 or more wash cycle options on the average washing machine, in practice most consumers limit their choice to just one or two familiar programs. The way forward is for the appliance itself to choose the best program. By sensing the weight of the load, temperature of the water, amount of dirt and so on, the intelligent machine will deliver the best results with the minimum use of energy, water and user intervention. This level of intelligence requires even more intensive computation from the microcontroller.

Accompanying this trend toward more advanced processing, pervasive communication is placing further demands on microcontroller performance. Protocols such as integrated Ethernet, USB, ZigBee and Bluetooth are becoming routine additions to many products. What’s more, these trends span many application domains. From refrigeration to home automation, and automotive to industrial domains, processing demands are growing.

32-bit production cost benefits
While the applications themselves are becoming more technically challenging to design and develop, it’s not all bad news for the manufacturers and designers. Because a single 32-bit processor can perform the tasks of several 8-bit processors, many manufacturers are taking advantage of the opportunity to aggregate several 8-bit devices into one 32-bit processor. This brings two major benefits: a simplified development process and reduced production costs.

Cost has been a barrier to the adoption of 32-bit processors in the past. There are two main cost components: the development (or NRE) costs and the unit volume costs. In 2009, the average selling price of an 8-bit microcontroller across various market segments varied somewhere between 54 cents and $2, with an average selling price of around $0.86. By way of comparison, devices based on the ARM 32-bit Cortex-M0 processor can now be purchased for as little as $0.65 in volume.

As discrete devices, most MCUs are pad-limited – in other words the opportunity to shrink the die is constrained by the need to fit the bonding pads around the perimeter of the chip. As minimizing the size of the die is not a priority, MCU manufacturers will typically target previous-generation processes such as 0.25μm and 0.18μm, where the process development costs have already been amortized and the cost of mask sets is considerably reduced compared with the latest state of the art 90nm and 65nm processes. Today, unit volume cost is therefore much less of a barrier to designing with 32-bit processors.

Controlling development costs
With regard to development costs, design teams are experiencing a raft of benefits from the move to the 32-bit architecture. There are few standards in the 8-bit MCU world, which has led to a proliferation of different development platforms and tools. It’s not unusual for a single company to be burdened with maintaining an array of different tool chains depending on which 8- and 16-bit devices they support.

This severely constrains the opportunity for code re-use, and significantly increases overheads for tool chain maintenance and their engineer’s training requirements. ARM Cortex-M series-based MCUs provide high levels of software portability. Although there are multiple MCU vendors and multiple MCU tools vendors, software can be easily ported via the Cortex Microcontroller Software Interface Standard (CMSIS). Software developed with CMSIS can be easily ported between different Cortex-M series-based MCUs, enabling OS or middleware products to support multiple device and tools vendor suites, and helping safeguard investment in software development by enabling better software reusability.

Additionally, while programming 8-bit devices often still relies on writing the application in assembly code, 32-bit platforms are much more amenable to taking a high-level approach to code development. This could be in C, UML or from a high-level tool such as MathWork’s Simulink. This means that the platform approach can often be taken, where an application can be developed once and re-targeted multiple times.

ARM defines 32-bit
With application processing demands up and volume production costs down, there is an unmistakable migration to 32-bit processor architectures, even for product applications that have previously been tagged as “low-end”. Increased functionality and integration is driving the move from traditional 8-bit MCUs toward the greater performance and flexibility provided by cost-effective 32-bit designs.

Standardizing on a processor architecture and peripherals will help to manage the complexity, connectivity and compatibility of systems while decreasing the resource requirements during the development cycle.

The ARM Cortex-M family of processors for MCUs comprises of the Cortex-M0, Cortex-M3 processors and the recently announced Cortex-M4 processor. These processors feature leading MCU technologies including a highly efficient CPU core, integrated interrupt controller, memory protection unit, low power sleep modes and low cost debug/trace. With Cortex-M4 processor-based devices expected for release later in 2010, easy-to-use signal processing will now be made available to the market in a MCU form factor and cost structure.

ARM’s 16-/32-bit architecture is forward-compatible and protects investment in hardware and software. Most of the world’s leading MCU suppliers now provide ARM processor-based MCU products, offering a broad choice of manufacturing sources. The ARM Connected Community also offers a broad choice of application-specific intellectual property, operating systems and development environments from over 700 partners, making it the best supported microprocessor architecture available.

ARM

Author Information
Shyam Sadasivan is a Product Manager for ARM. He has been with ARM since 2002, contributing in many areas including designing I/O interfaces, developing new design methodologies and product marketing for leading processors. Shyam is now involved in extending the reach of Cortex-M processor technologies into new markets and applications. Shyam holds degrees in Physics and Electronics from BITS, Pilani, and also the General Management Certificate from the University of Cambridge.