issue: September 2003 APPLIANCE Magazine Electronic Controls and Embedded Systems MCUs and DSPs: Worlds Apart?	Share  Printable format  Email this Article  Search
by Lisa Bonnema, Managing Editor Traditionally, MCUs and DSPs have been perceived as appropriate control options for entirely different types of applications.

Therefore, MCUs are designed to respond rapidly to I/O interrupts and to read in the code and data associated with several programs stored throughout a wide memory address range. In addition, MCUs are known to feature integrated peripherals and Flash memory that help keep embedded system costs down and enable reprogramming for faster system development.

The drawback in MCUs is that they are not designed for high precision and the parallel execution of instructions, so that they have limited utility in systems that need a high level of performance for control in real time.

DSPs, on the other hand, are specifically designed to deliver high performance in real-time applications. The parallelism inherent in DSP architectures permits several multiply-accelerate (MAC) operations to be performed simultaneously in computationally intensive tasks. DSP C/C++ compilers are also designed to be extremely efficient, so that the hardware architecture is supported to the fullest in software. With their performance overhead, DSPs are able to extend embedded control capability while reducing components and saving system costs.

For instance, in motor control, DSPs can perform rotor speed and positioning tasks in software without the look-up tables and additional feedback sensors required by an MCU. In power supplies, DSPs can support new requirements, such as power factor correction, without the addition of complex analog control hardware.

The high-performance architectures of DSPs, however, have tended to make them specialty engines for single-task systems. Earlier generations of DSPs were slower in handling interrupts and had a more limited address range than MCUs, as well as offering fewer integration options. In addition, developers have perceived programming DSP assembly code as a complicated process, compared to the familiar high-level MCU programming flow. As a result, system developers with an investment in MCUs have been reluctant to make the transition to DSPs for control applications.

The Best of Both Worlds

Texas Instruments (Houston, TX, U.S.) says it has eliminated these drawbacks by combining the high performance of DSPs with the general-purpose control features of MCUs.

The company's C2000 DSP controllers are reportedly capable of executing up to 150 million instructions per second (MIPS), supplying the high computational speed needed for real-time control.

The C2000 DSP controllers also feature the low interrupt latency and a wider address range typical of MCUs, which is said to make them better adapted for use in systems with multiple I/Os and control tasks. Integrated analog peripheral functions such as analog-to-digital converters (ADCs) and pulse-width-modulation (PWM) outputs save system space and cost, and on-chip Flash memory permits easy reprogramming for faster code development. While the Flash memory supports up to 120 MIPS with no wait states, additional zero-wait-state SRAM is also available on chip for time-critical code.

This information is provided by Todd Solak, Texas Instruments, Inc.