Motor Control Myths and Reality - DSPs Make a Strong Run at MCUs

eeProductCenter (05/24/2004 3:00 PM ET)

Motor control designers have faced a roadblock until recently in applications as diverse as white-goods and servo drives, having to choose between controller performance, or a high price. Most motor control applications are by nature low-cost. Attractively priced products, a must for market acceptance, has meant choosing the cheapest controller that would do the job and not much else. Smart DSP-based controllers are changing this, and the logjam has been broken, and modern DSP-based controllers offer a huge performance improvement at a very modest price.

Simple control algorithms, such as constant volts-hertz and six step commutation, lack the performance needed for efficiency and optimal machine sizing. Intelligent DSP-based controllers change this in two ways.

Firstly, they add numerical ability. This enables designers to implement higher performance control algorithms such as field oriented control. Secondly, the more numerically intensive algorithms also enable designers to use a more efficient motor. For example, replacing an AC Induction motor with a permanent magnet motor to further improve efficiency and dynamic performance.

Of course, this performance helps designers cost optimize their system, by eliminating mechanical components, right-sizing motors, and integrating more functions into the controller. The ability to integrate functions such as speed, position, torque profile generation, power factor correction enables designers to do more, and with less cost.

The "problem" with advanced techniques such as compute-intensive vector control, is that multiply-and-accumulate (MACs) operations dominate these algorithms. Standard 8, 16 or 32-bit microcontrollers can not handle them, due to the lack of a proper bus architecture to be math efficient. Ultimately, that means a fundamental design shift from microcontrollers and not just to DSPs, but to 32-bit DSP-based controllers.

Not surprisingly there are some popular myths related to the adoption of DSP controllers, for example:

1. DSP controllers do not have motor control peripherals

2. Code density issues make 32-bit DSP-based systems unattractive

3. DSPs might be a good fit for the control algorithms, but they do not handle other control tasks very well

4.Software for DSPs is hard, designers must do without real-time operating systems and good tools support

Peripheral integration

The first thought likely to enter a motor control engineer's mind when heavy-duty digital signal processing is mentioned is that DSP controllers are good at number crunching " but how about commonly integrated peripheral functions?

Optimized for motor control applications, today's motor-control DSPs integrate on chip such functions as pulse-width modulators (PWMs), encoder interfaces, communication ports, and analog-to-digital converters (ADCs). They also pack substantial flash memory and RAM to eliminate external memory devices.

Code density

The truth is that traditional 32-bit microcontroller architectures have had an inherent disadvantage in code density, a critical issue in low-cost applications with limited memory. Modern 32-bit DSP architectures deploy a carefully chosen mix of 16- and 32-bit instructions providing best in class code density, 32-bit instructions are used only when needed.

Control oriented architecture Best in class 32-bit DSP controllers have features such as atomic read-modify-write instructions simplify programming, and low interrupt latencies less than 100 nanoseconds create ultra-responsive CPUs.

Software Support

DSP compliers have advanced to where writing assembly code is rarely required. Further, algorithm development has become a lot easier as well. Texas Instruments, for example, offers a large library of common algorithms that are ready to run on its processors, and BIOS, a real time operating system specially optimized for DSPs.

When a custom algorithm is needed, help is available as well. To create their algorithms, engineers can leverage the IQMath tool, which simplifies the development of mathematical functions in the fixed point domain.

Texas Instruments is putting Moore's law and creativity to work, making the motor control designers' task easier.

As part of the Digital Control Systems group at Texas Instruments, Kedar Godbole mainly focuses on digital signal processor applications in digital control, and works on content within the Digital Motor Control software program and enhanced software reusability deployability. He also holds a bachelor's of engineering degree from the University of Pune in India.

He can be reached be at controlanswers@list.ti.com