On-chip flash memory makes motor control chips easier to program (in development) and reprogram once in the field. Based on an industry standard 16-bit DSP core, the new chips offer plenty of connectivity and simultaneous sampling (eight channels) at 10, 12, or 14 bits. Processor speeds range from 10 to 150 million instructions per second (MIPs), providing ample bandwidth for motor control peripherals, namely two 16-bit, three-phase PWM generators (with auxiliary PWM), dual encoder input, and watchdog and event timers.

Magnetic energy is the invisible force that makes motors go, but software decides where they go and how they get there. Software, in the form of digital current loops, fieldoriented control functions, and estimation algorithms, is as essential to motors these days as copper, iron, and steel.

Whether the goal is positioning or variable-speed control, the software engine of choice for motors is a special type of chip called a digital signal processor. Although DSPs have been around less than 15 years, these speedy devices have nailed down a preeminent spot in motor control, handily beating out standard microprocessors and microcontrollers.

DSPs have the edge in motor control not because they’re manufactured with more care or made of a better grade of silicon; they just happen to be “wired” differently. While general-purpose processors do a myriad of things reasonably well, DSPs are geared specifically for math, particularly vector and matrix manipulations. In a single 20-nsec clock cycle, a DSP can multiply two numbers and add the result to a running total. This simple sequence – the core of most motor control algorithms – could take dozens of clock cycles on a general-purpose microprocessor.

DSPs have architectural advantages as well, facilitating data transfer both internally and externally. When integrated with motor-based I/O circuits, such chips not only perform multiply-and-accumulate operations in a single clock cycle, they can also actuate a motor control signal – say a pulse-width modulation (PWM) output – in the same period of time. And that’s not all.

DSPs are usually fast enough to execute more than the average motor control algorithm between sampling periods. This opens up all sorts of possibilities from model-based sensorless and anticipatory control to diagnostics, load modeling, and adaptive control.

Why DSP?

Control algorithms commonly used in variable-speed drives consist largely of transfer functions and state-space equations readily expressed as a set of vector calculations. These operations execute efficiently on digital signal processors, architected for digital filters and vector mathematics

A standard fixed-point DSP core has three independent computational units – an arithmetic logic unit (ALU), a multiply- and-accumulate unit (MAC), and a shifter. Each connects to two data memory buses and a results bus, which carry information (input) from data memory, program memory, and the output registers of the units themselves. By integrating this architecture with motor control I/O, the control algorithm results can be written directly to a peripheral, such as a PWM output, in a single DSP cycle – all with a minimum of coding.

In review, the most important feature distinguishing a DSP from a standard microcontroller is the single-cycle multiply and accumulate (MAC) function. A DSP with a 20-MHz clock, for example, can perform a 16 X 16-bit multiply and a 32- bit accumulation in only 50 nsec. Such a manipulation, when combined with program and data memory reads, forms the core of a digital filter algorithm.

The basic signal functions comprising an ac motor control system include a processor (DSP core), a PWM generator, and an a/d converter. Timers and I/O ports are also required.

Digital signal processors are also well suited for real-time control. For starters, they’re equipped to handle a number of interrupts from external signal sources. They can also perform context switches in a single instruction cycle with the help of “shadow” registers – independent arithmetic registers – on each arithmetic unit. This eliminates the need to save register contents when servicing an interrupt or calling a subroutine, conserving both program memory code space and execution time.

What goes in must come out

Controlling a variable-speed drive takes more than digital signal processing circuits. It also requires special I/O circuits that interface to power and motion components. At the very least, these peripheral functions should include a pulsewidth modulation generator and an analog- to-digital conversion system. Other peripherals required for real-time embedded control include parallel I/O blocks, serial communication interfaces, and watchdog and event timers.

A sensorless variable-speed control system “sees” its way around with the help of a motor model that estimates the required voltage based on set speed and measured winding current. All functions, including input and output peripherals, have been integrated on a single-ship DSP.

PWM outputs provide the greatest flexibility in motor control because most low to medium-power ac motor control applications employ three-phase IGBT or MOSFET inverters. The power device switching signals are usually fixed-frequency PWM timing signals, ranging from a few kHz to tens of kHz.

Continue on Page 2