Motion ICs make their move

Designers integrating motion-control functions on their system designs have always had to make tough decisions, such as what motor to use and how to gear it. But now, with the emergence of the motion-control-IC market, designers have the choice of either implementing their designs in the traditional manner on a number of pc boards all connected by cables or using a new method that integrates several control functions on ICs. The new method has a steeper learning curve than the traditional method, but it requires that systems use fewer boards and fewer cables to connect those boards; this scenario means that designs will likely be more reliable and ultimately cost less.

The classic bus-based motion-board architecture uses a controller card and separate amplifier modules (Figure 1). Cables connect the motion card to the amplifiers, the amplifiers to the motors, and the motors to the card. Because the popular standards for motion cards have evolved to support PCI, PC/104, CompactPCI, and others, this architecture is still relevant today. But its weakness is the number of cables necessary to interconnect everything. This complexity costs both dollars and reliability, because connectors are important sources of failures, and signals on wires can degrade with distance.

The attendant costs of this approach represent a major challenge in motion control, and solving this problem by using alternative architectures has been a priority for the past 10 years. Generally, the goal is greater integration at the card level to eliminate connectors and cables.

Along these lines, an alternative control architecture integrates the controller function and the amplifier function onto one card (Figure 2a). A corollary version adds host software on the control card using an on-card microprocessor (Figure 2b). Industry experts sometimes refer to this second architecture as a "machine controller," because it provides all control functions necessary to run the machine. The advantage of both of these approaches is that the controller and the amplifier require no cables to interconnect. Another advantage is that the amplifier is much less expensive because you integrate it at the IC level rather than purchase it as a stand-alone unit.

When is it appropriate to integrate the motion controller with the amplifier onto one card? There are many aspects of this question, but the most important factor is the power rating of the amplifier. Combining the motion controller with the amplifier tends to be most viable in lower power systems, such as those that drive NEMA 34-sized or smaller motors.

Another consideration is bus architecture. If other parts of the control system use a standard parallel bus, then it makes sense to locate the motion card on the bus and separate the amplifier function. Although convenient in one respect, this scenario represents a constraint to integrating the controller with the amplifier. By comparison, network and serial buses do not have this limitation, because there are no physical form factors to which to mate the control card. This situation is one of the reasons that so much excitement exists about distributed networks for motion control.

Whether you build the integrated control card for a stand-alone configuration or you locate it on a network, you must keep in mind certain design considerations. In a typical integrated motion controller, the major elements are the motion processor, the signal-conditioning circuitry, and the amplifier (Figure 3). Other sections, such as a network-interface chip and a host microprocessor, may exist, depending on the overall control and communications architecture.

The motion processor is the central IC that performs most of the motion-control functions. These functions include quadrature-signal decoding, trajectory generation, servo-loop compensation (if you use servo motors), PWM (pulse-width modulation), analog functions, and pulse and direction motor-command output generation. Other functions may include commutation, digital I/O, analog I/O, breakpoints, servo trace, and motion-performance monitoring.

Since about 1985, it has been possible to purchase the motion processor off the shelf from a number of vendors. Varying in their degree of sophistication, the number of axes they support, and the motor types with which they work, these handy products provide high-level motion commands and manage all low-level interfaces to motion peripherals.

Motion processors connect to the outside world through a parallel microprocessor-style interface; a serial interface, such as RS-232 or RS-485; or, more recently, networks such as CANbus. Early on, their motion features were less powerful than those that off-the-shelf cards provided. However, in the last five years, that distinction has disappeared with the addition of S-curve profiling, dual-biquad filtering, and onboard trace (see sidebar, "Hardware-trace buffer aids in performance tuning").

The other major option for the motion processor is to purchase a DSP or microprocessor and directly program the motion functions. This approach is cheaper, not only because most DSPs are less expensive than dedicated motion ICs, but also because, if the application is simple enough, it may be possible to eliminate the need for a separate microprocessor that contains the host code (the code that controls the overall machine function). Combining the host code with the motion-processor function has drawbacks. Motion processors generally require high-speed, synchronous attention to motion peripherals, and the host code spends much of its time servicing communication requests or determining the next action. So, combining the two functions does not always make for a reliable and responsive system.

Another factor is motor type. Programming a servo loop is complicated, and few designers attempt this programming themselves. Getting the microprocessor to generate pulses through a digital-I/O port is simpler, so if the application uses a step motor, a home-built motion controller may be a good option.

On top of these product-design considerations, however, it is important to consider overall project costs, as well. Using off-the-shelf motion chips results in a shorter development cycle, because the design effort focuses on the application itself, rather than on the low-level motion routines. In addition, motion-IC vendors provide developer's kits for their products. These products contain PC-based cards, along with programs and software tools that save time because they allow members of the design team to start working earlier. The software engineer can start writing motion sequences early, and the mechanical engineer can prototype motors and linkages before the application motion card is available. Once the application motion card is ready, the actual user-developed card can host the code for final integration and testing.

The other major part of the integrated control card is the amplifier. The amplifier takes relatively weak signals from the motion processor, phases them for the given motor and application, and amplifies them. The motor command from the motion processor can be a PWM signal, a ±10V analog signal, or a pulse and direction signal. Servo motors generally use the first two methods, and step motors use the third.

A newer variation for analog-motor output is the SPI (serial-peripheral-interface) format. In this scheme, the motion processor outputs a 16-bit signed motor command on a digital serial line. This method is useful because a number of DAC ICs and other amplifier controllers directly accept it. It's an improvement over the traditional ±10V scheme because it avoids conversion to analog on both ends—a noise-inducing and cost-incurring process.

Depending on the motion processor's architecture, the amplifier or the motion processor can do commutation. Note that only brushless-dc motors require commutation. If the amplifier performs the commutation, then the motion processor sends a single phase signal for each motor, and the amplifier uses Hall sensors to distribute the power to the correct motor coil. If the motion processor performs the commutation, then it sends multiple motor signals per motor (one for each motor phase), and the amplifier performs no commutation. The advantage of using the motion processor to perform commutation is that it can perform more advanced motor-control techniques, such as sinusoidal commutation and field-oriented control, because the motion processor "sees" all of the signals from the motor, including the encoder data stream, whereas the amplifier "sees" only the motor currents.

Another important consideration for amplifier design is whether you use current (also called "torque") control. With current control, an additional control layer exists between the motion processor and the motor. This layer measures the actual current through each phase of the motor and adjusts the drive voltage to match the desired current (from the motion processor) to the actual current that the motor measures. Current control generally increases the bandwidth of the motor, which means it can react more quickly to outside disturbances. Therefore, it is a must for high-end applications, such as machine tools. Lower power or lower performance applications may consider omitting a current loop as long as they include some type of overcurrent protection to protect against short circuits at the motor.

Once you make some of these decisions, two major design approaches exist for motion amplifiers. The first uses all-in-one low-power-amplifier ICs, and the other uses discrete components, such as predrivers, MOSFETs or IGBTs (insulated-gate bipolar transistors), and other circuitry. All-in-one amplifier ICs integrate PWM-signal input, current control, a charge pump, and a switched-output voltage drive into one IC unit. Although convenient, they top off at approximately 36V and 4A of output—often less, depending on the application. If your application exceeds these numbers, or if you are looking for higher performance and more control over the design, you should likely use the discrete approach to assemble the amplifier.

Until recently, designing a discrete-component amplifier from scratch, particularly one with current control and high-efficiency MOSFET switchers, was a complex undertaking. But in the last year, several companies, including International Rectifier and Performance Motion Devices, have developed a new type of motion-control IC. These devices, intelligent motor controllers, aim to interface with external switchers, such as MOSFETs or IGBTs, and to internalize functions such as current control, PWM generation, and shoot-through protection.

You can program a typical intelligent amplifier IC to perform commutation, torque control, and velocity control (Figure 4). It can communicate by means of a serial port to a microprocessor, or it can accept a direct SPI data stream for autonomous operation.

Future improvements for intelligent amplifier ICs may include more power-efficient commutation techniques and other features targeting high-power, high-efficiency motor applications. Power efficiency in traditional positioning motion control is not usually a primary concern, but for ancillary markets, such as white goods; electric vehicles; and industrial applications, including pumps, compressors, and air-conditioning units, it is rapidly becoming important.

Indeed, new motion ICs have made it easier to integrate amplifiers with motion controllers, thereby lowering cost and improving reliability. But designers need to carefully consider the build-versus-buy trade-off for motion-control applications and get a realistic idea of overall cost of the system, including maintenance and time to market.