Electric servo systems provide the most advanced and precise motion control available for increasingly versatile industrial applications. Servos excel in two distinct working modes: rapid point-to-point load positioning and smooth, accurate trajectory control between points, as in surface contouring.

One characteristic of a maturing technology like servo motion is the body of preferred design approaches and application guidance it builds, which can offer end-users valuable tips for implementing their specific projects. These “tips” range from rules-of-thumb (for inertia loads, proper grounding/shielding, and motor cooling) to sophisticated servo tuning algorithms. Several tips focus on the latter area because servo motion systems must be well-tuned to realize their full precision and dynamic performance.

Some application “art” goes along with the science of making a successful servo motion system. A logical way to start is to look at the overall system.

Buying the motion controller, drive, and servo motor designed to work together—as a system—will eliminate numerous problems, including wiring, configuration, and communication issues, according to George Ellis, chief engineer, servo technology, at Danaher Motion. One complication with separately sourced components is connecting leads, which might have different sequences at the motor, drive terminals, and feedback devices (encoder, resolver, Hall sensor, etc.). “Swap two wires and the motor might seize up or spin out of control,” says Ellis. “Similarly, get one of many configuration parameters in the drive or controller wrong, and poor system performance may result for no obvious reason.” Compatibility of set-up and tuning software also favors getting both motor and drive from one vendor.

While separate purchases of components fit some cases, Control Engineering’s Product Research surveys on servo motors also verify the system approach. In the latest survey (March 2006), respondents indicated a 71% preference for buying matched servo drives and motors versus 12% for separate units. “When you purchase servo products as a system, your servo system will come up faster and work better,” adds Ellis.

“Tuning is the process of setting several gains (typically between three and five) to get fast, stable response without excessive noise,” continues Ellis. However, tuning servo systems can be confusing, due more to unfamiliar principles than complexity. “Avoid a hit-or-miss tuning method, where gains are adjusted up and down in hope of obtaining good response,” he suggests. Without a tuning plan, gain setting can get out of control. “The servo vendor can provide tuning procedures for its products appropriate to the user, without the need for Ph.D.-level knowledge in control theory,” Ellis adds.

John Mazurkiewicz, servo products manager at Baldor Electric Co., notes that today’s servo controls allow manual or autotuning; seasoned engineers typically tune manually, although it takes longer to perform. For autotuning, he advises first to do a no-load trial, then tune under load, once the user becomes familiar with the control, load behavior, and load location. “Controls typically autotune current and velocity loops, however, Baldor’s controls are unique and tune the position loop—making setup even easier. Full autotuning is usually easily accomplished in 10 minutes,” he says.

When a servo application becomes difficult to tune—by PID methods alone—to precisely accelerate and position high inertia loads, consider using current feedforward, suggests B&R Industrial Automation Corp. Presence of a mechanically soft drive train (with transmission belts or long, slender shafts, for example), or high backlash, high friction, overhanging loads, or mechanical cam spring loads likewise call for use of this tuning method, also known as inertia feedforward (IFF).

“Instead of producing the required torque by the PID control loop, current feedforward method lets the drive calculate the appropriate current (torque) internally,” says Markus Sandhoefner, motion control specialist at B&R Industrial Automation. “To be effective, current feedforward calculations must be performed in real-time inside the drive. While current is fed forward, PID control remains active but does not have to be tuned stiffly to reach maximum precision.”

Substantial lag error is inherent to all PID control loops. In contrast, IFF effectively handles high load inertia and other difficult loads mentioned by cutting lag error dramatically (see “Inertia feedforward” diagram). B&R offers inertia feedforward tuning as a standard user-accessible function on its Acopos Series servo controllers.

A wide range of servo tuning techniques exists. However, machine tool builders often load only default parameters into their servo controls without tuning, notes Paul Webster, servo product manager, GE Fanuc Automation Inc. The machine may run well and nothing more is done. “Default servo settings are established to give basic performance in general cases. Just doing basic servo tuning can bring significant increase in machine performance,” states Webster.

Beyond the default stage, initial parameter settings for high precision (and high speed) are available via “one-shot function,” said to be easily set in Fanuc’s Series 0i-C low-end controllers, or by Servo Guide PC-based software on more advanced series Fanuc CNCs. The next important tuning step to improve servo performance is setting of velocity gain and resonance elimination filters using a Bode diagram, explains Webster. Tuning software, such as Servo Guide aids the process.

The Bode diagram evaluates servo system stability by analyzing frequency response of a control loop relative to its gain magnitude (measured in dB) and phase angle (deg.). Important results to look for include wide as possible bandwidth (near constant gain vs. frequency), with values under 10 dB, and gain margin frequency roll-off under -20 dB, states Webster. Filter tuning is effective for resonance elimination when positive frequency spikes are removed on the Bode plot and frequency roll-off is held under -10 dB.

To obtain full performance from servo systems, GE Fanuc provides advanced tuning functions on its CNCs. Simple, intuitive tuning navigators help guide users through the process steps, according to Webster. Noteworthy is a dynamic learning function that automatically reduces quadrant protrusion (or lost motion) in CNC machining applications. Quadrant protrusion (QP) consists of backlash—due to physical clearances between machine parts—and system “springiness,” also known as wind-up. The two distinct lost-motion components combine to act as system delay, producing a QP effect on the circle graph (see “Lost motion” diagram).

“The feature we use to remove quadrant protrusion is called 'backlash acceleration,’ which tunes backlash amount and adds a correction to the velocity command for lost-motion factors,” Webster says. Dynamic tuning occurs in successive steps—until the user or application is satisfied with the motion path accuracy. “Once started, the learning process runs by itself and, in a sense, the CNC learns from its initial “errors,” adds Webster.

Other servo system vendors also offer some form of lost-motion correction.

The ratio of load inertia to motor inertia is an important consideration. Different rules of thumb exist, with load/motor inertia ratio of 10:1 (or less) often recommended—for example, by Baldor Electric. “Some may suggest higher ratios, but the point is to limit the range of 'inertia mismatch,’ to allow tuning to be accomplished easier,” says Mazurkiewicz.

Bosch Rexroth Corp. similarly advises examining inertia magnitude of the “driven load.” Quick speed changes and positioning become very difficult if load inertia is high compared to the motor’s rotor inertia, explains Brian Van Laar, senior applications engineer at Bosch Rexroth. “In some cases the load may actually drive the motor during deceleration, causing overshoot and long settling times,” he says. The company recommends the following “good standards” for inertia mismatch: <2:1 for quick positioning, <5:1 for moderate positioning, and <10:1 for quick velocity changes.

One way to improve inertia mismatch is to increase gear ratio or ball screw pitch—if the application allows, suggests Mazurkiewicz. This has the effect of reducing load inertia magnitude as reflected at the motor. Select gear ratio, screw pitch, and motor at same time, and don’t leave motor sizing as an afterthought.

On the other hand, much higher inertia ratios can be accommodated by proper tuning. For example, Baldor mentions the case of manually tuning a servo system with a 144:1 inertia ratio mismatch for a packaging machine. However, it took six hours of tuning to fully satisfy the customer. Reduction in response can be expected at higher inertia mismatch values. Other unusual examples of high inertia applications are discussed in CE Sept. 1997, “Electric Servos Do More with Less” (pp.103-115).

Baldor’s Mazurkiewicz cites several common issues found in servo motor installations:

Motor not reaching speed results from insufficient voltage available. Select motors with at least 10% voltage “headroom” allowance to handle low-line conditions; and verify voltage by measurements at the motor.

Insufficient torque may be due to underestimated load (motor selected was too small) or magnet demagnetization (with ferrite magnets). “Demag” is easy to check via voltmeter or scope: Measure voltage with motor running at a test speed, then back-drive it at the same speed using another motor. If output voltage is not the same, the test motor is demagnetized. Re-magnetization must be done by the motor manufacturer, but check and correct what caused demag so it does not recur.

Overheating of servo motors comes from excessive torque demand—indicating motor undersized for the application—or presence of current ripple (check with a scope). “Ripple may be caused by improper tuning or improper alignment of the feedback device,” says Mazurkiewicz. He warns about “measuring” motor temperature by the touch of a hand. The housing temperature of a properly sized brushless servo motor will be 100-125 °C (212-257 °F)!

For resolvers, verify resistance with an ohmmeter, checking through motor-connector as well as connected cable. For encoders, use a 5 V supply and scope to check channels A and B for 5 V square waves measured between ±A and ±B signals. Do not coil up feedback cabling.

Some motor controls offer an alternate way to check feedback devices. For example, Baldor Series II drives include a “feedback fault enable” function to check for missing signal complements and can set a “fault.”

Internally generated heat affects all motors negatively. However, direct-drive motors (rotary and linear) also become a heat source affecting the accuracy of high-precision production machines they serve, because of their physical integration into the machine structure. As a result, direct-drive motors (DDMs) are mostly liquid cooled for optimal heat control. DDMs are often found in high-speed metal cutting machines, including laser type, wood-working machinery, and gantry stacker/destacker systems.

Karl Rapp, Bosch Rexroth automation and machine tool branch manager, recommends detailed attention to selecting the chiller capacity for the number of installed motors, their locations on the machine, their respective loads, and temperatures (min/max and ambient). Coolant pressure drop is a prime factor, influenced by location (height) of each motor relative to the pump, tube diameter, and length of the cooling system flow path. “In case of multiple motors, manifolds must be used to properly supply each motor,” says Rapp. “Proper flow must be measured with sensors at the coolant output of each motor.”

Water- or oil-based coolants are recommended, but not tap water, service water, or machine-cutting fluid. Chemical composition of such liquids can lead to coolant flow blockage and motor overheating with attendant stoppage of the machine and production. “Monitor coolant pH value monthly to avoid chemical erosion and leaks,” notes Rapp. He adds that automated control and monitoring of the cooling system has been simplified and made affordable by off-the-shelf temperature, flow, pressure, and pH sensors.

As the number of servo axes increases on a machine, desired throughput, precision, and flexibility become tradeoffs with processing power needed at the motion controller. B&R Industrial Automation recommends distributed motion control for certain multi-axis applications to ease the processing burden and communication between servo drives and motion controller. Target applications include printing, packaging, and filling machines, and assembly lines.

Distributed motion control (DMC) differs from the centralized approach that uses a standard controller—for example with 8- or 16-axis capability—then is forced to add a second controller if machine expansion later requires 10 or 18 axes, respectively. “In fact, DMC eliminates the need for a centralized motion controller or CPU, since all high-speed motion calculations are performed locally inside the drive,” says B&R’s Sandhoefner. “Also eliminated is the need for precise synchronization between motion controllers, and wiring is much simplified.” However, logic and system I/O functions remain in a centralized master controller.

With DMC, a master position is broadcast over a fast, real-time network (such as Ethernet Powerlink), allowing each drive to simultaneously calculate its own slave position from “a drive-internal cam,” he explains. Then, each slave position is processed in real time as the new setpoint for the position control loop—also closed inside the drive. “As a result, motion control performance no longer depends on number of axes installed in a machine, even if a new option module is added after the machine is in full production at an end-user site,” adds Sandhoefner.

As a final tip: don’t overlook manufacturers’ literature and Web sites as valuable information sources for your servo motion application.