Manufacturing processes often march to the beat of short repetitive moves. These incremental or indexing cycles are prevalent on all sorts of production machinery, and they frequently hold the key to throughput and productivity.
Today, indexing speeds are determined largely by motors and drives and their ability to accelerate and decelerate quickly. Although many factors go into this, the main contributors are motor response, control techniques, and current generation. A little understanding here goes a long way toward increasing index speeds and making machines more valuable.
In many closed-loop motion control systems, acceleration is limited not so much by the motor, but by the way the motor is made to respond. And here, the one factor that's often overlooked is feedback resolution.
“Too often, designers select encoder resolution on the basis of the positioning precision required,” says Yves Villaret, Yet U.S. Inc., Manchester, N.H. If the resolution is too low, it will limit system cycle time. The answer, according to Villaret, is to select a higher resolution, which in turn, decreases settling time.
Rick Amendolea of Centricity Corp, Girard, Ohio echoes these thoughts. “Relatively low resolution — 4,096 post quadrature counts/rev — can prevent systems from achieving a high response. Slight deviations in counts per update can generate errors from which the control loop must respond.” These errors can be tuned out, however, using low-pass and notch filters.
Poor tuning — improperly setting gains to axis mechanics — can be a problem in and of itself, limiting motor control and response. “To correct this, designers should connect motor feedback directly to the drive, ensuring quick velocity and current loop updates,” says Rick Rey of Bosch Rexroth Corp., Hoffman Estates, Ill.
In applications with stepmotors, response times are optimized using both open and closed-loop control. “With both types of control, a stepmotor can run closed loop when it's about to lose steps, then revert to open loop once it regains control,” says Nick Johantgen of Oriental Motor USA Corp., Torrance, Calif.
Steve Bartz of Emerson Control Techniques, Eden Prairie, Minn., describes a similar approach for servosystems. “A servo drive employing state-space control with feedforward uses a position capture to mark the index's true beginning (in time and position). The control loop then uses this information to apply a gain to the algorithm, resulting in an accurate move.”
Another source of limitation may be found in the drive controller. “Drives have to process reference and feedback data, convert it to current in the motor, then deliver it in phase with shaft movements,” explains Ed Lee of Powertec Industrial Motors Inc., Rock Hill, S.C. Response tends to suffer if these steps aren't executed quickly enough.
One way to counteract drive limitations is by improving control algorithms, modulation techniques, and current feedback precision. “Increasing PWM and feedback resolutions strengthen response, as do advanced commutation methods such as sinusoidal and field oriented control (FOC),” says Sergey Lototsky, AC Tech Corp., Uxbridge, Mass.
John Chandler, Automotion Inc., Ann Arbor, Mich., adds this: “The field alignment technique, whether trapezoidal full step, sinewave microstep, or FOC, is important to motor control. The goal in rapid indexing applications is to maximize acceleration, and from a motor control viewpoint, the challenge becomes how to maximize torque. The solution is to quickly drive current to a high level at an appropriate field angle.”
Next in line when it comes to optimizing indexing speed is the motor drive. There are two main ways to drive synchronous motors. “The first is setting the current phase by presetting a driver's output voltage to one of six possible states,” explains Villaret. “State selection depends on shaft angle and is sensed by pole-sensor switches. However, this trapezoidal method does not deliver precise phase control since the driving voltage changes step every 60°. Sinusoidal drives, a second option, employ a short cycle for the current control loop and continuously change the voltage output at each phase, keeping current at a desired amplitude and phase.”
FOC is also well suited for indexing because it decouples time varying and spatial aspects inherent in motor control. “FOC offers direct control over torque and helps improve feedforward, gain scheduling, and notch filtering effectiveness,” explains Chandler.
“Drives that provide high current in a short time period, specifically unipolar PWM drives, are also very effective,” adds Johantgen. Those with large continuous or peak ratios let the motor briefly see a high peak load while accelerating. “Digital drives allowing precise tuning are helpful as well because they calculate motion trajectories internally (versus an external controller),” says Lototsky.
Lee also supports using digital drives. “This is because they evaluate feedback at very high rates and determine analog reference with high accuracy and quick sample times.” Rey concurs, “Digital commands preserve signal integrity. They permit high velocity loops, current loop gain settings, and control acceleration jerk to smooth velocity.”
For high acceleration, a drive must supply high current quickly. High current is supplied from high voltage. “Theoretically, both supply voltage and motor resistance limit current output,” says Chandler. According to Amendolea, “The output stage of drives with insulated gate bipolar transistors (IGBTs) sources power from capacitors (bus), supplying current quickly to the motor.”
For a driver to change current rapidly and control a motor at high speed, its voltage must be greater than the motor's back emf. “This is because only the difference between the two is available to drive current,” says Villaret.
Besides drive voltage and electronics, bandwidth also influences current levels. “Bandwidth is a measure of how fast a current loop commands output current to the motor,” explains Lee. DSP speed, control algorithm efficiency, feedback resolution, PWM frequency, and motor inductance all limit a drive's current-loop bandwidth.
Lototsky explains further: “Designers can implement forward modeling to increase bandwidth, essentially increasing the current supplied.” Forward modeling overcomes natural feedback delays and switching between modulation methods. “For more current, a drive with three different current controls can shift to a higher acceleration, then return to a run state,” adds Bob Parente, Intelligent Motion Systems Inc., Marlborough, Conn.
Vector drives are another way to supply a lot of current quickly because they maintain a magnetizing current to the motor. “With little time spent magnetizing the motor, acceleration is instant,” states Rey. In addition, properly shielding and grounding motor cables minimizes current loss.
Controls using feedforward look ahead to avoid reacting to lag errors, which can interfere with acceleration. Different types exist, namely velocity, acceleration, speed, torque, current, and position. Feedforward works best during motor acceleration and deceleration, versus continuous operation.
“Feedforward can look at a velocity increase, take the derivative, and calculate acceleration needed in the drive,” says Rey. The drive then provides a higher current when an acceleration command is given.
However, velocity feedforward can hurt settling time by creating unwanted overshoot. “For this reason, most drive software lets users tune the percentage of feedforward to apply,” says Villaret.
“Acceleration feedforward, by contrast, uses information about a moving system to estimate torque output.” It lowers tracking error at acceleration changing points, including at end-of-movement, thereby reducing settling time. Acceleration feedforward must also be tuned for the specific movement and system.
Feedforward is a “non-error” based gain, and because it adds an additional speed or torque command to the control loop, anticipates and compensates for system error. “Specifically, it minimizes following error during the move and raises other gains,” says Amendolea.
The time a system spends looking ahead is less than the natural resonant frequency of the motor. “Torque feedforward minimizes tracking error by predicting how much torque will advance a rotor's position and then commanding the phases to move there,” says Johantgen.
Controllers running feedforward in current and speed-position loops are helping digital drives with high-speed processors outperform other models. “This is because digital drives often employ algorithms that predict current or velocity, based on a present situation,” says Lee.
All closed-loop system regulators are driven from the error between actual and desired set points. “With models that most closely match the actual system, an input reference is obtained exactly for a desired output. And, there is less overshoot and a shorter settling time, thus minimizing tracking error,” explains Lototsky.
However, feedforward does not change the dynamic response of a feedback loop to error. “But it does reduce disturbances that would otherwise excite or produce error in the loop,” explains Chandler. “Because feedforward operates in open loop, it is most effective when the actual demand signals are repeatable.”
Steve Bartz of Emerson Control Techniques, Eden Prairie, Minn. expounds on feedforward, adding another twist. “Drives employing state-space control with feedforward use a position capture to mark the index's true beginning (in time and position). Then, the control loop uses this information to apply a gain to the algorithm, resulting in a fast, accurate move.”
Thin is in
A big drive train is like a freight train — anything but nimble, totally incapable of frequent stopping and reversing. This presents an optimization challenge.
Naturally, a drive must be powerful enough to start and stop its attached load as well as its own weight. Some loads demand larger, more powerful drive components — but larger drive components further increase system inertia, requiring a greater percentage of power to move the drive train versus the load. This is especially so when a (low inertia) servomotor does the cycling. “One alternative here is to let the motor run constantly, and meter its power to the system with a clutch,” says Jeffery Watkins, Force Control Industries, Fairfield, Ohio.
Another option is to reduce drive train mass with alternate designs and materials. “Most industrial power transmission products were originally designed for use with motors that produce 1,750 rpm of shaft velocity, so high speed as we know it today wasn't a common objective. Bulky steel parts were often designed to maximize durability and torque-carrying capacity over time,” explains Andrew Lechner of R+W America, Inc., Bensenville, Ill. Today, the number of applications requiring high speed and efficiency with low inertia has significantly increased, with dramatic changes in mechanical designs for motion control. The next step, says Lechner, is the use of thermally stable and robust composites in the components themselves.
But the issue isn't just mass. “The distribution of the load and how it's coupled to the motor are two major response limitations,” says Rick Amendolea of Centricity Corp, Girard, Ohio.
Quantified as moment of inertia, a body's resistance to change in angular velocity is determined by mass and its distribution about the axis of rotation. This inertia — along with system acceleration — dictates the motor torque required for an application. “System maximum response, or bandwidth, is directly related to the ratio of reflected load inertia to motor rotor inertia,” adds Amendolea. The higher the ratio, the lower the bandwidth and system response.
The most effective way to keep inertia down? Keep component diameters small. For example, for belt drives, versions made of newer, more sophisticated materials can bend around smaller-diameter, low-inertia pulleys without fatiguing. Similarly, “multiple-disc clutches generate increased torque without the penalty of increased inertia that comes with larger diameters,” says Watkins.
The opposite is also true. Hollow shafts may be lighter and stiffer than solid shafting, but because mass is distributed farther out, they carry higher inertia about their central axes.
Slack causes shock
Quickly reversing directions can send shock rippling through a system. Looseness and backlash exasperate the problem, because temporarily unengaged components drifting in one direction literally get slammed into reversal, over and over. Thus, keeping backlash low reduces speed buildup of power-side components prior to reengagement with their freewheeling mates — so there's less impact. This is why tight connections are crucial.
For example, quill and keyed coupling and pulley connections fit loosely to assemble easily, but jarring from indexing reversals quickly wear their keys. Compression and interference-fit shaft connections, on the other hand, are more suitable for high cycling because they reduce and withstand impact from frequent torque reversals. Gears are the same way — looser-meshing teeth, while beneficial in some cases, don't cut it with indexing — because the repeated reversals make for frequent tooth knocking.
There's no substitute for a tight mechanical system, either; backlash always reduces the responsiveness of high-speed systems. “Control-loop tuning gains are based on load inertia and friction, general losses in a mechanical system from bearing drag, seal friction, and gear losses,” says Amendolea. “So if system gains are set for maximum-response bandwidth with these loads in place, the system will probably go unstable if these loads are suddenly absent. This is essentially what happens when there is backlash in the mechanism.” The motor is decoupled from the load and what might be a 20:1 inertia mismatch with significant friction has now become a 1:1 inertia match, with very little friction. The result is instability.
The right rigidity
If backlash is looseness between components, torsional compliance is looseness within components. An expression of material rigidity, torsional compliance is how much mechanical parts wind up under torque. Here though, some give can be beneficial. For example, servo-insert couplings are actually designed for lower torsional stiffness. It serves to damp impact loads and reduce their potentially harmful effects on the mechanical system. “Insert couplings do in fact store energy, and spring back during fast indexing,” explains Lechner. Why doesn't this storing and releasing of energy increase shock then? “Any impact involves a rapid velocity change of one element relative to another,” says Lechner, “but low torsional stiffness tends to increase the load's rotational acceleration and deceleration time.”
That's how insert couplings decrease force of impact resulting from torsional compliance. Proof is that insert couplings allow loads to oscillate, however slightly and briefly, once the motor has stopped: “This persistence of motion subsequent to the motor reaching zero speed is evidence that the energy is being transferred more slowly than in systems lacking torsional compliance,” continues Lechner. One caveat: Reconfiguring mechanical component geometry can reduce material, inertia, and potential imbalances, but the challenge here is that a level of rigidity must be maintained. “That's why replacing metal components with lighter, high-strength plastics will most certainly be a future trend in power transmission design,” says Lechner.
If compliance doesn't serve a specific design function, it should be minimized as much as possible. “Compliance in drive shafts, belts, and gear trains limit overall system response because it introduces resonant points and nonlinear effects difficult to correct with control loops. Even if control loops can compensate, often the torsional windup of the mechanics, from a positioning standpoint, cannot be tolerated,” Amendolea explains.
Because cams roll smoothly through reversals with one-way rotation, they've always been a natural for indexing. It's like doing laps around a track versus running back and forth between two points — their reversals are inherently easier. Cams are also a low-compliance option. “Not only do cams produce complex motion, they're also rigid — making them particularly suitable for high-speed indexing applications,” explains Mark Combs, automation sales manager for Sankyo Inc., Sydney Ohio. Modern cam profiles are synthesized from motion curves calculated to generate optimized movement. This technique allows one well-designed cam to control an entire machine — with more sophisticated performance than one might expect by just eyeballing the cam profile. “For this reason, mechanical cams can actually have shorter cycling times than electric and fluid-power actuators,” — with cycles in excess of 1,000/min. in some cases — continues Combs. That's because cams avoid the other flavors of compliance — namely, electrical saturation delay and decay, and soft response due to air compressibility.