Zeroing in on Precision and Accuracy

Oct. 1, 2001
Hitting the bull’s-eye time and again is the ultimate goal. Understanding the main sources of position errors and having an arsenal of devices to fight against them is the best defense against sloppy motion control.

Ah yes, the servo loop. It sounds so simple. And when it comes right down to it, there are only two issues involved — how close you come to hitting the final point (accuracy) and whether or not you can hit it again and again and again (repeatability).

The trouble begins with the forces of nature and physics. Foes such as friction and external torque impact things like rate feedback, current feedback, and position feedback. Then there’s gain and damping to be considered and manipulated in order to achieve positional accuracy. The good news is that many components have been designed to help fight against the enemies of precise motion and enable your system to work in a reliable and repeatable manner.

The following journey through words and images starts with a look at the three main sources of position errors and winds through a forest of solutions and examples that may just spark the ideal design idea in your quest for positional perfection.

Motion’s holy grail — precision

To discuss system precision in motion control systems, an understanding of the various sources of position errors is a must. These include three main elements: errors due to instability, errors in response to external disturbances, and errors due to mechanical transmission. In the following discussion these elements are considered and addressed by Jacob Tal, president, Galil Motion Control, Rocklin, Calif.

Instability errors

When a closed loop motion control system becomes unstable, it produces a cyclic motion, known as a limit cycle, superimposed on top of the motion. In extreme cases, the back-and-forth motion can overheat the motor and even damage it. Fortunately, this type of error is easy to detect and may be eliminated with something as simple as a digital filter. Furthermore, most controllers have the ability to do automatic tuning that selects the filter parameters such that the resulting system remains responsive and stable over time.

The degree of stability can be determined by a test called the step response. Here, the system is required to move back and forth, and the actual response is noted. An overshoot indicates less damping and a less stable system.

The following figures, Figures 1 & 2, show how an overshoot is corrected by increasing the D (damping) term in a PID filter.

Disturbance errors

To a motion control system, all outside effects are considered disturbances. These include mechanical friction, noise on input signals, and torque ripples in the motor. All of these inputs create disturbances that take the motor away from its nominal path, creating position errors.

To handle disturbances, the designer basically has two choices. One is to make the control system “stiff” and responsive. This means the control loop must have high gain that makes it reject the disturbances. However, as loop gain increases so too does the risk of the system going unstable, forcing the designer to walk the fine line between high gain on one hand, and system stability on the other. This is the area where automatic tuning software programs can be very useful.

A system’s ability to reject disturbances is related to its bandwidth — the range of frequencies it can follow. If the system can follow command signals at high frequencies, it can also reject disturbances at those frequencies.

Another way to deal with disturbance error is to reduce the disturbances themselves. This approach, however, is quite costly because it requires high quality elements. For example, friction disturbances can be eliminated by using air bearings, which are normally very expensive. Likewise, lead screw errors can be reduced with precision parts, and motor torque ripple by purchasing “smooth” motors. Given the cost associated with this approach, it is often used as a last resort.

Transmission errors

The most common transmission errors are caused by lead screws and include backlash and lead screw non-linearity. However, transmission errors exist in all mechanical systems. The common situation is that even if the motor is driven correctly to its proper position, the load may be off due to mechanical errors.

Continue on page 2

When such errors are repeatable, it is possible to compensate for them in the controller. For example, the non-linearity of a lead screw can be mapped, and a correction table stored in memory. Fixed backlash may be treated in a similar fashion.

In those cases where the error varies, however, the only alternative is to measure it by placing a sensor right at the output, rather than at the motor. This bypasses all the transmission errors and reads true output position.

There is a catch though. When the feedback device is placed on the output element, all the transmission dynamics become part of the closed loop and will affect loop stability. In particular, any flexibility in the mechanical elements or any backlash can easily lead to system instability.

To deal with the latter situation, it is recommended to use a control system based on a dual loop; namely, placing position sensors on both the motor and the load. The two feedback signals allow development of a more efficient control method that results in both stability and accuracy.

Tough torque tracking

The one motion variable on which it has been the most difficult to close the loop is torque. Until recently, if torque was measured at all, it was done with great expense and effort using strain gauges or other such sensors. But Fast Technology, Livonia, Mich., has developed a new way to measure torque that is both inexpensive and suitable for a range of applications.

The method, called Embedded Magnetic Domain (EMD) technology, employs a combination of magnetic induction and electronic processing to directly, instantly, and accurately measure torque from a static or rotating shaft without contact and in hostile environments. Thus, it is now possible to know both torque and power transmitted through rotating shafts spinning up to 100,000 rpm.

How it works

The EMD method involves several steps. First, a region of the shaft is magnetized, imparting a specific magnetic pattern; then, proprietary circuitry and signal conditioning are used to measure the magnetic field alterations generated through the twisting of the shaft as torque is applied.

The non-contact sensors give an output signal proportional to the applied torque. Using this method, torque can be measured from as far as 10 mm from the shaft — through dirt, fluids, and non-magnetic materials — at any shaft speed and with minimal hysteresis.

The torque sensing system consists of three basic components, including the magnetized shaft, a detection head that measures torque-induced variations in the shaft’s magnetic characteristics, and the signal conditioning electronics, which connect to the detection head and provide an electrical signal indicating the magnitude and direction of the torque applied to the shaft.

Sensors are supplied as complete measurement systems packaged in a single housing including torque shaft, signal detector, electronic signal processing unit, connector, and cable. The technology works in temperatures from -40 to 150°C. Low power consumption, high output signal, long-term measurement stability, and low cost make EMD technology viable in automotive, aerospace, power tool, machine tool, well drilling, and other applications.

Digital control freak

There’s a strong link between precision and digital control. Motors, for example, employ digital electronic control to compensate for load changes as well as regulate speed without the help of gears or pulleys. Digital control also opens the door to multimotor synchronization, efficiency, and reliability.

The key to digital control, however, is software, calling for ever faster processors capable of executing more instructions in less time with greater precision. To meet these needs, Texas Instruments Inc., Dallas, has developed a new highpowered DSP (digital signal processor) core. Known as the TMS320C28x, the new chip family features an instruction set that reduces common control operations to fewer steps. These DSPs can save costs by controlling motors with fewer motion sensors and minimizing motor vibration and noise by limiting signal resonances. The processors also are capable of sampling signals (such as line power) more frequently, reducing sample jitter and improving stability. In many applications, the additional DSP processing power may even be employed to analyze motor operation characteristics and optimize control, essentially using the motor as a sensor.

Picture this — high speed motion analysis

System designers often find it difficult to analyze and fix problems on automated production lines when processes are happening at what seems like the speed of light. Now there’s a new tool to the rescue. A high speed, high resolution video camera system for solving difficult motion analysis applications — by viewing mission critical operations with precision — is available from MCT Inc., Charlotte, N.C.

Continue on page 3

The Photron Fastcam- Ultima 1024 motion camera with advanced CMOS technology assures mega-pixel resolution and high speed digital imaging performance. This makes it ideal for numerous production and research applications where slow motion playback would optimize quality, yield, and productivity by monitoring performance and troubleshooting equipment problems.

For example, a packaging equipment OEM that makes machinery to place wrap-around labels on plastic bottles recently faced a big dilemma — labels were not being uniformly positioned on the bottles, causing an unacceptable level of rejects. After reviewing the slow motion video of the label placement mechanisms, necessary mechanical design changes were identified and implemented. Adjustments were then made to the timing system, allowing the packaging line to run smoothly.

Video cam tech specs

Image resolution of the camera is 1,024 x 1,024 pixels at a rate of 500 frames per sec. Advanced CMOS imaging technology allows for the combination of both high speed and high resolution image capture, which was unattainable using CCD image sensor technology. The camera provides high speed image capture of up to 16,000 images per sec and the ability to freeze individual frames of motion. Unique to motion analysis is the system’s capability to network up to 63 cameras and synchronize them from as far away as two km, assuring fast and safe control communications and image download from a remote location.

DSP goes the distance for Air Force

To build the aft fuselage for the Air Force’s new F-22 Raptor, the Boeing Co., Kent, Wash., faced a major challenge: Fabricate two titanium booms to support the enormous load of the horizontal stabilizer and keep the weight of each boom to 300 lb. And to ensure the aerodynamic properties of the airplane, the windswept surface of the booms needed to be formed to extremely close tolerances.

The booms are formed by joining subassemblies of titanium into a whole unit, using electron-beam welding. Once the joints of each subsection are complete, excess material is removed and the surface is smoothed so that each joint complies with the shape specified in the CAD design.

To achieve both strength and accuracy in the parts, Boeing adapted a number of technologies that were uncommon to aircraft production. One involved the FAROArm measuring arm, from FARO Technologies Inc., Lake Mary, Fla., that is used in a variety of manufacturing industries for taking quality measurements and reverse engineering. This is an articulating instrument that uses optical encoders at the “joints” to record X-Y-Z location and I-J-K orientation data of any point within its reach. Typically, the optical encoders provide a dimensional accuracy as close as ±0.001 in. However, for the project at Boeing, engineers needed to make more precise measurements and perform more complex model analysis.

In meeting these requirements, design engineers at FARO needed to work within limited space, power consumption, and reliability parameters. The additional functionality had to be added without increasing the size of the joints and compromising rigidity. With six joints and numerous discrete components at each joint, the probability of field and manufacturing failures would be increased with the addition of new components; adding discrete components to an already discrete-heavy design was not going to get the job done.

DSP decision

The decision to proceed with the DSP56F803 digital signal processor (DSP) from Motorola’s DSP Standard Products Division, Tempe, Ariz., was based on the device programmability and feature set. With embedded Flash memory, the DSP can be programmed during manufacturing allowing full functional unit testing before system integration.

The move to incorporate a processing unit in each of the six FAROArm joints provides additional computational power right at the point of measurement. Spatial measurements from the optical encoders in each joint may now be processed by the DSP. As a by-product of local encoder data processing, the amount of data traffic to the FAROArm main processor is reduced. Each DSP forwards its completed computations to the arm’s main processor, where 3D models are created.

The result is improved noise immunity and reduced power consumption, enabling lower weight and component cost, and longer battery life. Another benefit of incorporating multiple functionality into a single device is increased reliability. With a single device, a number of discrete components were eliminated from the design, increasing reliability while reducing costs and time-to-market. Chip debugging capabilities enable FARO engineers to program, test, re-program, and retest the DSP software many times daily until the desired behavior/performance is obtained.

Source: Victor Berrios, Product Marketing Engineer, Motorola, Semiconductor Products Sector

Continue on page 4

Consistent motion eliminates artifact lines

Large format digital image scanning has broadened into many fields including architecture, map making, and medical x-rays. In each of these fields, the need for clean scans without the interference of artifact lines is critical — none more so than medicine, as poor image quality leading to misdiagnosis can prove fatal. At the center of many of these systems is a linear motion device that controls the scanning motion. Consistency in this motion is the essential issue in a scanner design. Two main factors influence linear motion consistency during imaging applications — positional accuracy vs. repeatability and equally applied torque. Regarding linear slides, positional accuracy refers to the variation between a slide’s desired position and its actual position. This covers position in any plane or direction, not just linear position on the slide. Figure 1

Repeatability refers to the variation between a slide’s first movement to a position and its subsequent movements to the same position. This can be in either one or both directions. Variations in carriage movement and the ability to repeat those movements will always be present to some extent. These variations can be from end-to-end (pitch), side-to-side (roll), or twisting movements (yaw).

Built-up components increase error

The amount of error in a linear slide can be traced back through the manufacturing process of its individual components. Each is manufactured within its own tolerance range, but the more components required, the greater the combined tolerance or margin for error that is built into the final product.

In the case of imaging or scanning applications, variations in the slide’s carriage movement will result in artifact lines on the output, so it is critical to control the amount of variance built into these base components. Traditional component build-up of a linear slide includes the bearing shell, balls, raceways, seals, anti-rotation pins, retaining rings, shafting, mounting bolts, base rail, and carriage housing. For an average 32-in. stroke system, 300 individual components are required. In this type of slide, much of the tolerance stack can be taken out of the equation by applying “preload” or over-tightening the assembly to remove play between the components. However, preload may cause the unit to twist resulting in greater variance in motion. In addition, preload can reduce the life expectancy of the rolling elements and raceways.

Consistency of torque

Once a slide’s dimensional accuracy and repeatability are assured, the consistency of torque required to drive the slide becomes the next hurdle in producing quality images. Variations in torque can be caused by a wide variety of ingredients within the system. These may include the surface finish of the rails, type of lubrication, type of drive mechanism selected, types of support bearings used on the drive mechanism, connection of the drive mechanism to the carriage, type of motor, coupling to the motor, and more.

For this comparison, consider all factors to be equal outside of the base slide system; the type of linear guide bearing then becomes key in contributing to consistent torque. In the built-up components system, rolling ball elements are used; in the 2-piece Dolphin Guide system, the bearing surface is a Teflon-based composite called FrelonGold.

Rolling ball element linear bearings provide low friction upon initial installation, usually in the range of 0.05 COF. If maintained and lubricated properly, they will hold close to that friction over the entire bearing life. However, this low COF does not translate into consistent motion. Each linear bearing contains upwards of 40 hardened steel balls that continuously move in and out of the load zone while counter rotating and banging against each other, the raceways, shafting, and bearing ends. Each movement provides the potential for stick-slip, galling, and vibrations that are fed back through the entire system. Vibrations are magnified by the metal-to-metal contact of each ball with the shaft. Add to that the high susceptibility for damage to the individual balls from shock loads, corrosive wear, or foreign particulate. Flat spots, skidding, and galling all fight against smooth motion.

In contrast, the Teflon-based composite is bonded to the carriage at the molecular level, resulting in a bearing and carriage that have no internal moving elements. Because of the Teflon, the composite is a relatively soft material that eliminates all metalto- metal contact, which also allows it to handle foreign particulates well; any that migrate into the bearing are embedded into the material without damaging the bearing or raceway. The COF is higher at 0.125, but the material properties remain consistent for the entire bearing life, resulting in repeatable performance.

Voice your opinion!

To join the conversation, and become an exclusive member of Machine Design, create an account today!