Machine Design

Feedback for servos

The right feedback sensor is a key in motion systems that do what they are supposed to.

Rick Armstrong
Development Manager
Danaher Motion
Wood Dale, Ill.

These rotary encoders are typical of the wide range of styles, accuracy, and resolution available today.

Many servomotor manufacturers now integrate a feedback device directly into the motor housing. This AKM motor from Danaher Motion is designed to accommodate most types of feedback devices without requiring extensive modifications in the housing or on the feedback device.

Hall-effect devices are digital on/off sensors constructed of semiconductor material which senses the presence of magnetic fields. Here sensors are integrated into the end turns of the motor stator winding and are actuated by the rotor magnets.

Most encoders (both absolute and incremental) use a light source which projects a beam through a narrow slit in a code wheel and through a precision aperture onto a light sensor. The wheel revolves in synchronism with the servomotor rotor. Light falls on the sensor when the slit and aperture align.

Only the rising edge of one channel, typically A, is used when counting output pulses of encoders. Quadrature detection counts the rising and falling edges of both channels, increasing resolution by a factor of four.

Absolute encoding disks create a unique code for each disk position. Multiple tracks with one sensor per track generate a special digital pattern known as Gray code. Gray code's uniqueness stems from only 1 data bit changing state with each count. This eliminates ambiguity in output values caused by multiple bits changing at slightly different times. Since only one bit changes, the reading has only two possible sequential values. Other codes such as binary or BCD can be derived from the Gray code value.

Incremental disks are simple, having only one track shared by two sensors. The sensors are mounted so their signals are in quadrature, or 90° out of phase with each other. A special index slot provides a fixed reference for a starting shaft position.

Servomotor-powered motioncontrol systems are expected to be fast, accurate, and reliable. Yet even the most expensive servomotors can be victims of external factors degrading their performance. Leading the list is the feedback device sending position and other information back to the motor controller. So a fundamental understanding of the different types and uses of feedback devices is a prerequisite for speed and accuracy.

Servoapplications may use one or more feedback schemes. These schemes can be based on their location in the system, whether they measure incrementally or absolutely, or the sensing mechanism for detecting motion.

For position control the best location for the feedback device is at the load. Position errors created by mechanical play and slippage place loads in locations where they shouldn't be. Loadmounted position sensors eliminate the effect of this end play. These sensors are in addition to any feedback device already mounted inside the motor. Brushless motors require rotor-position feedback for electronic commutation and speed control. Some systems use these internal feedback sensors to monitor position by determining the direction and distance the motor shaft moves. When motor-mounted feedback devices can't be avoided it is important to keep the total error between sensed and actual true position of the load within acceptable limits.

There is one scenario when using internal sensors is okay. External sensors aren't needed when motors drive the load without any intermediate mechanism. Known as direct drive, the effect is the same as connecting external sensors directly to the load. There are rotary and linear motors with enough resolution that include feedback from their internal sensors for direct-drive applications. Direct-drive motors also minimize maintenance and eliminate compliance, or lost motion, that reduces the responsiveness or bandwidth of the system.

Feedback sensors report position in either absolute or incremental formats. An absolute position sensor reports a pattern or code that signifies its exact position within one electrical cycle of system power-up. By contrast, the incremental position sensor typically provides output pulses for each increment of motion. By means of counting the pulses, the system knows how far and in what direction it has moved.

Systems equipped with incremental encoders must be "homed" or set to a known starting position at power-up. From this home position incremental movements track how far the load has moved and in what direction. Combined with a position memory, the system reports an absolute position as long as power remains on. On power loss the system "forgets" its current location and must be reset to the home position before position data is again usable. Connecting the controller to an uninterruptible power supply prevents such problems for critical applications.

Some sensors are extremely rugged and target the industrial machine-control industry. Others are relatively fragile, intended more for precision laboratory equipment. And then there are applications where the requirements overlap. Semiconductor manufacturing calls for high accuracy in a clean environment with high-speed throughput to meet demanding production schedules.

Motion systems are either linear, rotational, or a combination of the two. Feedback sensors are specifically designed for each case. They may have different mounting features and motion direction, but the basic principle of feedback operation applies to either configuration. For linear systems like those used in XYZ positioning, the data indicates the exact location of all axes simultaneously. Information about exact position can be crucial in some situations like an E-stop (emergency-stop) event. Being able to restart the motion components at the exact point they stopped prevents machine jams and reduces waste.

Some systems calculate speed from position data by taking the derivative with respect to time. But such readings are valid only with no acceleration or deceleration component. Applications where speed information is important use a feedback device designed for that specific purpose, such as a precision analog tachometer.

Many motion-control manufacturers offer complete systems where the motor, feedback device, and drive are an optimized package. Such packages handle more than 90% of the motion applications today. The feedback device no longer mounts separately to the servosystem. Some manufacturers offer "smart" feedback devices that send the drive an identification signature listing motor and feedback details. These devices automatically set up and tune the drive system to the motor and feedback devices.

The best device for a particular application is the one that gives the required accuracy and resolution. Accuracy identifies how precisely the feedback sensor reports the position of its mechanical shaft. A 360° rotary device with a position accuracy of ±1% will report its shaft position within an accuracy of ±3.6°.

Resolution defines the number of divisionsper-revolution for rotary encoders or per inch (or millimeter) for linear encoders. It identifies how far an encoder must move to indicate a change of position. Resolution is specified in various ways: pulses per revolution, cycles per revolution, counts per revolution, lines per inch (millimeter), bits per turn, bits per inch (millimeter), etc. A rotary encoder that produces 1,000 pulses/rev need only turn 1/1,000 of a revolution to generate an output pulse indicating the encoder has turned.

If a 2,000 pulse-per-revolution (ppr) encoder has an accuracy of ±0.5%, position will be accurate to ±10 pulses. Doubling the resolution to 4,000 ppr also doubles the range of accuracy to ±20 pulses. Though the 4,000-ppr encoder has a higher resolution, it is no more accurate than the 2,000-ppr unit.

However, if accuracy is given in lines rather than percent, increasing the total ppr will actually increase accuracy. A 1,000-line encoder is accurate to ±10 lines. Its accuracy is 1%. But a 4,000-line encoder with the same number of lines is more accurate at 0.25%

Position sensors include Hall-effect, resolvers, general-purpose encoders (of a wide variety), and specialty encoders. Many servomotor suppliers offer multiple feedback options to accommodate a wide range of performance or environmental needs.

Among the simplest and least-expensive feedback devices are Hall-effect sensors. These on-off devices detect magnetic fields and trigger an output pulse as the field passes. Made of semiconductor material, they are rugged, operate at high frequencies (equating to tens of thousands of motor rpm), and commonly serve as sensors in six-step commutation for brushless motors. Drive electronics are simple because the Hall-effect sensor directly switches phase-power semiconductors. This makes Hall devices widely used for torque or coarse speed control.

Hall-effect devices come in standalone packages that mount within the servomotor housing. These sensors are sometimes embedded in the stator windings of brushless servomotors where the rotor magnets trigger them. The devices report the rotor-shaft position which software converts to speed or acceleration.

Six-step drives that employ Hall sensors produce less efficient torque than a true three-phase drive. Worse yet, they can generate high torque ripple. Torque ripple results from abrupt current changes that generate torque fluctuations. The fluctuations produce minute but detectable speed variations that may seriously degrade overall drive performance.

Resolvers are rotary transformers well suited for harsh environments of temperature extremes, vibration, and shock. Units are commonly rated at 155°C, with special models withstanding 230°C. Some models are radiation hardened. Frameless brushless types used in servomotors have little need of maintenance. A large throughbore can accommodate motor modifications such as hollow shafts and extension options. They handle motor speeds in excess of 10,000 rpm and provide moderate accuracy and resolution that suit most industrial applications at a low to midrange cost. Principle drawbacks include their lower accuracy and more sophisticated interface needs compared to encoders. Combined, resolvers and encoders account for more than 80% of the closed-loop feedback devices in motion control.

A resolver has one primary winding and two secondary windings wound 90° to each other. A typical excitation voltage from 400 to 4,000 Hz is applied to the primary winding. The secondary windings couple the input voltage of the primary trigonometrically according to shaft position. The voltage values of the two sinusoidal signals, one sine and one cosine, are compared and converted into digital signals in the drive controller by resolver-to-digital converters (R/D or RDCs) or by interpolation software in the drive.

Resolver-to-digital resolution (RDRES) specifies the R/D resolution of the controller for one complete electrical cycle and is usually stated as a power of two. An RDRES of 12 means the resolution is 2 12 or 4,096 counts/rev for single-speed resolvers. Typical RDRES values range from 12 to 16.

Resolvers can be single speed or multispeed. Speed refers to the number of electrical cycles per mechanical revolution. Single-speed units produce a single electrical cycle for one complete revolution of the shaft. A 12-speed resolver creates 12 electrical cycles for a single mechanical rotation. The impact of multispeed resolvers is felt when combined with the RDRES values. The counts-per-revolution increase by a factor of the resolver speed. An eight-speed resolver connected to a controller R/D with an RDRES of 12 produces 2 12 8 or 32,768 counts/rev.

Resolver output signals are relatively clean because they are analog devices. Their typical operating voltage of 12 Vac or higher makes them less susceptible to EMI noise. Even with higher noise margins resolver resolution is still limited by rpm. Single speed resolvers under 1,500 rpm can use RDRES values up to 16. Between 1,500 and 6,100 rpm the RDRES drops to 14. Over 6,100 rpm the maximum RDRES value is 12. Higher resolutions usually demand slower motor/resolver rpms.

While optical encoders are generally more accurate than resolvers special manufacturing techniques help improve resolver accuracy. Tooth-wound resolvers keep part-to-part variation to a minimum increasing output accuracy up to 50% over standard resolvers.

Encoders are classified by whether they're rotary or linear, incremental or absolute. They are also categorized by the method of sensing: optical, magnetic, or contact. When optical encoders first appeared they were praised for their ability to offer high accuracy in both low and high-speed applications. Despite this praise, at one point they were viewed as unreliable. Many of the problems stemmed simply from misapplication. They were installed on heavy industrial equipment where vibration and temperature took a toll on the fragile electronics and glass-encoding disks.

Today optical encoders find use in many areas. But even though they are more rugged than ever before, most manufacturers still recommend encoders only for light industrial applications typified by exposure to temperatures below 70°C and vibrations below 20 g.

The key to an optical encoder is the code disk. The disk has either slits or masked gradations that divide it into dark and light areas. Light sensors read the on-off pattern created as light passes through the disc creating output pulses equal to the number of divisions or lines on the disk. Hence optical encoder resolution, or granularity, is given in lines per revolution or pulses per revolution.

Absolute encoder disks have every position marked with a unique light and dark pattern. Individual light sensors detect the pattern sending the exact disk position to the output. Hard-coding the information to the disk means position information is available almost immediately after power is applied. Resolution of an absolute encoder is given as the number of bits in the output word or steps of the encoder. A 10-bit absolute encoder has a resolution of 2 10 or 1,024 steps. A drawback of absolute encoders is the number of wires that must be connected. A 10-bit encoder requires a minimum of 12 wires — 10 data, power, and ground — connecting the encoder to the controller.

Code disks for incremental encoders are much simpler. This disk is divided into equally spaced divisions of light and dark. Two light sensors generate two square-wave signals 90° out of phase with each other. The two square waves become output signals called channel A and channel B. The phase relationship between the A and B channels determines if the encoder is turning clockwise (B leads A) or counterclockwise (A leads B). Resolution is given as the number of output pulses in either channel A or B for one complete revolution.

A process known as quadrature detection senses each rising and falling edge from both channels to detect position. Quadrature detection boosts resolution 4 while maintaining the same accuracy. Higher resolution increases system repeatability and makes it possible to operate position and velocity loops at high gain ensuring system stiffness. Encoder resolutions of 50 to 5,000 lines/rev are standard, but line counts to 100,000 or more are possible.

Some encoders have additional channels to track shaft position or help noise immunity. Index channels output a pulse once each revolution to provide a reference or start position for the shaft. Complement channels output the complement of A and B channels for use by differential inputs reducing EMI and RFI on long wire runs. Hall equivalent and commutation channels offer other means of tracking shaft position.

Linear encoders contain a linear track and a read head. Linear tracks range in length from a few inches to several feet. Graduations etched in the track are scanned by the read head as motion components move. Like the rotary encoder the read head senses multiple channels, A and B, to provide position and direction data. Encoders with sinusoidal outputs use additional interpolation circuitry to electronically increase resolution.

Linear encoders are generally the best choice for equipment demanding high resolution in straight-line travel. Resolutions to 0.1 m are common. Some systems resolve down to 20 nm. Accuracy is typically 20 m/m, but may diminish linearly over the travel distance of the track. To compensate, a technique called slope error correction can bring any error to below 5 m/m.

Machines operating at high speeds use linear-encoder feedback because these devices typically work at higher speeds than other feedback devices. The main factor limiting speed is whether or not the electronic counting circuitry can keep pace. High-accuracy applications can have problems with errors from other sources such as leadscrew cumulative error, thermal expansion, or nut backlash. Linear encoders overcome these challenges.

Sine encoders offer high-level performance. Although more expensive than resolvers or incremental encoders, they are best in applications that need high accuracy coupled with high resolution. They are as rugged as a resolver and operate at speeds over 10,000 rpm.

Sine encoders resemble incremental encoders except the A and B data channels go to the controller as 1-V peak-topeak sine waves instead of square waves. The controller interpolates each complete sine wave as a means of increasing system resolution. This reduces truncation and quantization errors, allowing higher loop gains. Sine encoders can produce over 2 million counts/rev, or about 0.62 arc-sec of resolution. Such capability suits applications that handle high inertia or highmass loads. The greater resolution generates-more output pulses at slow speeds for finer control in accelerating, decelerating, and positioning these loads.

Like other encoders, sine encoders also may have commutation tracks, Hall emulation tracks, or auxiliary sinusoidal channels called C and D. These tracks provide absolute position within one revolution. The C and D channels resemble the sine and cosine signals used in resolvers.

An absolute, multiple-turn sine encoder is a variation of a sine encoder. These devices contain gears between the shaft and position wheel so the system knows the shaft position upon start up. They offer high precision, resolution, and accuracy for applications such as high-speed registration, film coating, and web control. Sine encoders also are candidates for low-speed operations where smooth rotation is critical. They help enable high gains, superior stiffness, and accurate positioning.

Feedback devices can generate electrical or optical signals. One advantage of using optical transmission lines for feedback signals is immunity to electrical noise or EMI/RFI environments. Noise can interfere with electrical feedback signals and distort data sent to the drive. It may be necessary to boost feedback signals out of the noise via amplifiers or signal conditioning devices. Newer feedback devices use ICs to convert and interpolate signals into more robust waveforms that overcome noise and propagation loss in connecting cables.

Electronic commutation in brushless motors

Commutation is the control of current to produce torque. In permanent magnet motors, torque arises when the magnetic field from the winding interacts with the field from the permanent magnet. Torque is optimal when current is channeled to the proper windings to produce the greatest interaction. As the rotor moves, the position of the windings change relative to the magnets. This means the optimal path to channel the current changes. In a brush motor, brushes and a commutator connected to the armature windings automatically route current to the optimal path. In a brushless motor the rotor position is fed back to the drive where power semiconductors switch current to the appropriate windings.



Danaher Motion,

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.