Drives and Controls: Feedback Elements

3.1 Rotary and Linear Incremental Optical Encoders

Optical encoders (ill. 3) can be characterized by the physical measurement principle they use (diffraction or directed light), their design features, and the protection requirements to which they are built. They range from completely enclosed and sealed units to open-frame kit units. They are typically used in velocity- or position-feedback systems such as those found in tape transport equipment, machine-tool spindle controls, bed positioning equipment, woodworking machines, robots, material-handling equipment, textile machines, plotters, printers, tape drives, and a variety of measuring and testing devices. Commercial encoders are generally defined as being capable of measuring angles of up to 3" . For higher resolutions, an angular measurement device must be used. These devices are capable of measuring angles as fine as 0.00001" (0.03μ).

There are three categories of encoders from an environmental protection view-point.

Sealed encoders are generally protected to the levels of IP 64 or better. These are stand-alone units that have internal bearings and seals and are not intended to allow user access to internal workings. Self-contained encoders are not necessarily dust proof. These have internal bearings and are stand-alone units, but some customer access may be possible or may even be necessary during installation. Modular encoders are completely open units which rely entirely on the application for protection.

These units don't contain bearings. They are sometimes referred to as kit encoders or tach kits.

Sealed units are the most expensive, and generally are not well suited for high-speed operation because of the seals. However, these can be very high accuracy, high-resolution devices, capable of resolutions ranging up to 10,000 cpr.

Modular encoders are the lowest in cost. These units generally have the best price-to-performance rating, but they require some care on the part of the user as they can be damaged if not installed properly. Modular units are available with resolutions up to 2500 cpr.

Self-contained encoders span the entire performance envelope, at a slightly higher cost than modular devices. The self-contained hollow-shaft encoders are widely used in the drive industry, as they eliminate coupling resonance.

Hollow-shaft encoders are also widely used with integrated commutation electronics.

This provides a simplified assembly process to the manufacturer by allowing elimination of the Hall board. This approach also simplifies overall alignment.

FIG 3 Rotary optical encoder.

Terms

amplitude modulation Using the code wheel and mask as an optical shutter, or to create Moiré patterns to modulate the intensity of light impinging on the photodetectors.

code wheel A circular disk of transparent material with patterns of transmissive and opaque regions equally spaced about the perimeter. Light shining through the clear regions is passed onto the mask. The spacing on the code wheel defines the line count of the encoder. 1024 opaque regions separated by 1024 clear spaces will create a 1024-cpr encoder.

disk Another term for code wheel.

grating A pattern of closely spaced lines which is used to shutter light passing through the code wheel.

index See reference mark

mask A glass plate mounted on the encoder housing so as to remain stationary with respect to the rotating code disk. The mask supports the optical gratings or patterns.

Moiré patterns When light is transmitted through a set of gratings that are equally spaced but at a slight angle to each other, patterns of brightness and dark are created.

These patterns are called Moiré patterns. As the gratings are moved relative to each other, periodic brightness fluctuations can be seen.

phase modulation Using a reflective mask with a stepped grating pattern to modulate light impinging on the photo-detector via constructive and destructive interference.

phase plate Another term for mas k. More appropriately used when referring to encoders using phase modulation of light rather than amplitude modulation.

reference mark A once-per-revolution output that's one period wide.

reticle Another term for mask.

Principles of Operation. The basic components of rotary optical encoders are as follows (see ill. 4):

Light source, which can be a lamp or a light-emitting diode (LED) Collimating (condenser) lens to improve light power density and reduce diffraction effects Code wheel Mask Signal detectors Output-conditioning circuitry

ill. 4 Directed light scanning.

The sensor operation results from photo-electrically scanning very fine gratings on the disk. A disk with a radial grating of lines and gaps serves as the measuring standard. The opaque lines can be made using a number of methods, such as plating chromium onto the glass. The lines are placed so that the spacing between lines and gaps is equal, and the lines are spaced uniformly around the circumference of the disk, so as to make a circular graduation.

In close proximity to the rotating disk is a scanning reticle, with grating fields for the data channels and one or more fields for the reference mark. The data-channel windows are placed onto the scanning reticle such that they are phase-shifted in relation to each other and the graduation pitch by one-quarter of the grating period.

All of these fields are simultaneously illuminated by a beam of collimated light. As the graduation rotates, the light is modulated onto the sensors, and the sensors then output two sinusoidal signals with a 9 0 deg phase shift between them.

Reference Mark. The reference mark is created by a peripheral set of gratings in tandem with or adjacent to the main data windows. Sometimes the reference is made by a constant light source outside the code wheel and a single window in the mask area. The reference mark produces a single pulse that's one period wide. The reference can also be digitally combined with the main quadrature signals so that it's active only during a specific portion of the quadrature cycle. This is called gating the reference mark.

Typical Resolution. Commercial encoder products are available with resolutions up to 10,000 cpr. Above this value, different techniques must be used in the design and manufacture, increasing overall cost significantly.

Methods of Fabrication. Rotary optical encoders can be constructed using either amplitude modulation (AM) or phase modulation (PM) techniques, but AM is far more common due to the lower cost of manufacture.

PM methods are used for very high resolution devices which might be found on the z axis of a machine tool or, more commonly, in a linear optical encoder used for measurement equipment. For a discussion of PM methods, refer to subsec. 3.3, Linear Optical Encoders.

Graduations. Three major materials are used for manufacturing the mask or disk graduations:

Chrome on glass Estar-based film or photoplastic Metal The disk graduations can be made by either expose-and-etch processes or plate-up processes. Expose and etch is very similar to processes used by the printed-circuit board industry. The plate-up approach was developed by Dr. Johannes Heidenhain, GmbH, and is called the Diadur process.

Plate-up processes yield much better edge quality, but require extensive investment by the manufacturer. Etch processes utilize the same materials and techniques developed for the semiconductor industry, and so require very little investment on the part of the manufacturer to implement. Etch processes are used exclusively for graduations on photoplastics and metal.

The expose-and-etch printing method is as follows:

1. The graduation is produced by placing a master plate against a blank plate that has been coated first with chrome and then a photoresist.

2. The blank is exposed using a high-intensity ultraviolet (UV) light.

3. The exposed blank is then developed and etched. The etching process produces a duplicate of the master image.

4. The disks are cut from the blank, cleaned, and are then ready to be installed.

Centering Process. The centering process for placing the disk onto the encoder shaft is crucial to the performance and accuracy of the encoder. The graduation must be placed as precisely as possible with respect to the rotational axis of the shaft or hub. Typically, the concentricity of the disk pattern to the rotational axis must be better than 0.0004 in (1 0 deg u m).

Light Sources. Light sources can be incandescent or solid state (LED), depending on the environmental constraints and cost targets.

Solid-state light sources are used more predominantly due to their long service life (in excess of 100,000 h). They also have excellent resistance to shock and vibration.

However, because they are silicon devices, they are limited to junction temperatures of approximately 15 0 deg C. This results in limitations on their use at high ambient temperatures. The output of LEDs also drops about 1 percent per degree Celsius, so use at higher temperatures must be evaluated carefully.

Incandescent illumination sources are used when environmental temperatures are extreme, 12 5 deg C or higher, due to their ability to withstand a higher ambient temperature (up to 20 0 deg C). They also have about twice the output of LEDs.

Most sources provide light in all directions, most of which won't fall on the detectors. To improve this situation, a collimating lens is used. Collimation gathers the light and focuses it at a point at infinity. The result is a parallel beam which can be precisely directed at the photo-elements. This provides three improvements:

It serves to combat the intensity loss due to the inverse-square law.

The reduction in scattered light reduces crosstalk and noise at the detector.

When parallel light passes through the disk-mask "shutter", there is less leakage due to stray light, which results in better modulation and more useable signal from the detectors.

Photo Detectors . There are three primary types of photodetectors used in optical encoders. These are the solar cell or photovoltaic device, the photodiode, and the phototransistor.

Photovoltaic devices. These are solar cells, or photodiodes being used in a photovoltaic mode. These devices generate electricity when light impinges on the detector surface. They don't require external power. When connected to a load, the voltage potential created by the illumination results in the generation of a current. These devices have a very broad spectral response, and are particularly sensitive to the infrared region. They have excellent frequency response and are resistant to most environmental contaminants.

Photodiodes. By connecting a photodiode anode to a power supply, and the cathode to a load resistor, the photodiode operates in the photoconductive mode.

In this mode, the device acts like a valve which controls the amount of current flowing through the resistor from the voltage source, depending on how much incident light is present. These devices retain the excellent frequency response characteristics of the photovoltaic devices, and generally need less detector surface area to develop equivalent output signals.

Phototransistors. These devices trade off the frequency response of photodiodes for increased output levels. Phototransistors can generate significant output voltages > 1 V), which makes them superior for use in noisy environments. However, they are significantly slower to respond than photodiodes, which results in reduced frequency response of the encoder. Phototransistors can also be implemented in less area than can photovoltaic devices. Significant signals can be generated for a device as small as 0.021 in^2 (0.5 mm^2 ).

Signal Conditioning. ill 5 diagrams a typical single-sided-supply photo-diode sensor arrangement. This circuit uses comparators to create square-wave quadrature output signals with 50 percent duty cycles.

The selection of values for R2 and R3 controls the amount of hysteresis in the circuit, while the values for R1 and R4 control the photodiode output-signal levels.

Balance Adjustment. To develop a 50 percent duty cycle at the output of the comparator, the input offset levels must be identical. This will never occur naturally for a number of reasons, but primarily because the amount of light shining on the two detectors will never be exactly the same, and the photo-detectors won't have exactly the same characteristics. Most encoders therefore require adjustment as part of the final manufacturing process. This can be accomplished in the following ways.

Shading screws. By physically blocking light to one or both of the complementary sensors, their outputs can be matched. This is a robust process, but somewhat slow and difficult to automate.

Analog and digital pots. By replacing one of the load resistors with a potentiometer, the input voltages to the comparator can be made equal. This process is very easily accomplished, but the different resistance values can cause problems over temperature, and potentiometers decrease the overall reliability. Digital potentiometers can be adjusted via a computer interface, which makes this approach very amenable to automation.

Test-Select. This process selects fixed-load resistor values at the final test of the encoder. This is very time consuming, but the fixed values are more stable than a potentiometer. This process is moderately difficult to automate.

ill. 5 Optical encoder signal conditioning.

Balanced sensor array. This process uses an inter-digitated pattern of detectors to eliminate the need for balance adjustment. The distribution of sensors throughout the area of illumination results in an overall averaging which balances the outputs.

This process is very efficient to manufacture as it eliminates adjustment entirely.

However, each resolution must be tooled uniquely, which makes the capital investment in this approach very high and causes long lead times when new resolutions are needed.

Output Signal Qualities. Signal processing of the sinusoidal outputs is handled in two ways, either as analog information or as digital information.

Analog outputs come in a number of flavors, the most basic form of which is to supply the raw sensor signals. These are low-level signals in the microamp range, which must be carefully shielded and cannot be sent over long distances. The next most common analog output is to provide simple amplified signals. These can be implemented as dc-biased ac signals, which can be driven by a single-sided power sup-ply, or as amplified zero-referenced signals using a dual power supply. Amplitudes of either approach are somewhat user driven, but 100 mV for the single supply and 2.5 V peak to peak for the amplified approach would be common values. The third form is very common in Europe , and is termed the 1-V peak-to-peak output. This output is guaranteed to hold this level + 0/ - 3 dB) over the rated frequency range, which can be as high as 200 Hz. These units are also capable of driving significant lengths of cable, and the constant level sinusoidal signal is excellent for use with interpolation electronics.

Digital signals also come in some variety. When the encoder produces quadrature outputs, the signals can be formatted as TTL, HTL, line-driver, high-voltage line-driver, complementary metal-oxide semiconductor (CMOS) line-driver, buffered, and open-collector variations. In any case, the signals are digital in nature, switching between ground and the supply or high-voltage value determined by the application.

The goal of these outputs is to retain pulse width and symmetry over all frequencies and temperatures. These signals cannot be interpolated, although edge counting of the quadrature signals is common. Another version is direction sensing.

These signals are usually either TTL or HTL, but anything is possible and has probably been sold at one time or another.

Accuracy and Resolution. The resolution, or measuring step, of an encoder is the angle corresponding to the distance between two edges of the square-wave pulse-train output. Basically, this is one-quarter of the grating period. Accuracy can usually be approximated as 5 percent of the grating period for resolutions up to 5000 cpr.

Between 5000 and 10,000 cpr, accuracy is basically constant at approximately 12 arc sec. (Dr. Johannes Heidenhain, GmbH). There are many texts discussing this issue (Electro-Craft Handbook, 1980; Ernst, 1989). Error consists of intrinsic instrument errors in the encoder, plus system errors.

System errors are due to the following causes:

Hysteresis effects. The amount of hysteresis used to control noise will effect over-all accuracy, as this changes the switching point of the output in a TTL system and introduces phase lag in an analog system.

Runout due to eccentricity of the disk and hub assembly with respect to center of rotation. Eccentricity errors are created by manufacturing process accuracies associated with putting the disk on the hub, tolerance between the hub and the motor shaft, bearing runout, and the accuracy of the pattern itself. This type of error will result in amplitude modulation of the output A, approximated by

R = 1 in.

delta R = 0.0005 in

delta A = delta R / R + delta R

A = 0.05%

Surface runout. This is either due to poor mounting of the disk to the hub flange or, in a modular encoder, due to tolerances between the hub and motor shaft of the motor shaft runout. All of these can result in variations in the gap between the disk and the mask. The angular error due to shaft runout (arc minutes) can be approximated as follows:

60 x sin^-1 (TIR / Rt)

where

TIR = motor shaft runout

Rt = nominal data track radius For a 0.75-in track radius, this is 0.45 8’ /0.0001 in.

Pattern errors which cause both amplitude and frequency variation errors. Frequency errors appear as "flutter" on an oscilloscope. This is caused by irregular spacing of the opaque patterns on the code wheel. These errors can result from errors in master generation or from printing errors. Many times, these errors will be cyclic, occurring every 45 or 9  mechanical. These errors result from certain types of master generation processes in which a section of the disk pattern is stepped and repeated to make the entire 36 0 deg. pattern. Other errors occur with pattern-generation equipment, called closing errors. These occur when a small error results over the 36 0 deg. printing cycle, so that the last line generated is slightly larger or smaller than all the rest.

Jitter. This can occur when the alignment of the elements in the optical path is incorrect, the illumination source is poorly collimated, or contamination is present on the disk or mask surface.

Sensor output drift. Most encoders use a push-pull configuration to minimize the effects of detector changes, light variation, and voltage variation. When the sensors drift out of balance with each other, symmetry in the quadrature output will change.

Interpolation. There are many methods of developing higher-resolution TTL out-puts by processing the analog sinusoidal signals developed in the measurement sys-tem.

One consists of developing phase-shifted copies of the original signal using resistor networks. Taking advantage of the relationship

sin ( )  cos  sin  sin  cos

the base sinusoidal signals sin  and cos  are multiplied by phase-shifted copies. E.g., a  interpolator would use 5 sets of signals, each shifted 1  . The results are converted into square waves via comparators, and all of the outputs are routed through an exclusive OR gate. The result is a set of square waves in quadrature at a frequency equal to 5 times the original. ill 6 shows how interpolation of  would compare with the original output.

Interpolation of this type can be used for multiplication up to 2  with reason-able success. Higher subdivisions are obtained using digital methods. One such method computes the arctangent using the values of the two analog quadrature signals as the sine and cosine values, then uses table look-up methods to determine the corresponding angle. Quadrant detectors complete the calculation. Another method makes readings of the analog values at two discrete times, and then creates an artificial pulse train to get between the two at the desired resolution. Similar methods are used for resolver-to-digital converters, and are discussed in that subsection.

Application Considerations

Environment. Encoders are very robust sensors, but they need to be selected for the intended environment. The main limitation in the application of encoders is temperature. Most commercial encoders are rated at 8  C or lower. Industrial ratings increase this to  10 to 10  C. Severe-environment encoders operate up to 12  C. Shock and vibration are rarely a problem. Even though many encoders utilize glass disks, these assemblies are very robust and can withstand most military levels of shock and vibration. In fact, it's very difficult to damage an encoder mechanically and not damage the motor it's mounted on.

ill. 6 Interpolation of  compared to original output.

Interface Requirements. In any encoder application, it must be decided what signal levels are needed for interface with the controls, what type of circuitry the encoder will be connected to, what frequency response is needed, and what type of signal will be sent through the cable, as well as mounting and coupling requirements.

Slew Rate. The encoder slew rate is limited by either mechanical or electrical considerations. Mechanical limits are encountered when bearing limits are exceeded, or when testing has shown that the assembly isn't capable of remaining intact under the rotational stresses. Electrical limits are encountered when the input frequency from the sensors to the signal conditioning circuit exceeds the response capabilities of that circuit. This relationship is stated as follows:

nmax  10 3

 60 rpm

fmax

z

where f  scanning frequency, Hz

z  encoder line count, cpr ill 7 shows how frequency response, encoder resolution and input rpm are related.

ill. 7 Frequency response capability.

Interconnection. Applications in very noisy environments, or which must drive long cables, should use differential line drivers. Shielding and grounding are also important, but this can also drive sensor cost dramatically. Cable length, environ-mental protection, and signal types all play together. Long cables in a high-noise environment are best dealt with using amplified sinusoidal signals with current drivers, feeding into a line receiver. For cables less than 100 m in length, TTL signals can get by, but care should be used at this distance. For best control, cable shields should be tied to the control, and the control to ground. In Europe , it's also desired that the encoder case be tied to the cable shield and that the power ground remain isolated. Some examples of suggested interface circuits are shown in ill. 8.

Mounting Requirements. There are several standard mounting patterns. For hollow-shaft encoders, there are also various styles of spring-plate adapters, which are very important to the performance of the installed device. These couplings must be designed to allow for high torsional rigidity, while being compliant in the axial direction. It is a design goal for these couplings to have a natural frequency exceeding the application bandwidth by a factor of E.g., a system with a planned servo bandwidth of 100 Hz should have an encoder flex-coupling mount with natural frequency of  1000 Hz.

Motor End-Play. A modular unit will require approximately 0.010 in of motor shaft end-play to maintain disk integrity. Hollow-shaft encoders with flexible mounting plates can usually accommodate as much as 0.040 in.

Power Supply Constraints. Because encoders utilize LED or incandescent illumination, they can draw significant amounts of power. It isn't uncommon for an encoder to require 250 mA or more in a high-temperature brushless servo application. The designer should check to be sure that the drive system has sufficient power to support an encoder application. This is especially important in commutation encoder applications, in which a system that was designed to support Hall sensors is now being connected to a commutation encoder. Most power supplies for Hall sensors are very low wattage units, and they may not be able to support the encoder requirements.

3.2 Single and Multi-turn Absolute Rotary Encoders

Absolute encoders are manufactured in exactly the same manner as incremental encoders. The main difference is that more sensors are used, and so they are more complex than incremental encoders. The overall complexity depends on the number of bits, or word size of the encoder-the more bits, the more complex and expensive.

They are used where motion can occur when power is removed, such as to provide level control or fail-safe operation. Machine-tool and robotics applications are the primary users of these devices.

Principles of Operation. An absolute encoder uses one track of the code disk for each bit in the output. Therefore, an 8-bit absolute encoder has 8 tracks on the disk and requires at least 8 sensors to detect light passing through these tracks. Depending on the size of the encoder, the sensors, and the tracks, it may be necessary to use multiple sources of illumination to assure adequate signal levels.

The data tracks can be encoded to provide position information in a number of ways. One method is to encode the data as pure binary information. In this approach, each track is equal to a power of 2. One disadvantage of this approach is that it requires many simultaneous bit transitions. E.g., when counting from 15 to 16, 4 bit transitions are required simultaneously.

15 = 01111 binary

16 = 10000 binary

ill. 8 Interconnection schematics: ( a) TTL buffered output (7404/7406),

( b) RS-442 line driver, ( c) voltage comparator, and ( d) voltage comparator with improved noise immunity.

This situation is distinctly unique to the absolute encoder, as it cannot occur with an incremental device. The binary code is termed polystrophic because of this characteristic of multiple bit changes.

Polystrophism is a problem because in a real-world situation, all these bits won't change simultaneously. There will be some slight ambiguity, for however small a time, which will result in the possibility of the encoder generating incorrect outputs. All of the problems associated with the manufacture of an accurate incremental encoder apply here, compounded by the number of data tracks being implemented. Hysteresis, eccentricities, noise, and so forth can all add up to slight variations. Were the bit error to occur in the most significant bit (MSB), the user could receive a feedback signal that's in error by 18  . Encoder manufacturers have developed specialized scanning methods, called U-scan and V-scan, to orchestrate the transitions of the many bits simultaneously. V-scan uses the least significant bit (LSB) to determine which direction the scale is moving-that is, is the bit transition from high to low or from low to high. The sensors are arranged in two banks, in a V shape, the distribution allowing for tolerances in the system (ill. 9). Once the direction is determined, logic selects the correct side of the V to obtain the reading without transition error.

Although proper design can result in the successful implementation of an absolute encoder using binary encoding, the problem is real enough that many other codes have been developed. The Gray code is a monostrophic code. This is a very popular code which allows only one bit change between any two monotonic values.

Table 3 shows the difference between decimal, binary, and Gray coding. Once the values are read by the computer, they can be readily translated into whatever form is most appropriate.

FIG 9 V scan.

TBL 3 Differences in Monotonic Values Decimal Gray code Binary:

0 0000 0000 1 0001 0001 2 0011 0010 3 0010 0011 4 0110 0100 5 0111 0101 6 0101 0110 7 0100 0111 8 1100 1000 9 1101 1001 10 1111 1010

Note. strophe (from Greek, act of turning; to turn; to twist; action of whirling): The movement of the classical Greek chorus while turning from one side to the other of the orchestra (Webster's Seventh New Collegiate Dictionary, 1971).

Methods of Fabrication. A single-turn absolute encoder can generally be produced with up to 14 bits of position information. 14 bits results in 16,384 unique positions per revolution of the encoder. In many cases, this isn't enough. For a machine tool, where the bed must traverse several feet and the absolute encoder is connected to the lead screw, each rotation is unique, as well as the angle within each rotation.

To accommodate these requirements, multiturn encoders have been developed. Typical multiturn absolute encoders provide 13 or 14 bits per turn, and up to 12 more bits for turn counting. The combination provides up to 26 bits of absolute position data. Even if the resolution per bit were 0.000007 in, this would allow for over 39 ft of absolute position control.

The manner in which turn-counting is implemented determines the cost of the device. The least expensive approach is to use a battery backup for the encoder. The disadvantage of this approach is that, during power loss, the battery must also energize the encoder so that information won't be lost if movement occurs during this event. Because the LED can be a significant drain on the battery, an encoder like this can usually not last more than a few days before power must be restored or information is lost. Many companies have developed ingenious methods to improve the battery life for these devices, and they are widely used throughout the industry.

The most robust multiturn absolutes are built using gearboxes driving additional code wheels for turn counting. By continually gearing down the output shaft, and using this gearing to drive smaller encoders, an additional 12 bits of information can be obtained. The multiple encoder outputs must be carefully combined, using over-lap bits, to ensure that transition errors won't occur. Of course, these devices are complex and require that precision mechanical components work properly. They are available from a number of manufacturers.

Application Considerations. Because of large output word sizes (up to 26 bits), absolute encoder interfaces have developed many interface methods. For word lengths up to 10 bits, parallel interfaces are used. All 10 bits, and sometimes an additional quadrature channel, are provided via direct wiring. For larger word sizes, this isn't practical. For these encoders, there are typically two forms of interface. Since the encoder is used as the primary feedback device during operation, a standard incremental encoder is provided with standard wiring. When used as an absolute reference at power-up, some of the databus system is used to pass the longer digital value over to the main controller. This eliminates the need to handle long cables with many wires. Once the drive has been initialized and begins operation, the incremental encoder interface is used exclusively.

3.3 Linear Optical Encoders

Linear optical encoders are no different from rotary optical encoders. However, their form factor and the way they are used result in some differences in the typical manufacturing processes and end-user handling.

Linear optical encoders are available in lengths from several centimeters to hundreds of meters. In their most basic form, they are comprised of a graduated scale, a read head, and mounting hardware. The read head contains the illumination source, the scanning reticle, and the signal-conditioning electronics. The scale can be made of glass, steel, or plastic. Linear encoders are found in a wide variety of applications, and because of this, there is a need for various types of environmental protection, just as is the case for rotary encoders. However, because the linear encoder must of necessity include a large opening over its entire length for the read head to exit, sealing and protection methods are quite different and not as robust as for rotary encoders. Like rotary encoders, linear encoders have frequency response limitations.

However, these are defined as meters per minute or feet per minute rather than revolutions per minute. Unlike rotary encoders, linear systems are usually found on machine-tool beds and measuring systems, neither of which are normally subjected to ambient temperature extremes. For this reason, they are generally limited to operation over lower temperature ranges. Linear optical encoders are capable of very high resolution, in some cases rivaling that of laser interferometers. They are far more accurate than similar devices using magnetic or inductive systems, as their grating periods can be much smaller and they have superior interpolation accuracy.

Terms

Abbe error Measuring error caused by guideway imperfections and the distance between the tool point and the scale. This results from deviations between the linear scale straight axis and curvatures in the machine tool.

carriage The framework which connects the read head to the scale.

read head The movable portion of the scale containing the signal-conditioning electronics, illumination source, and scanning reticle.

response threshold Error which results from hysteresis and backlash as a result of a directional change.

Principles of Operation. Linear encoders can be manufactured to use either the directed-light principle or the diffracted-light principle. When grating periods of less than 8  m are employed, the diffracted-light method must be utilized. ill 10 depicts a linear encoder scanning mechanism which senses movement by diffraction and interference techniques. Note that only three photodetectors are rewired because of the use of the interference mechanism. As the plane wave of light generated by the collimating lens passes through the transparent scanning reticle, it's diffracted into three directions. At the phase grating of the scale, the light is reflected and diffracted again. The diffracted light returns back through the scanning grating and is diffracted a third time, resulting in three interfering unidirectional light beams. These are collected through a lens and projected onto the photodetectors.

FIG 10 Scanning using diffraction and interference.

Scales using this technique can achieve measuring steps down to a few nanometers and can be as accurate as a laser interferometer when temperature and atmospheric errors are accounted for. It is of interest that although these devices are quite sensitive to angular alignment of the scanning reticle to the scale, they can be relatively insensitive to gap. This isn't true for scales using the directed-light principle, in which gap must be very tightly controlled at small grating pitches or diffraction effects will destroy the signal.

Reference Mark. Most linear encoders contain at least one reference mark. Since some linear scales are quite long, it can be awkward to attempt to find this mark when the system is started or when the power has been lost. To minimize this problem, linear scales sometimes use distance-coded reference marks. In this approach, many reference marks are used. The distance between every other mark is constant, but the distance between any two will vary by a line width. In this way it can be known what section of the scale is in use and how far it's to the last mark, so the position is absolutely determined. This method can reduce the seek motion to 100 mm or less, instead of the entire scale length.

Methods of Fabrication. Linear scales can be glass, metal, metal tape, or Mylar, and can measure over inches or feet.

Scales. Typical grating pitches are 10 and 20  m for encoders using the transmitted-light principle. For protection against contamination, the glass gradua-tion is mounted within an aluminum extrusion, which is sealed to the environment with lip seals. At these grating pitches, the diffraction of light is significant, so it's important to maintain a precise gap and alignment between the carriage and the scale. The mounting of the glass scale to the housing is done with an elastic com-pound so that thermal differences can be accommodated. The housing is mounted firmly to the machine at its midpoint, with elastic blocks at the ends. This also is done to allow for thermal differences between the scale and the machine it's mounted on.

The maximum glass scale length is 3 m in a single piece.

Steel scales can be manufactured in any length, and they are designed to use reflective techniques. Highly reflective gold is plated onto the steel scale, with opaque etched spaces defining the grating pattern. The typical pitch for steel scales is 40  m.

Resolution using interpolation of the 40  m pitch can be as good as 0.2  m (20  ).

For long sections, the scale is supplied in sections which are assembled at the site.

Interferential measuring systems can be made of steel, and a reflective steel phase grating is used to define the graduation. An 8-  m grating results in a 4-  m signal period, which can be interpolated up to 400 times to produce a 0.01-  m measuring step. This is in the same range as a laser interferometer. In some cases this is a better system, because the steel scale is thermally matched to the steel workpiece, so it will better track the machine tool than would a laser interferometer.

Read head. There are two basic methods for controlling gap. One method is mechanically simple, and involves coating the glass scale with a friction-resistant coating. The carriage is then allowed to ride in contact with this coating, allowing very close gaps with good consistency. Because the gap is based on the thickness of this coating, it's very important to apply it in a manner which promotes a constant film thickness. Encoders of this design are termed contacting encoders.

Another method is to use ball-bearing rollers to support the carriage of the read head above the glass scale surface. Although the rollers contact the scale, there is no interaction in the region where light is being transmitted or reflected. For this reason, these are termed noncontacting encoders.

In both cases, angular alignment is provided by ball-bearing rollers riding on the outer edge of the scale.

Light source. The light source is designed to illuminate as large an area as possible so that the photodetectors will average the light and eliminate any problem resulting from contamination or scale imperfections. With a 10-  m pitch, several hundred lines can be averaged to develop the detected signal. Typically, light sources are LED type.

Output Signal Qualities. Output signals are equivalent to what is available for rotary optical encoders, but TTL outputs tend to be only of the line-drive type (RS-422).

Analog outputs are either amplified sine-wave or current outputs of the 11-  A peak-to-peak type.

Accuracy and Resolution. Linear encoders are capable of accurately measuring 1  m/m, [1 part per million (ppm)]. An absolute accuracy of 0.5  m is readily available as well, but not as common. The major source of error in a system using linear encoders is the Abbe error. Abbe error can be compensated for by calibration of the machine after the scale has been installed.

Typical resolution for scales using a 10-  m pitch is 0.25  m, using a 1  interpolation and edge transitions as the measuring step.

Application Considerations. Although there are no real standards, linear encoders have consumer-oriented requirements that become industry standards.

Most scales limit traverse rates to 30 m/min due to frequency-response limitations.

Traverse rate also has an impact on the distance design life, depending on the type of scale. Contacting scales have design lifetimes of  1 million ft. Bearing systems can exceed this, but bearings have lifetimes as well. The user should consult with the manufacturer for this information should this issue need to be addressed.

Flatness of the mounting is very important to preserving accuracy. Because of this, many scales can be significantly more troublesome to use than others. The user should evaluate the bracketry and adjustments available in the scale mounting hard-ware for ease of use and practicality. It will do no good to have a scale capable of 0.5  m accuracy if the installation is good to only 5  m.

(cont. to part 2)