Drives and Controls: Feedback Elements (part 2)

3.4 Magnetic Encoders

Magnetic encoders have many useful features. They have lower power requirements than optical encoders, a simple and robust structure, good performance characteristics, excellent resistance to humid and dirty environments, and they are well suited to large-volume manufacturing techniques. Some of the disadvantages are sensitivity to temperature effects, and generally lower resolution capabilities.

Terms

gauss (G) Unit of magnetic flux density. 0.05 T  500 G.

tesla (T) Unit of magnetic flux density in the SI system of units. 1 T  1 Wb/m 2 .

Principles of Operation. Magnetic encoders utilize a magnetoresistive (MR) sensing element and a magnetic code wheel (ill. 11). The MR sensor changes resistance in the presence of a magnetic field. MR sensors are slightly more sensitive than Hall devices. MR elements can sense fields of approximately 0.005 T (50 G). This is at the low end for Hall devices. This is necessary because flux density is proportional to the pole width, so as line count goes up, density goes down. High-resolution magnetic encoders result in very modest magnetic flux densities.

As the magnetoresistive elements pass through the magnetic field of the code wheel, the material resistance changes by approximately 1.6 percent, depending on the polarity of the field. The sensor is constructed with an aspect ratio that's much wider than it's high. A change in resistance is observed when flux passes across the width, but when the MR element is over a magnetic field that's normal to the conductor, no change in resistance is observed. When connected as shown in ill. 11, if a 5-V potential is applied to the element, the output at phase A or phase B will vary by 0.040 V peak to peak. Because both magnetic poles affect the sensor in the same way but with opposite polarity, when a code wheel of radius r is rotated through an angle tan  1  , the output will complete one electrical cycle.

Methods of Fabrication. Magnetic encoders consist of a sensor element and a code wheel. In many cases, these are provided to the end user as separate components, and the application must provide suitable mounting structures.

The code wheel is often constructed as a cylinder, with magnetic poles recorded on the outer circumference. This allows the sensor to be placed along the outside surface of the drum. This is a major advantage for this type of sensor in that it pro-vides for substantial motor end-play allowance. This is a benefit to motor manufacturers, and will sometimes allow lower-cost motors to be used. The code wheel is generally made by injection molding of an isotropic magnetic material which has been mixed with a plastic carrier, such as nylon or polycarbonate. It can be made in arbitrarily large or small sizes, depending only on the mechanical structure required to support it. For this reason it's well suited for large through-shaft requirements, such as in spindle motors.

ill. 11 Magnetoresistive (MR) sensor operational principles.

The molded code wheel is magnetized in the same way as a recording tape. The wheel position is carefully controlled while a series of magnetic pole pairs are recorded onto the magnetic media. This provides a very flexible manufacturing approach. Any line count can be created by simply recording a different "song." The minimum width that can be recorded is proportional to the coercivity of the magnetic material, so small pole pitches require low-coercivity magnets. Isotropic ferrite is most often used for high resolution.

The MR sensor is made by sputtering a magnetoresistive substance, such as Ni-Fe permalloy, onto an insulating substrate, usually glass. The permalloy is etched to pro-duce a grating structure with spacing appropriate to the resolution of the encoder being fabricated. The film thickness of the MR sensors is on the order of 200 nm (0.2  m). Usually, the grating structure is created so that more than one resolution can be achieved. This is done by connecting the appropriate grating fingers at final assembly. ill 12 shows how one such sensor was implemented. Various resolutions could be supported by soldering connections to the appropriate "fingers" on the etched pattern.

FIG 12 MR sensor.

Output Signal Qualities. Frequency-response capabilities into the megahertz range have been documented (Hunt, 1987) but there are few if any manufacturers printing this on a product data sheet. A magnetic encoder with this performance would have to have a very high resolution to be practical, hence a small gap. In this situation, many of the advantages of the device would be lost. Most manufacturers specify output frequency capabilities of 100 to 200 kHz.

Signals are usually provided in TTL square-wave format. This is because the interaction between the magnetic flux and the MR sensors does not easily yield sinusoidal signals. As a result, they are not easily interpolated, nor used as sinusoidal devices. Typically, the MR outputs are digitized via comparators, as shown in ill. 13.

Accuracy and Resolution. Many manufacturers produce magnetic encoders, Sony being one of the more prominent in the United States . The Sony RE20, RE21, and RE30 encoders support resolutions up to 2048 cpr, and will also provide commutation signals. Special versions are also being fabricated with resolutions of 8192 cpr.

VPL of Topanga, California , has developed an incremental magnetic encoder with resolutions of up to 26,112 cpr, with 402,124 cpr being possible.

Accuracy information isn't readily available. Accuracy can be affected by uneven heating of the MR elements. As current is supplied, the resistive elements heat up. If the heat loading is uneven, the resistance values change unequally, causing the sensor to deviate from its balanced condition. Careful design of the sensor to provide heat-sinking capacity can minimize this issue.

Interpolation. It is quite easy to develop multiplied outputs for this type of encoder. ill 14 shows a proposed method for developing an output for a magnetic encoder. In this method, 16 MR sensors are equally spaced over distance. Although the output frequency is multiplied by a factor of four, the input circuit frequency requirement has not changed. In this application, the author was able to create an encoder capable of developing 10,000 ppr (2500 cpr) and operating up to 700 kHz.

ill. 13 Typical MR sensor signal-conditioning circuit.

Application Considerations. Air gaps must be approximately 80 percent of the pole pitch. For a 40-mm code wheel, 500 ppr, pole pitch is 0.25 mm. A gap of 0.2 mm would suffice. Successful operation with variations of as much as 50 percent has been documented ( Campbell , 1990).

Adequate shielding must be provided, as these devices can be affected by a strong magnetic field. This is particularly important in machine-tool environments where magnetic chucks can be placed in close proximity to the ways where the sensor might be located. It is sometimes more difficult than expected to completely shield these devices from magnetic effects in motor applications. Improper handling of this issue can result in partial or total demagnetization.

One environment in which magnetic encoders work well is one with substantial contamination. Dust, chalk, oil, and other substances don't affect them (Micro Switch, 1982).

3.5 Hall Devices

Hall devices offer the following characteristics (Micro Switch, 1982):

True solid-state construction Long service life (20 billion operations) Reasonable frequency response (100 kHz) Work at zero speed Noncontacting system, Good output interface characteristics, Good operating temperature limits  40 to 15  C)

ill. 14 Interpolation scheme for MR sensor elements.

Principles of Operation. Dr. Edwin Hall discovered the Hall effect in 1879. He was using a foil of gold to study electron flow. When he placed a magnet so that its field was perpendicular to the current flow, a voltage potential was developed across the two edges of the foil.

By placing output connections perpendicular to the direction of current flow, no potential difference exists in the absence of a magnetic field. When a magnetic field is present, current flow is acted on by Lorentz forces, disturbing the current distribution and causing a potential difference across the foil. This is called the Hall volt-age.

A balanced field will produce no output.

Linear sensors have an output voltage characteristic like that shown in ill. 15.

Output varies from 1.5 to 4.5 V as magnetic flux varies, and can be approximated as follows:

Vout  (6.25  10  4 ) V_supply) B  0. 5 Vsupply

where  400  B  400 Digital output devices have on/off characteristics. The on values are related to supply voltages, and can be anything up to 15 V. These devices basically consist of a linear sensor driving a Schmitt trigger, which is simply a comparator with hysteresis.

The effect of hysteresis on the sensor output is shown in ill. 16.

Both unipolar and bipolar digital devices are available. Unipolar devices switch at increasing or decreasing levels of the same pole. Bipolar devices can be selected to have the hysteresis require positive and negative polarities to switch. The response curve for a unipolar Hall element is shown in ill. 17.

The sensor must be brought up to D1 before it will switch on. The sensor must be moved away to point D2 before it will switch off.

ill. 15 Linear Hall sensor output voltage characteristics.

Methods of Fabrication. Hall sensors are entirely solid-state devices. They are mass produced by large manufacturers such as Seimens, Honeywell, and others.

Each of these suppliers has its own method of manufacture.

Magnets, on the other hand, can be purchased, and many times can be optimized, when specified by the end user. The user can implement them in a system in many ways. Unipolar slide-by Hall sensors (ill. 18) are used in low-precision applications and are not suitable for linear sensors. Bipolar slide-by Hall sensors (ill. 19) are more accurate and provide crisp response. Unipolar Hall sensors with pole pieces added (ill. 20) allow gain adjustment. The addition of a pole piece allows actuation at a greater distance ( D4 instead of D3; D2 instead of D1). Bias magnets can allow the addition of an offset to the flux curve, allowing the user to fine-tune it.

Care should be taken to avoid demagnetization if opposing fields are used.

ill. 16 Sensor output hysteresis curve.

ill. 17 Digital unipolar Hall sensor operational characteristics.

Output Signal Qualities. Both linear and digital devices are available. Either may contain internal voltage regulators, or may require regulated voltage inputs. Both linear and digital devices are reasonably slow, with switching times on the order of 1  s.

For linear units, the output loading must not drop below the minimum resistance.

For microswitching this is around 2200  . The output source can be up to 10 mA, and be used with supply voltages from 6 to 16 V dc. The sensitivity for linear sensors is typically on the order of 7 mV/G nominal, with a range of 400 G.

Digital devices are basically the same, but there are sourcing and sinking types.

Sinking types are active low, so when operational, the output drops to 0.4 V maxi-mum.

Sourcing typos are active high, with output voltage V_supply of 1.5 V typical. They can source or sink about 10 mA.

Accuracy and Resolution. These systems are generally used for coarse angular measurement. A typical accuracy value for rotation is between 1 and  mechanical.

Resolution to 1  mechanical is less common but possible. This is basically due to difficulties in getting ring magnets with adequate pole densities, and the reduction of flux density as pole density increases. Accuracy is obtained by operating the Hall device in a steep portion of the flux curve, so that the distance required to cause switching is minimized, and the transitions are crisp. Most applications of this type require a bipolar magnet design. Most devices use internal circuitry to provide temperature stabilization, but the overall accuracy of these devices can be expected to average about 5 percent.

ill. 18 Unipolar slide-by Hall sensor.

Application Considerations. The Hall voltage is defined as follows:

VH  kIc (B sin  ) where Ic is the applied current; B sin  is the component of magnetic field perpendicular to the current path; and k is a function of the geometry of the Hall element, ambient temperature, and the strain placed upon the Hall element. Compensation for temperature can be provided in the device. Typical compensated outputs are 3 percent null shift and 0.077 percent gain shift over a  25 to 8  C range.

In general, systems using Hall sensors can be designed readily using trial and error. If the signal isn't big enough, get a bigger magnet or add a pole piece. How-ever, sometimes the user must specify the magnet and application mechanics, and there are a few rules which can help to support this.

If the magnetic properties are not known, they can be measured using a calibrated Hall sensor. These are devices which are the same size as the part planned for use, but which have calibrated linear outputs that can be used to determine magnetic field strength in the desired application configuration. Once this is known, operating margins can be determined. If a specific magnet needs to be specified, it's best to work closely with a magnet supplier. Here are a few issues which need to be addressed in any magnet design consideration. Magnet selection uses the BH curve (ill. 21) to select between materials. The peak flux available from a magnet, and which must be used to trigger the Hall sensor, depends on the magnetic geometry and the material.

When a magnetic material is magnetized by magnetizing force H, the material moves from the origin to the maximum flux density B_max. At this point, the material is saturated. When H is released, flux density reduces to the residual value B R. The amount of magnetizing force required to then demagnetize the material is the coercive force H c. The shaded region in ill. 21 is the demagnetization curve. Plotting the product of BH in this region against B produces a curve known as the energy product curve, which has an extremum value known as the peak energy product. This value is used to compare magnetic materials. Although the designer can perform the magnetic calculations, and sometimes must, many manufacturers supply charts showing flux density versus distance from the detector for various standard magnet materials and shapes. In most cases, this will be enough to allow determination of design parameters. Handling of magnet structures can also be an issue.

If the magnetic material is purchased already magnetized, handling requirements are more stringent, and the handling of bulk magnets can be quite an issue for the stock room. Accidental demagnetization can result in a number of ways:

Dropping magnetic materials can result in "knock-down" demagnetization.

Placement near uncontrolled strong ac fields can result in demagnetization.

Bulk handling sometimes places magnets into temperature extremes, which may result in loss of magnet integrity.

ill. 19 Bipolar slide-by Hall sensor.

ill. 20 Unipolar Hall sensor with pole piece.

ill. 21 Magnetization curve.

Contact of the magnetic surfaces by ferromagnetic materials (screwdrivers, etc.) can affect flux levels.

Lodex is a material which can be affected by temperatures as low as 10  C. How-ever, for most magnets the threshold is 25  C or above. Still, temperature-related loss of magnetization can account for as much as 5 percent flux change. These changes are nonreversible, requiring remagnetization to correct. Although this may have no impact on a digital sensor application, linear circuits are more susceptible to this problem. Keepers are highly recommended to minimize flux interactions, improve handling, and provide overall structural protection. If large volumes of magnetic material are to be handled, it's recommended that the magnetization be done as part of the manufacturing process. When properly designed, handled, and applied, both linear and digital devices are well suited to high temperature operation, with  40 to 15  C operational limits.

3.6 Linear and Rotary Inductosyns

Inductosyn is a trademark of Farrand Industries, Inc. Both rotary and linear versions are manufactured. Rotary Inductosyns consist of two disks, each carrying radial meander-shaped circuits. Linear versions have a long portion carrying the measurement standard and a slider which moves across it, acting as the output device. They can be manufactured using the same materials as the machine they are to be used on, so thermal mismatch effects can be minimized. This ability to use steel and other engineering materials also makes them mechanically robust and insensitive to contamination.

Although the intrinsic resolution of these devices isn't high, they are readily interpolated to a high factor. Rotary devices generate 2048 poles, and linear products have cycle lengths of 0.01, 0.02, 0.1, and 0.2 in and 2 mm. Interpolation electronics allowing the base resolution to be multiplied by 1000, 2000, or 2048 are readily available. The use of 2048-pole rotary device with a 2048 converter yields  4 million cpr.

ill. 22 Inductosyn operating principles.

Principles of Operation. ill 22 provides a schematic representation of the manner in which the Indoctosyn operates. Electromagnetic coupling between conductors on the slider and the scale causes signal output. The excitation frequency can range from 200 Hz to 200 kHz, but the recommended input is 2 V peak to peak at 10 kHz. With this excitation, an output signal as large as 100 mV can be obtained. The sensor output is sinusoidal and depends on the relative motion between two sets of conductors. Inductive coupling between the two sets of windings is used to measure displacement. An alternating current supplied to the excited conductor induces volt-age in the other which changes as the relative position varies.

Coupling is at a nominal maximum at the position shown. After displacement of one-fourth of the conductor pitch  , the coupling is nulled out. Displacement of another 1/.4  produces full output again, but with reversed polarity to the original position.

Rotary devices obtain an additional benefit by the nature of their design. In this case, the sensor and the excited conductor are the same size. As a result, the sensor is able to simultaneously scan the entire circular pattern of the code disk. This helps to minimize both eccentricity and graduation errors.

Methods of Fabrication. Generally, these devices are manufactured using hot-rolled steel or aluminum. Special designs can use stainless steel or beryllium and can vary in size and thickness. Rotary devices consist of two plates, which must be mounted on a surface prepared by the user. This eliminates coupling errors, but places the burden of assembly and final device accuracy on the skill of the user.

Standard devices are rather large, 3 in being the smallest. Both rotary and linear devices are produced using plate-and-etch techniques, similar to those used in the printed-circuit board industry. Once the base pattern is created on the substrate, the entire device is coated with a shielding material to minimize the effects of stray EMI.

Output Signal Qualities. Inductosyn digital interfaces typically provide excitation for the device and a user-selectable interpolation value, such as 1000 or 2000. These devices are closed-loop servos using null-seeking techniques, and have tracking rate limits from 840 to 3600 in/min at 0.0001 in resolution. The Inductosyn elements behave as transformer windings, with coupling ratio k that's at maximum when the windings are directly adjacent to each other. The sine and cosine windings can be excited by constant amplitude sine and cosine signals, or both can be excited with a common carrier. When sine and cosine excitation is used, the output is a constant amplitude signal that's phase shifted 36  for each movement of the slider by a distance  . When common excitation is used, the output amplitude varies according to the following relationship:

Vout  kVe cos (  pi x/  )

where

X  relative displacement (0 < = x < =  )

Ve  excitation voltage

Vout  output voltage

These devices typically consume approximately 3.75 W of power at 5 V dc.

Accuracy and Resolution. Standard linear products are capable of 0.0002 in (0.005 mm) accuracy, rotary products to  (0.0005  ). Selected units can reduce these numbers by half. Repeatability is actually better than the rated accuracy by a factor of

Bigger is better in terms of accuracy for rotary devices, because error terms are reduced as diameter increases.

Application Considerations. Abbe errors and misalignment errors are probably the biggest error contributors in Inductosyn applications. The next is electrical noise.

Direct attachment to the machine tool provides a major advantage in that it eliminates the possibility of backlash. However, the alignment must be done properly and is nontrivial. For rotary devices, the recommended pattern concentricity requires a TIR of 0.0002 in. The smaller the disk, the larger the impact of eccentricity becomes. Eccentricity error is inversely proportional to the diameter of the disk.

As an example, if the same amount of eccentricity were to be measured on two rotors, one 12-in. in diameter and the other 3-in. in diameter, the error would increase by a factor.

The allowable air gap is generous at 0.0005 to 0.015 in, and the devices will operate well at any value in this range. However, for rotary products, TIR of the surface must be held to less than 10 percent. This can be a challenge for a 12-in disk. Linear products must be mounted on ground surfaces such that the overall gap variation is less than 0.002 in.

To control noise on the output signal, twisted-pair cables with good separation and magnetic shielding sleeves are required. Cable shields must be insulated from each other and carried independently through the harness to be grounded at one common point in the system (Farrand Controls). EMI shielding is normally built into the device, but it must be properly grounded to the machine. Errors can be generated if the sine/cosine excitation is unbalanced or if there is cross-coupling. Shielding, ground loops, stray coupling at terminations, power-supply regulation, or phase-shift problems can all contribute to errors in the output.

3.7 Synchros and Resolvers

The synchro and the resolver are probably the earliest applications of inductive measuring techniques. Synchros resemble three-phase motors, and in fact are in some ways used like them. The torque transmitter (TX) is a synchro that's capable of driving enough current to actually do work. When the output of a TX is supplied to the stator windings of a torque receiver (TR), the rotor of the TR will rotate. The TR is simply a low-impedance synchro being run backward. The rotor of the TR will take a position closely approximating that of the TX in order to balance the current flowing in the loops. The control transmitter (CX) and control transformer (CT) are high-impedance versions of the TX and TR. Now real power is developed in the loop, and the CT simply generates an error voltage.

Resolvers are similar, but they have two stator coils at a 9 0 degree orientation rather than the 12 0 degree configuration of the synchro. The output signal amplitudes vary with the sine and cosine of the rotation angle, as the inductive coupling between rotor and stator coils varies with the angular position of the rotor. Resolvers are absolute-position devices, because the outputs of a simple two-pole resolver complete a full electrical cycle for each mechanical revolution. Additional poles can be added to increase resolution, with 16- or 36-pole configurations being fairly common, but when this is done the device is no longer absolute. The two-speed resolver contains a two-pole and a multipole device on a common housing. This allows the coarse sensor to be used to determine the basic angle, and the fine-speed device to give additional accuracy to the measurement.

The rotary differential transformer is another variation. The stator has three windings, and a magnetic iron core is rotated within it. One winding is excited with alternating current, and the voltage on the secondary windings varies depending on the position of the core.

Resolvers are not available as torque components.

Terms

CDX Control differential transmitter.

CT Control transformer.

CX Control transmitter.

RC Resolver control transformer.

RDC Resolver-to-digital converter.

SDC Synchro-to-digital converter.

TDX Torque differential transmitter.

TR Torque receiver.

TX Torque transmitter or resolver transmitter.

Principles of Operation. Synchros and resolvers both have a rotor which rotates

inside a fixed stator. Schematically, they are constructed as shown in ill. 23.

The three stator windings of the synchro are physically wound 12  apart and are connected in a star fashion. The three terminals are brought directly out. The rotor can be accessed via slip rings and brushes, or it can be excited by an additional circular transformer wound internally at the end of the unit. When the rotor is excited with an ac voltage, there are induced voltages in the stator that are proportional to the sine of the rotor coil axis and the stator coil axis. E.g., if the voltage across R1 and R2 is A sin  t, the voltages at S1, S2, and S3 will be

S1 to S3  A sin  t sin 

S3 to S2  A sin  t sin  12  )

S2 to S1  A sin  t sin  24  ) For a resolver, this relationship would be

S1 to S3  A sin  t sin 

S4 to S2  A sin  t cos 

Actually, since more and more servo systems are going digital, understanding these synchros and resolvers requires knowledge of the digital converter, as well.

These devices are little servos in and of themselves, and the selection and setup of these devices is crucial to the performance of the overall feedback system.

ill. 23 Schematic drawings of (a) synchros and (b) resolvers.

TABLE 4 RDC Update Rates-Output Frequency, kHz Motor speed, rpm | Resolution, cpr (1 128 256 512 1024 2048 4096) 3,000 0.1 6.4 12.8 25.6 51.2 102.4 204.8 4,500 0.1 9.6 19.2 38.4 76.8 153.6 307.2 6,000 0.1 12.8 25.6 51.2 102.4 204.8 409.6 7,500 0.1 16.0 32.0 64.0 128.0 256.0 512.0 9,000 0.2 19.2 38.4 76.8 153.6 307.2 614.4 10,500 0.2 22.4 44.8 89.6 179.2 358.4 716.8 12,000 0.2 25.6 51.2 102.4 204.8 409.6 819.2 13,500 0.2 28.8 57.6 115.2 230.4 460.8 921.6 15,000 0.3 32.0 64.0 128.0 256.0 512.0 1024.0 17,500 0.3 37.3 74.7 149.3 298.7 597.3 1194.7 18,000 0.3 38.4 76.8 153.6 307.2 614.4 1228.8 Resolver-to-digital converters (RDCs) interpolate the resolver output signals and provide 10-, 12-, 14-, and 16-bit results, depending on the converter used. (Some even have programmable word sizes.) When a single-speed resolver is coupled to a 12-bit RDC, a position measurement resolution of 36  /4096  0.087  (5.2  ) is obtained. The frequency at which this conversion must be obtained depends on the speed in rpm of the motor, and Table 4 shows this relationship. E.g., a 12 bit RDC must be able to update a conversion result at the rate of 200 kHz, or 5  s, when the motor is turning at 3000 rpm to avoid data latency. Although most converters can present data without a problem at this rate, usually needing only 1  s to transfer data, the data may not be completely accurate. Most RDCs are tracking converters, which are implemented using a type 2 servo. The type 2 servo is a closed-loop control system which is characterized as having zero error for constant velocity or stationary inputs. Conversely, this type of system will demonstrate errors in all other situations, and the magnitude of these errors must be controlled through optimized tuning of the converter. The frequency response of the converter will play an important part in the overall loop stabilization.

The tracking converter works by multiplying a "guess" angle by the input volt-ages.

If the resolver is at an angle  , then the resultant will be

V sin  t sin  cos and V sin  t cos  sin These can be subtracted from each other, and when the reference voltage is factored out, the remainder is reduced by the trigonometric difference relationship to V sin  t sin  ) This signal is demodulated and sent to a phase-sensitive detector, which results in the generation of an error signal that's proportional to the difference between  and …

This difference is then used to modify the guess, and this is fed back to close the loop. It is the value of that's output as the converter result. Because it's a type 2 system, which means that the error is integrated until it goes away, the guess will reach the exact value quickly and with no error for a constant position or velocity situation.

There will be errors during acceleration, but one isn't normally trying to control this variable anyway.

Methods of Fabrication. Synchros and resolvers are constructed much like a motor. Iron laminations are created for the stator, and the windings are inserted. The rotor is usually made from solid iron, and one or more windings are applied, depending on whether it's to be a CX, CT, CDX, or whatever.

Output Signal Qualities. Amplitude or phase evaluation can be used for interpolation.

By driving the stators with ac signals 9  phase shifted to each other, an ac signal is developed in the rotor which has a phase relationship to the supply voltage that depends on the rotor position.

Accuracy and Resolution. Resolvers have accuracy ratings of 2 to 2  . The corresponding RDC adds an uncertainty of 2 to  1 LSB.

Resolver errors also have both static and dynamic contributions, which result from the acceleration error in the RDC tracking loop, offset voltages that are uncompensated, phase shift between the signals and the reference voltage, and capacitive or inductive crosstalk between the resolver signals and the reference cabling. Noise in the interconnection or on the reference will generate speed-dependent errors proportional to the phase shift in reference and inversely proportional to the reference frequency. Errors can develop if sine/cosine gain is unbalanced, or if there is cross-coupling.

Shielding, ground loop, stray coupling at terminations, power-supply regulation, or phase-shift problems are also resulting errors.

Application Considerations. In addition to the problems just discussed, high-speed operation of resolver systems generally requires higher reference frequencies.

Whether this is an acceptable configuration depends on the system and the application.

For a retrofit or refurbishment situation, it may be that the reference frequency and voltage is already set, thus constraining the upgrade possibilities. In the past, 400- or 60-Hz at 26- or 115-V RMS references have been used extensively, with occasional 1200-Hz applications. Today, it seems that users are moving toward higher frequencies and lower voltages in order to obtain higher tracking rates. The result is that there are many opportunities for low-level modifications in most refurbishment applications, as references are not standard and probably never will be. New drive systems must be able to accommodate this as well, resulting in added complexity.

The fact that the converter itself has dynamics becomes an important part of the system design. Being a type 2 device, the converter can introduce up to 18  of phase lag into the system. For a 12-bit converter using a 400-Hz reference (Analog Devices AD2S80 A), the RDC bandwidth  3 dB point) will be less than 100 Hz. Using the same reference, a 14-bit converter will have a bandwidth of 66 Hz, and a 16-bit converter will have a bandwidth of 53 Hz. A 100-Hz  3-dB bandwidth means that there will be approximately 3 dB of peaking and 4  of phase shift at 40 Hz. As many servos attempt to close position loops near these frequencies, and an added 4  phase shift would be undesirable, it should also be noted that although the RDC tracking rate may not be exceeded, a system with difficult load dynamics could well prove unstable when the RDC dynamics are introduced. The situation only worsens when 14- or 16-bit converters are utilized.

3.8 DC Tachometers

Tachometers are used as velocity feedback sensors on speed-control systems. These devices generate an electrical signal which is proportional to the angular velocity of the motor shaft. They are used for monitoring open-loop systems and as the primary feedback element in velocity control systems. As shown earlier, they can also be used for inner-loop stabilization in position-control systems, of which there are three basic types, iron core, moving coil, and brushless, the most predominant type being the iron core.

Terms:

dc generator Used interchangeably with dc tachometer.

Kg Tachometer voltage sensitivity, V/krpm

Ripple Noise voltage which can be assumed to be superimposed upon a linear output.

Principles of Operation. A tachometer is the opposite of a motor. A motor converts electrical energy into motion, and a tachometer converts mechanical motion into electrical energy. The output of the tachometer can be modeled as a main signal and the ripple component. The main signal is directly proportional to the rotor angular velocity, much like the counter-emf generated by a motor. The constant defining this proportionality is termed the gain K g. Typical values for Kg range from 0.3 to 25 V/krpm.

Methods of Fabrication. Iron-core tachometers are made using rotor laminations, much like a motor. The rotor for a tachometer has more slots than a rotor for a motor would. Because of this use of iron, these devices have significant inertia.

The moving-coil tachometer is a wound-coil rotor with magnets on the stator.

These have very low inertia. For both iron-core and moving-coil tachometers, the current generated in the individual windings is routed to the output terminals via a commutator and brushes. Once again, this is similar in a fashion to a motor, but the current flow is in the opposite direction. The current flows out of the tachometer, and into the motor. Brushless tachometers are like iron-core devices, but instead of brushes they use optical or magnetic commutation circuits for current steering.

Output Signal Qualities. Tachometer outputs are bipolar, and are positive for one direction and negative for the other. A perfect device would have a completely linear relationship between rpm and output voltage, with zero ripple. This is never the case, however, and voltage ripple defines the quality of the output. The output volt-age sensitivity is affected by the load the tachometer sees.

The tachometer output is dependent upon the load resistor value by

K  RGI = RLI

Vout  ( RLK  )/RL  RG RL also affects the current in the circuit, shown in ill. 24.

ill. 24 Tachometer circuit.

Accuracy and Resolution. The accuracy of a tachometer is determined by its linearity, ripple voltage, and temperature stability. Accuracy values needed vary depending upon the application.

For the paper and pulp industry, accuracy values of 0.03 percent are standard.

Linearity of 0.5 percent is normal up to 3000 rpm. At higher velocities, nonlinearities become more apparent.

This can result from brush bounce, commutator eccentricity, and brush skew due to directional changes at high speeds. Linearity is also affected by eddy current and hysteresis losses in the armature due to shorting during switching.

A good value for voltage ripple is 1 percent. With ripple values this small, small-order effects like shaft eccentricity can be seen if present. Levels this low are generally achieved only by moving-coil types because of their very low inductance. Ripple is composed mainly of noise created by brush transition between commutator segments.

The signal is periodic and is related to the number of commutation segments and the shaft velocity. The frequency of this second-order contributor to ripple is armature eccentricity. It causes low-frequency amplitude variations of the same frequency as the rotor rpm. Finally, inductive effects in the windings can also affect the tachometer output. This contribution is generally very high frequency, however, and can be easily filtered out.

Temperature stability can also have some effect on tachometer performance. A thermal stability of 0.01 percent per degree Celsius is the best available, while the low-end commercial grade can be as poor as 0.2 percent. For a particular application, look at the temperature differential expected during operation. If the servo needs to maintain 2 percent of set point over this range, then 2 percent divided by range equals the percent stability required. Choose a tachometer with a temperature stability rating that's better than this value.

Application Considerations. In order to eliminate backlash, tachometers are generally mounted directly onto the motor shaft. In some cases, the tachometer is actually built right into the motor. This manufacturing approach is cost effective, but leads to coupling between the motor and the tachometer due to interaction between their magnetic fields. The electromagnetic coupling between the motor and the tachometer will be stronger at higher frequencies, so it's a sort of high-pass filter. The phase relationship of the electromagnetic coupling with respect to the motor voltage depends on the angular orientation of the tachometer to the motor. If the tachometer is improperly aligned to the motor, the combination of the high-pass filter characteristics and an improper phase relationship can result in instabilities in the motor/tachometer outputs even when used in an open-loop sys-tem.

Many manufacturers provide motors with integral tachometers, and when properly built, they can be used for systems with servo bandwidths in the 15- to 30 Hz region. However, above this frequency, magnetic coupling between the tachometer and the motor won't be manageable, and the two devices should be separated.

Tachometers are not designed to provide any significant output power. In order to maintain commutation quality, they should be terminated into a load resistor which will maintain current levels on the order of 1 mA. If linearity is the prime consideration, the RL should be chosen to be at least 100 times the dc resistance of the tachometer. If the intended application will run at very low speeds, such that this will be difficult to maintain, a silver commutator should be used.