Have you heard of
EnDat yet?

If not click here!

 
 

Rotary Encoders Make Digital Drives Dynamic

Rainer Hagl
Steffen Beilski
Dr. Johannes Heidenhain GmbH
Traunreut, Germany

Examine the control circuitry of a traditional NC machine and you are likely to find that a microprocessor is controlling the position loop. However, the velocity loop is probably an analog circuit, receiving velocity feedback from a tachometer. Commutation signals for the brushless motor comes from Hall Sensors.

Newer systems, however, are making a transition to a fully digital approach. The microprocessor takes over both speed control and electronic commutation, as well as position control. Here, it is desirable to minimize the number of lines running between the motor and the control electronics. This provides a motivation to let the encoder provide position, speed and commutation data, rather than using separate subsystems for each.

Moreover, it is desirable to keep the encoder size as small as possible. Space in drive systems comes at a premium, making compact design and simple mounting important.

Not all rotary encoders measure up to the stringent demands of fully digital drives. A review of the system requirements shows what qualities to look for in encoders that must provide feedback in these complicated controls (Photo1).

The fast acceleration and high bandwidth demanded of modern drives dictates a need for sampling times of at least once every 200 to 600 µseconds. The controller must be able to process each sample in less than 25% of the sampling time, or between 50 and 150 µseconds. To minimize control-loop dead time, speed and position values should go to the control system with as little delay as possible, no more than a few microseconds.


Photo 1: Built-in rotary encoders are compact and come in styles that include incremental versions producing commutation signals, as well as single-turn or multi-turn absolute encoders which, in addition, generate incremental signals for digital speed control. One benefit of built-in models is that an already installed incremental encoder of the ERN 1300 series can be replaced without mechanical changes by a multiturn encoder to get an absolute axis. The stator coupling compensates axial misalignment of +/- 0.5 mm and radial misalignment of +/- 0.2mm incremental signals are typically 2,048 periods/rev at 1 V p-p. Single and multiturn absolute models feature synchronous serial data output. 

Digital speed control also demands position encoders with high resolution. Assuming a 500 µsec sampling time and that the encoder measures at least one minimum increment of motor shaft rotation for each sampling cycle, it must generate 1.2 million measuring steps/rev. or more to handle a speed of .1 rpm. Even higher resolution is needed in applications demanding lower speeds, such as the electronic synchronization of machine axes.

Another requirement levied by brushless dc drives is the need for the absolute value of motor rotor position on start up, or after a power interruption. Electronic commutation demands this information. However, it is impractical to obtain it in parallel from absolute encoders. The problem is the large number of lines (more than 20 for the 20-bit resolution producing 1 million steps/rev) running from the encoder to subsequent electronics.

The only alternative is to employ absolute encoders, but obtain position information from then serially. Of course, serial transmission causes other concerns relating to control loop dead time. For example, at a clock frequency of 500 Khz (which limits cable length to under 350 ft.), it takes over 50 µseconds to transmit 20 bits serially. This constitutes a delay that violates sampling time minimums. The way around this dilemma is to generate incremental signals as well as the serially transmitted absolute values.

However, it is advantageous to generate these incremental signals in the form of sine waves rather than square waves. The reason relates to maximum drive speeds. Consider a motor turning at 6,000 rpm and generating 500,000 measuring steps/rev. In the case of square-wave outputs, the minimum edge separation is less than 0.02 µsec. This corresponds to a frequency of 12.5 MHz, which cannot be processed by ordinary TTL-compatible input circuitry. And signals in this frequency range cannot travel over long cables. Such demands are destined to get worse. Drive manufacturers are striving toward maximum feed-drive speeds of 9,000 rpm and higher. Drives hitting 12,000 rpm are expected in the future.

Incremental rotary encoders are normally available with line counts to 5,000. For speeds anticipated in the future, line counts between 1,000 and 2,500 represent a compromise between high accuracy and the need to keep output frequency low to avoid restricting cable length (Figure 1).

With these constraints, 1-V peak to peak sinusoids are nearly ideal encoder signals. Encoders generating such signals have a -3db amplitude roll off at 240KHz. At these levels, cable length can be as long as 500 ft. (in contrast, encoders producing typical signal levels of 11 µAmps peak to peak work with cable lengths of 100 ft.).



 

Figure 1: Incremental rotary encoders, typically provide accuracies of +/- 1/20 of a grating period. For example, a 2,000 line encoder gives an accuracy of about +/- 30 arc seconds (at temperatures to 20° C). This allows them to serve in applications such as textile printing where the feed rate of material must synchronize closely with the printing cylinders to keep colors in registration. For comparison, resolvers with bearings have typical accuracies of +/- 8 arc min., and so-called pancake resolver accuracies are typically +/- 20 arc min., making them impractical for such uses.

A cross section of a HEIDENHAIN built-in rotary encoder illustrates some of the features that contribute to high accuracy. Stator couplings rigidly connect the motor shaft to the encoder rotor, often through the generally preferred conical seat approach. This sort of coupling boosts encoder natural frequencies to about 2 KHz, three to four times that of encoders with separate couplings. The stator-type coupling also helps compensate for deviations in eccentricity between the encoder flange and motor shaft. 


To provide enough resolution for digital speed control, signal conditioning circuitry must electronically subdivide, or interpolate, the period of the input signals. For example, the one million step/rev resolution necessary for digital speed control can come from subdividing each sine wave of a 2,500-line encoder by a factor of 512. Subdivisions of 1,024 or even 4,096 are possible to provide resolution exceeding 10 million steps/rev.

In addition to generating serial position signals, encoders must feedback the absolute value of the rotor position for commutating the dc brushless motor. Absolute encoders can provide this information. Absolute encoders in digital drives generally transmit data back to drive electronics using a 25 bit synchronous serial format. This format has become a standard way of minimizing the number of lines running between the encoder and the signal processing electronics. Most commercially available signal processing electronics support it.

Use of a synchronous serial interface and an enhanced data protocol allows the encoder to send data to the drive amplifier. For example, HEIDENHAIN model ECN 1313 and 1325 encoders use such a protocol and an internal EEPROM to store and send motor data, such as maximum torque, power, current and so forth, to set up and optimize the drive. This eliminates the need for a skilled engineer to tweak drive parameters.

Another way of commutating brushless motors is to use incremental encoders that contain commutation tracks. These tracks generate either sine waves or square waves, depending on the type of motor commutation.

Rotary encoders used in analog speed control typically deliver square wave commutation signals. Square waves are employed for compatibility with downstream PLCs and NCs which typically can handle only square-wave inputs. Commutation signals in this instance consist of three square waves phase shifted by 120° degrees. This usually corresponds to a mechanical phase shift of 40° or 60° degrees.

On the other hand, there are cost optimized rotary encoders designed specifically for dc brushless motors with sinusoidal commutation. These provide sine and cosine signals with one period per revolution. This permits rotor position to be determined to an accuracy of +/-5° degrees at switch on, normally sufficient for starting brushless motors. After the motor shaft rotates past the reference point, commutation accuracy improves to within a few arc seconds.

One benefit of generating sinusoidal signals for both position and commutation is that the same interpolation circuitry in the subsequent electronics can handle both signals. This simplifies the input processing.

Rotary encoders employ several techniques to ensure high-accuracy outputs. These techniques concern the use of rigid couplings between the encoder and motor shaft, compensation for high temperatures, and optical filtering of the signal shape.

Use of interpolation makes it imperative that the sinusoidal outputs encoders generate be clean and of high quality. One way of guaranteeing such quality is through use of optical filtering in the encoder. This filtering consists of performing an integration on the signals received from a large number of grating lines through the scanning reticle and encoder wheel. Accuracy improves becasue integration averages out small imperfections in the gratings. The signal shape (sine and cosine) is optimized by using a special scanning method to eliminate signal components caused by different light. Encoder optics can see temperatures between 100 and 120 degrees C. Consequently, they must carry specially designed LED light sources than can withstand such temperatures. LED intensity is also regulated to compensate for a high-temperature fall off in sensitivity of the photovoltaic cells used in the encoder. These cells are selected so they have nearly identical operating qualities over temperature. This also promotes high accuracy in the down-stream interpolation of the encoder signals.

Another factor crucial for accuracy is rigid coupling between the encoder and motor shaft. The conventional way of connecting an encoder to a motor shaft is with a coupling located on the rotor-side of the encoder. This configuration is unsuitable for digital drives. The problem is that even the best coupling acts as a torsional spring which, in conjunction with the moment of inertia of the encoder rotor, comprises a spring mass system. Upon reaching its natural frequency, this design limits drive stability.

The better alternative is to locate the coupling on the stator side of the encoder between the scanning unit and the encoder housing. The stator-side coupling is not strained during acceleration. The encoder rotor can mount rigidly to the motor shaft. The stator-side coupling sees only the low frictional torque of the bearings. A spring parallelogram coupling compensates axial motion and radial misalignment between the motor shaft and the encoder. Rotor couplings can also compensate for these effects, but can induce measuring errors of up to +/- 40° degrees in the process.

Similarly, dc brushless motors can be equipped with built-in rotary encoders having a rigid connection between the motor shaft and the encoder rotor. These also contain the compensating spring parallelogram. The encoder connects with the motor shaft either through a conical seat or through cylindrical clamping. The conical seat is more rigid and is thus the preferred approach.

Built-in rotary encoders have mechanical natural frequencies of approximately 2 KHz, which is three to four times higher than those of rotary encoders with separate couplings. The result is good dynamic behavior during high angular acceleration of the motor shaft.

Encoders that are built into the motor housing are also inherently shorter than those with separate couplings. And they are simpler to mount. For example, encoder conical shafts simply insert in an internal cone of the motor shaft and are fixed by a screw from the back of the encoder housing. The mounting ring can be rotated on the stationary shaft to the position at which a reference signal should be generated. Then the mounting ring secures by screws in fixing clamps or a tensioning ring.

Built-in rotary encoders have no need for an additional adapter flange. Moreover, it is significantly simpler to adjust the position of the reference signal output than for encoders with separate couplings. This reduces mounting time.

Sizing up encoder signals:

Incremental encoders generate two 90 degree phase-shifted pulse trains which are either sine (Figure 2) or square wave (Figure 3), and a reference mark once per revolution. Some versions also generate commutation signals for use with dc brushless motors. These are typically shifted 40° or 60° degrees depending on the number of motor pole pairs.

Rotary encoders are designed specifically for dc brushless motors can provide both incremental and absolute data. Sine and cosine signals with a period of once per revolution permit rotor position to be determined to an accuracy of +/- 5° degrees at switch on. An example is the ERN 1387, which is employed with specially developed electronics for interpolation and other signal processing.


Figures 2 and 3: Output Signals Of Incremental Rotary Encoders


Figure 4: Output Signal Of Absolute Rotary Encoders

Graphs of motor shaft speed versus encoder output frequencies show how the number of lines on the encoder wheel affects encoder output frequency, which tops out at around 1 MHz for a 4,096 count encoder spinning at 18000 rpm. Digital interpolation can boost the frequency of encoder outputs to increase the effective resolution of the system. For example, applying the 4,096 fold interpolation to a 5,000 line encoder signal produces about a 2-million count signal for processing by down stream electronics.

In contrast to incremental models, absolute multiturn rotary encoders typically produce 25-bit code values for a given increment of rotation (Figure 4). These values are then encoded into 25-bit words that are clocked serially to signal processing electronics. In digital drives, absolute encoders must also generate incremental signals for speed control.

 

© 1999-2004 Automation and Metrology, Inc.  All rights reserved.

Copyright © 2005 Optimized by Cherryoneweb.com Website Optimized by: Cherryoneweb.com