A new type of sensor that provides commutation, velocity, and position feedback may finally unlock the potential of digital drives.
Since the inception of closed-loop servo control, engineers have sought better ways to measure angular position and velocity. The current crop of devices that have sprung up as a result of this search – encoders, resolvers, and tachometers – have met most of the demands over the years. But lately, with the advent of digital drives, sensing technology is emerging as the weak link in the automation chain.
What digital drives need, and what ordinary sensing techniques can't provide, is a device that offers velocity, position, and commutation feedback in a single package small enough to fit, if need be, inside a typical motor housing. Besides being rugged and inexpensive, the device must also be easy to install and connect, self-diagnosing, and able to resolve shaft movements into diminishingly small steps. For the past several years, researchers have been working toward just such a device, and what they've come up with, though it may not be the "ideal" sensor, is better than anything else currently available.
It used to be that controllers were the main factor in determining the performance of an electronic motion system. But now, because of digital drives, motor feedback systems are rapidly becoming the most critical component.
Consider this: To achieve their potential for dynamic acceleration and velocity, digital drives require extremely short sampling intervals. Typical intervals range from 50 to 600 μs. Of course, as sampling times decrease, the number of counts from the feedback device must increase. A drive with a sampling interval of 400 μs, for example, requires a sensor capable of measuring 1.5 million steps/rev, assuming the motor runs at 0.1 rpm.
In the past, it took several sensors to provide commutation, position, and velocity feedback for servoloops, involving almost 20 wires. But now, because digital drives provide a single connection point for all feedback signals, it is not only possible but advantageous to integrate all feedback functions into a single device. Unfortunately, such a device has been hard to come by.
Conventional feedback devices have their strong points, but all come up short in one way or another when it comes to digital drives. On the surface, resolvers seem to be ideal because they combine practically all necessary feedback functions in a single package. However, their accuracy and effect on the control circuit can limit drive potential. For example, some of the benefits of digital speed regulation -- synchronization, dynamic response, and load stiffness -- are next to impossible to achieve with resolvers.
Hall-effect sensors, used for years to provide commutation feedback, aren't the answer either. In fact, they are rapidly falling out of favor in all but the simplest applications because even conventional position and velocity feedback devices have the ability to provide commutation information.
Encoders, though they've become more common in the past few years, have drawbacks too. For one thing, they're limited by resolution. Another problem, particularly with incremental types, is that they require additional lines for commutation feedback.
With absolute encoders, using serial interfaces like SSI, the connections aren't as bad, but the price – absolute types are often two to three times more expensive than incremental types – can be a major obstacle. Also conventional absolutes can’t provide updates quick enough to keep up with dynamic drives.
Better than the rest
Out of the struggle to keep sensing technology on pace with digital drives, sensor designers have come up with a new type of feedback device that combines the advantages of incremental and absolute encoders.
The optical configuration is similar to that of conventional absolute encoders, but instead of the pattern on the glass code disk producing a binary value, the new device produces three pairs of sine/cosine signals with an increasing number of periods (1, 8, and 64). These signals go to an on-board chip that calculates the vectors of each sin/cos pair and chains them together to form a 9-bit signal. The signal is then interpolated another 5 bits to get a final binary value of 14 bits.
Continue on page 2
The new feedback device saves money because it requires fewer tracks and related scanning reticules. And with just one additional outside track, it can generate another channel of sin/cos signals (1-Vpp) with up to 2,048 periods. These additional pulses provide sufficient dynamic feedback for digital drives.
For applications requiring the tracking of multiple shaft revolutions, an optional mechanical gearing system may be added. Here, four gears arranged in series turn four permanent magnets. The two-pole magnetic disks are related by an 8:1 gear ratio. They produce a single pair of sin/cos signals from two Hall-effect sensors.
Upon power up, the custom chip inside the sensor calculates the vector angle of each of the four signal pairs, starting with the coarsest stage. This unambiguously points to the circle of the next stage. The process continues, chaining all vectors into one final absolute value. When the single and multi-turn stages are considered together, an initial resolution of 26 bits is possible.
A special interface that consists of two channels – one for process data and one for parameters – connects the sensor to the drive. The parameter channel is a digital link, such as an RS- 485 interface, and is used to communicate initial absolute values and allow exchange of information between the feedback device and drive. The process data channel is a pair of differential RS-422 channels that transmit sinusoidal signals, which after further evaluation in the drive, are used for the dynamic velocity and position loops.
Initial absolute position is calculated only when the feedback device is turned on, and is typically a 14-bit word for single-turn devices or a 26-bit word for multi-turn devices. This information is transmitted over the digital link to the drive, which uses it for commutation and as a pointer to identify the period of the 1-Vpp sin/cos signal. After it retrieves initial values, the drive then increases the base resolution of the sin/cos signal, interpolating it using an analog-to-digital converter.
Because the drive is involved in evaluating sin/cos signals, the final resolution of the system depends on the drive. In practice, most drives can achieve resolutions between 1 and 8 million counts. By comparison, incremental encoders are limited to about 10,000 counts per turn; the maximum resolution for quadrature output is just slightly higher at 40,000 counts.
It is important to note that, because the signals are sent in a sinusoidal fashion, the bandwidth requirement for the drive input circuit is relatively low. For instance, using conventional digital feedback, a servomotor running at 6,000 rpm, with a resolution of 20,000 counts, requires a connection with a bandwidth of 2 MHz. But with the new system, a motor can run at the same speed, with a much higher resolution, yet the bandwidth requirement drops to somewhere on the order of 50 kHz.
The new feedback system is also more accurate than existing devices. It offers signal linearity of the sin/cos signals to ±5 arc-sec, resulting in total system inaccuracies (including mechanical mounting of the sensor and a/d conversion in the drive) of less than ±30 arc-sec. This is at least ten times better than conventional incremental encoders and resolvers.
Another advantage is that it provides all the feedback signals needed for a drive to operate in a single device that can be integrated into the motor. Current designs are available in package sizes as small as 2.0 in. diameter, permitting integration into size 23 (US) and size 55-mm (European) motor frames.
Moreover, while conventional encoders are rated to 1050C, the new devices are good to 1250C. Thus, it is no longer necessary to derate the continuous torque rating of a motor to protect the encoder. Another important advantage is the possibility of just one drive interface (only eight wires) for all applications, regardless of the required performance.
The mechanical interface also offers advantages. All encoders have two points of coupling; the shaft and body. One of these points must be flexible and the other rigid. In the case of a shaftstyle encoder, the coupling between the shaft and the load must be flexible. This coupling plus the hub and rotor inertia make up a spring-mass system, which resonates, limiting drive stability.
Continue on page 3
Resonance is not a concern with the new feedback device because it employs a stator coupling. This provides for the required flexibility between the encoder and motor. The rotor (shaft) rigidly mounts to the drive shaft of the motor, so there's no chance of slipping or resonance.
Besides the inherent design advantages, stator-coupled feedback systems don't require additional components such as flexible couplings, mounting brackets, and hardware. They also make fewer demands on the concentricity of the motor flange and shaft, and are quick and easy to install.
Moving to digital drives
Despite the advantages of digital drives, analog drives, in one form or other, are still the most common type. Here, the controller sends the speed command to the drive as an analog signal. The drive gets velocity and commutation information from the feedback device, while position information is sent directly to the controller.
One issue with analog drives is that they require feedback signals to be sent to both the drive and controller. The resulting split in feedback loops creates two sets of feedback connections for each axis. This increases the complexity and cost of wiring, and makes it more difficult to tune the system.
Power-block drives, another common type, are in the analog family, though one step closer to digital control. Here, the controller receives commutation, velocity, and position information from the feedback device. It then supplies the power block with a digital speed command and commutation information in the form of analog or PWM signals.
Because feedback signals go to only one location, power-block drive systems are less complex and easier to install. However, since the controller assumes most of the burden for servo loops, these drives impose limits in processing speed and the number of system axes.
The latest drives on the scene are, of course, digital drives. When these devices are connected to the controller over a digital link, such as Sercos or DeviceNet, all feedback loops can be closed in the drive. The drive then passes position information to the controller via the link, and receives speed commands over the same interface.
One advantage of digital drives is that the feedback device, motor, and any secondary devices (such as dual-loop encoders) connect only to the drive, reducing the complexity of system wiring. The drive also assumes many of the burdens previously handled by the controller, allowing the controller to perform additional tasks more expediently, or handle more axes. Digital systems afford one master controller the ability to connect to hundreds of drives.
They also open the door to "smart" sensing systems. By integrating a memory chip in a sensor for digital drives, important data can be stored in the feedback device and accessed over the digital link. An electronic motor label, for example, can be stored during manufacture, then accessed during system initialization to automatically configure the drive. Other information, such as diagnostics and internal device temperature can also be accessed.
A closer look at encoders
Most conventional encoders are alike in that they employ similar optical scanning techniques. Light emitted by an LED is interrupted as it passes through a rotating disk and a second stationary mask. The resulting light flashes illuminate a photodiode array, producing an output according to the pattern on the disk.
In an incremental encoder, the rotating disk has separate tracks for counting and commutation (optional). The number of equally spaced lines around the circumference of the disk (pairs of clear and dark areas) corresponds to the resolution of the encoder.
Absolute encoders use a similar scanning method, but the pattern on the disk is different. The code disk consists of concentric tracks that have precisely placed patterns of clear and dark segments that form a digital value when evaluated together. A separate track is required for each bit of resolution; for instance, a 12-bit (4,096 pulses/rev) encoder would have 12 individual concentric tracks.
Absolute encoders have been more expensive to produce than incremental encoders for several reasons. First, the absolute signals from individual tracks must be physically aligned with neighboring tracks. Second, the light source must be bigger and more complex because of the larger number of tracks. Finally, having to have individual photodiode arrays for each track adds complexity and cost.
Continue on page 4
Time for feedback
One place where the new motor feedback technology is making a difference is in the new 9/440 CNC controller from Allen-Bradley. According to Jay Jeffery, manager of advanced programs for A-B, several sensing technologies, including resolvers and incremental encoders, were reviewed in the process of selecting an appropriate feedback system.
The design objectives called for a device with the following features:
•Automatic adjustment of the controller at setup
•Resolutions exceeding one million counts
•High signal stability in transmission lines up to 100 m
•Motor standardization through a common mechanical package
•Higher accuracy relative to previous feedback technologies
•Multi-turn absolute feedback
The new feedback technology not only meets these criteria, it improves system accuracy by an order of magnitude to ±30 arc-sec. This was confirmed in a Ball-bar test, which indicated a 10-μm improvement over a comparable resolverbased system.
Andy Monnin is with Stegmann Inc., Dayton, Ohio.