Observers improve resolver conversion in motion systems
Resolvers are commonly used in motion-control systems as position sensors. Resolvers work like variable transforms, where the transformation ratio changes with the position of the motor shaft. A resolver works by taking a reference waveform, which is a fixed-amplitude sinusoid, and producing two amplitude-modulated output signals: one encoding the sine of the position and the other, the c...
Resolvers are commonly used in motion-control systems as position sensors. Resolvers work like variable transforms, where the transformation ratio changes with the position of the motor shaft. A resolver works by taking a reference waveform, which is a fixed-amplitude sinusoid, and producing two amplitude-modulated output signals: one encoding the sine of the position and the other, the cosine.
An observer is an algorithm that estimates what a plant (or element being controlled) is doing based on the feedback sensor's output and the excitation applied to the plant. Using the excitation signal, observers can tell us much more about the plant than the feedback device alone.
Observers have been used in complex control systems for years (and taught in universities). Still, many designers may be surprised to find observers can be applied to simple systems and produce impressive results. A relatively simple application of observers is used here to discuss the conversion of signals from resolvers.
The image on the left reflects results with an observer. The observer, used on the right, reduces lag and eliminates ringing.
Resolver output signals must be processed before a motion controller can use them. This is commonly done using a resolver digital converter (RDC) integrated circuit. The RDC chip takes sine morerand cosine signals and produces an output approximating motor position. To minimize noise and simplify the conversion process, most RDCs employ a so-called double-integration method. Here, the difference of the RDC output and the actual position is formed, then driven to zero using a control loop, as the block diagram shows.
RDC is a control system in itself. The PI gains are usually set with resistors and capacitors. The idea is to set these gains high to make the conversion process as fast as possible.
If the designer sets RDC gains too low, the process will inject phase lag into the motion-control loops, causing instability, and ultimately forcing down system gains. However, if gains are set too high, the RDC can become unstable or too noisy, degrading system performance. As with any control system, the final tuning gain values are a compromise and, in most cases, RDC will inject significant phase lag in the motion-control system.
The block diagram shows the signal used to convert a standard RDC to an observer-based converter with dashed lines. This signal adds the expected effects of the torque-producing current. The motor controller measures this current, and anticipates its effect on the system. Then, the observer combines the torque-producing current and resolver signals to estimate the motor's position with much less phase lag than the traditional RDC can provide.
Practice vs. theory
In principle, converting a traditional RDC to an observer is straightforward. The designer needs only to augment the RDC process according to the block diagram. However, there is more to the process, in practice.
First, to include the effects of torque-producing current requires knowing the motor and load inertias, and the motor torque constant. Values of these parameters need be only reasonably accurate (typically,
The real hurdle is that the system structure (block diagram) which amounts to a real-time model of the motor and load cannot be applied to a traditional RDC because the chips do not support such a mechanism. This implies the RDC process needs to be implemented in software as is done in motor controllers, such as Kollmorgen's ServoStar 600.
For products that can take advantage of the observer structure, the benefits are significant. In high-performance systems, the observer allows significant increases in control-system gains.
For example, the first part of the second graphic shows the response of a Kollmorgen ServoStar 600 using traditional RDC. The gains are tuned to levels high enough for a 200-rpm step command to settle in 10 msec. Such high gains induce excessive ringing, and probably would not be acceptable in a practical application. However, if the same system, with identical RDC and control-loop gains, is modified to include the observer, also shown in the second graphic, the phase lag of the traditional RDC and the ringing it produces are eliminated.
The observer makes practical the high gains needed for fast settling. George Ellis, senior scientist at Kollmorgen
For more information, visit www.motionvillage.com/training/handbook/resources/ pcim2001rd_observe.pdf or www.controleng.com/free info . Comments? E-mail: firstname.lastname@example.org