In a perfect world, you would just install a servo motor, turn it on, and everything would work perfectly. We don’t live in a perfect world, however, and when a newly installed servo motor gets its first command, it rarely winds up at the exact position demanded.
As much an art as a science, manual servo-motor tuning can be laborious and time-consuming. New automated tuning algorithms have taken much of the work out of the process, however, and provide a good solution for most applications.
The actual process entails placing a load on the system and sending a command to the motor, then monitoring the resultant output signal from the motor and the feedback signal from the encoder. Essentially, it’s a process of comparing the position demanded versus the position measured. Once that raw tuning at the drive level is taken care of, the proceedings shift to the controller and fine-tuning the performance of the PID loop.
Tuning is an iterative process, and as you might expect, can be time consuming for even one axis. It’s performed serially, one axis after another, and there can be interdependence. “You tune one and it will affect the others,” says Bill Leang, Manager of Motion Engineering for Yaskawa Electric America Inc. (Waukegan, Illinois). “Depending on the machine, once everything is tuned up and you look at the whole system, you may have to go back to some of the axes and fine tune them a little bit.” For a few axes, that could be complex; for a packaging machine with, say, 30 axes, manual tuning could take days.
Adding to the challenge is the fact that manual tuning requires a direct connection to the drive being tuned. Back in the day, that meant hooking up to a test point on the board itself, no matter where that board was. With today’s networked systems, technicians and integrators have the far easier option of linking into the system to tune every drive/servo motor on it from one position.
Loads and Damping
The problem is that tuning for the best response and minimizing settling time can cause vibrations if the machine chassis isn’t sufficiently rigid - vibrations that can cause or exacerbate oscillations. Such oscillation, or dithering, can compromise accuracy -- or worse. “If it’s sitting there at zero position and you hear that oscillation, you’re putting out a current to that motor,” says Bell. “If it’s overtuned or you have a lot of lash in the system, that motor’s sitting there fighting itself and it’s going to heat up.” Eventually, the drive will register an overcurrent error message and trip out.
In applications demanding high accuracy, such as nanotechnology or semiconductor metrology, the process is essentially to overtune the servo motor and then back off gradually to reach peak performance. “You can see it by looking at the waveform and also by feel,” says Leang. “You put your hand on the table and you can feel whether there’s any vibration. You can hear an audible sound and you sometimes tune it that way. We have different filters and things like that to try to dampen it.”
The load on the system is key. An unloaded motor can wind up dithering about the commanded position; the load adds the inertia necessary to damp that process. Tuning should take place with a load at or near that actually expected for the application. Of course, how a load couples to the servo motor can also have an effect, for example, if a screw has significant backlash or a belt has a lot of stretch. “If you do not have a proper low-backlash gearbox or a servo-rated coupler, it can actually amplify your dithering,” says Bell. “Your motor will be moving, trying to hold that position, but when it hits that load going back and forth it will start oscillating larger and larger until eventually the drive will trip out with an overcurrent error.” Not surprisingly, a change in actuators requires retuning the motor for best results.
Linear motors, which often lack belts and screws, present special tuning challenges, especially for users accustomed to rotary servo motor/actuator combinations. “Those systems have certain inherent features,” says Ben Furnish, Linear Product Manager at Parker Hannifin (Irwin, Pennsylvania). “The screw or belt drive mechanisms actually damp out some of the servo motor response. When you get a linear motor system, especially if it has air bearings, it’s much more difficult to tune because you don't have any of that extra damping. At that point, you’re strictly tuning the servo loop versus the load and it can be much tougher as the motor dithers more freely around the encoder counts.”
Once a servo motor is tuned, the load can be changed but only to a certain degree. “You could end up with issues there,” says Bell. “It really depends on your mechanical advantage, whether you have a gearbox, or how you’re sized. If you change loads, you want to stay within a 10:1 ratio of rotor to load inertia. If you get outside of 10:1, it makes it hard to tune in the first place. If you change your load you want to be sure you stay within that ratio.”
Best of all, a process that once took days of intensive work can be accomplished in minutes.
“Based on the feedback and the load change, which is the current to the motor, we have an algorithm that adjusts different parameters in the servo drive to get optimal performance,” says Leang. “The microprocessors are fast enough to process the information. The old way, you adjusted potentiometers -- typically the proportional, the integral, and the derivative -- and then maybe a couple of filters on top of that. When we went to digital servos, it become a parameter you changed by typing in a number. That’s all incorporated in an algorithm.”
The upside of autotuning is that it is fast and simple, yet effective enough for general use. The downside is that it still can’t quite achieve the performance of manual tuning. For applications requiring higher accuracy, users can run the autotuning routine for raw tuning, followed by manual adjustment; a hybrid format called guided autotuning offers better performance still. “You just tweak it, because even with the smartest application, you want to go back and make sure that you’re running to your peak efficiency,” Bell says.
In general, there’s not much of a cost tradeoff. The hardware incorporated in most components today can easily accommodate the autotuning algorithms.
As for market response, most customers are happy to have less to do. “I think people generally like it,” says Leang. “Some of the experienced guys might still want to do manual tuning but most people, in a lot of applications, just leave autotuning set as the default out of the box, put it on the machine and run with it.”