DAC102021

44 October 2021 www.drivesncontrols.com create a velocity loop. Around the velocity loop is a position loop. It is possible to feed the trajectory into the acceleration and velocity feed-forwards, then add that into the system current node. However, there are other ways to insert feed-forward, such as adding the differentiated trajectory to the velocity loop. In short, the technician can run a frequency response of the system, plot the inverse plant, and overlay the inverse feed-forward on the plant. Then, he or she can pick a gain that best matches the inertia line. Of course, there is one important thing to remember when increasing the gain: expect to see the current slope fall because it is the inverse of the feed- forward on the plot. Another way to achieve the same outcome is to tune with a closed-loop response, while a third method involves looking at the feed- forward current. Timing is everything Having determined where to put the feed- forward command and how to calculate the gains, the next task is selecting the optimum time for insertion. This step is critical to a successful outcome because failing to add the feed-forward at the correct time means the control effort will not be in phase with the system. Most modern plants take advantage of digital controllers, which typically allow technicians to insert the feed-forward at different sample times in the system. This strategy will help the feed-forward to be more in or out of phase with that of the plant. With digital control systems, optimising the timing of when to add the feed-forward command to the control effort will account for computation time and further minimise overshoot. Although usually only a small effect, it could make the difference in optimising the feed- forward gains and structure. To highlight the effect of feed-forward on a point-to-point move, consider the simple example of moving a 7kg mass mounted vertically on a linear axis (total moving weight 21kg) over a distance of 100mm at 0.5g acceleration. Here, important factors include peak overshoot, move and settle time, and rms (root mean square) tracking error. The first step is to make a standard trapezoidal move (no feed-forward) and look at the position/following error which, in this instance, happens to be 15.70μm rms. Then, start applying other methods to make improvements. For instance, entering a manually calculated value of acceleration feed- forward (Aff) – in this case, 629 – reduces the position error to 6.21μm rms, with the move and settle time now at 342ms (down from 350ms). Applying the controller’s autotune function makes a further improvement in position error to 3.74μm rms (324ms move and settle time). Aerotech offers a function called EasyTune that uses a special algorithm to look for optimal error, lowering the position error to 3.33μm rms (336ms move and settle time). For further refinement, a little handcrafting can deliver results. In this example, entering an Aff of 471 and velocity feed-forward (Vff) of 2, produces a position error of 2.47μm rms (325ms move and settle time). In any servo tuning exercise it is easy to ignore non-linearities, but this is a mistake. Among the common ones is the pulse-width modulation (PWM) amplifier, a drive widely used in automation projects. Most think that the output current is some scale factor of current input to the amplifier, or at best consider it a first-order system (as it is not really a straight gain). However, PWM amplifiers do not look like that in reality. At low currents, the PWM gain is substantially lower than at high currents. This results in a lowering of the overall loop gain and extends settle times when performing point-to-point moves. One solution is to measure the nominal current loop transmission and find a standard set of stable gains. Then, the technician can measure the low-current loop transmission and find an alternate gain set (which will be higher). After this task, he or she can monitor the current level in the system and, when it falls below a certain level, use the alternate gain set. Notably, it is viable to reduce the settle time by increasing the gain as the current reduces. Picking the right point is critical. Typically, this task will involve running a series of loop transmissions, looking at the open loop, and using different values of current through the system. As the amount of current put through the amplifier decreases, the change in gain will become clearly visible. The area of most distinct change will indicate the optimum point. The friction factor Another non-linearity to overcome is friction. This is often evident on Bode plots, where at low frequencies there is a clear flattening of the curve, indicating friction in the axis. To help overcome this problem, a simple answer is to relinquish phase margin in favour of increasing the low-frequency gain. Adopting such an approach mitigates the friction effect and minimises settle time. Another method is to use enhanced tracking control – in effect, a different control structure – to increase low-frequency gain. This structure type at low frequency recovers the rigid body (inertia) line, producing a much higher low-frequency gain and minimising the settle time. Furthermore, tracking error reduces in complex contour motion. A final factor to consider is the sensitivity frequency response of the system, which can help to determine stability and tracking performance. High sensitivity indicates the system is likely to become unstable at that frequency. Ultimately, the use of feed-forward presents many options to help optimise the tuning of servo systems. However, while many of these methods can prove highly effective, there is always more to consider, such as sensor resolution and servo sample time – and their effects on system performance – but that is for another day. n n PRECISION ENGINEERING AND MOTION CONTROL Modern motion control systems deliver precision automation and optimised throughput, with user-friendly interfaces