The following discussion will focus on how to implement advanced motion control technology to improve the performance of laser micro-drilling machines.
By William Yeh and Jerry Lin, Aerotech Taiwan
In micro-drilling process equipment, the most popular ways to create microvias (blind or buried vias with a diameter ≤150 µm) are mechanical drilling and laser drilling. Mechanical drilling is mainly used in larger apertures such as PCBs or IC substrates. Laser drilling is mainly used in finer apertures such as silicon wafers, ceramic substrates, and sapphire substrates. Although many of the requirements of mechanical and laser drilling are identical, this discussion will focus on how advanced motion control technology plays a critical role in the success of laser-drilling process equipment. Not only do we need to “tune” the axes to attain the highest possible process capacity and yield but, most importantly, we need maximum stability so that we operate without resonance or vibrations to ensure maximum aperture consistency.
Figure 1. A laser drilling machine using Aerotech Nmark AGV-HPO open-frame galvo scanners.
Laser micro-drilling applications include three common requirements: (1) ensure that the XY servo stages are “tuned” to be fast, stable, and accurate; (2) ensure that the roundness of the apertures meets the customer requirement; and (3) because the quantity of apertures is normally large, ensure the highest possible program execution efficiency.
Tuning Servo Stages
Part dimensions, mass, and the dynamic performance requirement will all influence the difficulty of XY servo stage system tuning. For small parts like silicon wafers or ceramic substrates, if the move and settle time requirements are not extremely stringent, the tuning process could be completed fairly easily. However, large substrates that require very high dynamic performance from the servo stages are common, and this will be the bottleneck for many machine makers to adjust the servo gains to meet the process requirements.
Normally a motion system control will provide an autotuning routine (Figure 2), or a “step response” tuning technique that is mainly done in the Time Domain. This type of tuning process could be effective in easier applications. However, because Time Domain tuning techniques cannot estimate the resonance, the pole zero map, or set-up the filters, many times the user has to reduce the servo gains to avoid system oscillation. Unfortunately, many times reducing the gains also means degraded throughput and process yield.
Figure 2. Using an autotune routine usually cannot provide the “ideal” servo gains.
Using frequency domain techniques (frequency response analysis) is a more advanced tuning approach and will yield better throughput. The frequency response analysis technique injects a sinusoidal signal from low to high frequency into the motor, and captures the phase and gains at each frequency. The phase margin and gain margin are then evaluated in order to make adjustments to the servo gains. High performance motion controllers, like the Aerotech Automation 3200, can easily shape the loop and adjust the Bode plot graphically, easily set up the servo filters, and ultimately increase the gains to maximize the throughput (Figure 3).
Figure 3. Using loop transmission to optimize the servo gains.
Improving the Roundness
Many laser micro-drilling applications require the best possible roundness. In mechanical-drilling applications, because the tool will “punch” the substrate, roundness is less dependent on the motion control system. But in laser drilling applications, where the micro apertures are mainly “routed,” the roundness is more dependent on the laser spot size deviation and, therefore, the following error of the motion system.
First, the user will need a graphical interface to be able to display the feedback signal (from linear encoder or the encoder from the galvo scanners) on the PC, based on different process parameters, to optimize the throughput and the quality. The common process parameters are: acceleration, speed, and radius of the aperture. When the motion system is routing the aperture, the acceleration can be calculated as:
Therefore, the smaller the aperture or higher the process speed will yield higher acceleration. However, higher acceleration of the motion system will result in higher following error, which means the roundness of the aperture will be reduced. If the user needs to improve the quality of the aperture by reducing the acceleration, the throughput will be impacted.
Figure 4. The acceleration is proportional to the following error.
In order to improve the roundness without significantly impacting the throughput, the Aerotech A3200 provides two features: (1) acceleration limiting and (2) Enhanced Tracking Control. Advanced motion controllers like the Aerotech A3200 can limit the coordinated circular acceleration (acceleration limiting) while still maximizing linear acceleration. This will increase the roundness without appreciably affecting throughput. The Enhanced Tracking Control (ETC) feature actually reduces the following error of the circles. For example, Aerotech’s ETC uses advanced algorithms that can increase low frequency gains, but without changing the higher frequency gain. This greatly improves following error at direction reversals, which inherently include high friction, thereby improving the roundness of the apertures.
Figure 5. Using Enhanced Tracking Control functionality greatly improves the roundness of the routing motion.
Program Execution Efficiency
Advanced motion controllers have many special features that allow the large quantity of aperture coordinates to be processed at the highest possible efficiency. For example, Aerotech’s Look Ahead feature can determine the trajectory of upcoming apertures. The Queue mode allows the data to be processed in the First-In, First-Out fashion and, therefore, the embedded memory size does not restrict how many points you can actually process. Although the performance benefits of these features are not as obvious as those previously discussed, these are still critical factors to the overall success of laser micro-drilling processing machines.
The following example demonstrates the process of optimizing a laser micro-drilling system.
When a user found the process yield was low and the roundness was not ideal from their machine, they analyzed the system with Aerotech’s 2D plot function to determine whether the roundness problem was from the following error. The user then increased the servo gains to reduce the following error. In this process, the excess gains made the system unstable. A tool was required to remove the oscillation and to increase the servo gains so that the peak following error would be greatly reduced.
Stroke: 2 mm
Velocity: 44 mm/s
Acceleration: 2000 mm/s2
Figure 6. The largest following error in this example is +7 µm to -5 µm at accel/decel zone.
Figure 7. In order to reduce following error, increase the servo gain from Kp=283247; Ki=2710 to Kp=565151; Ki=5407. Crossover frequency increased from 34 Hz to 40 Hz.
Figure 8. Capture frequency response plots from different locations using the Overlap plot function. When we overlapped the plots, we found the Gain Margin is less than 6 dB.
Figure 9. Use two notch filters to improve the Gain Margin. The Gain Margin is increased from 5.5 dB to 9.6 dB. The system is now stable.
Figure 10. When the system is stable, increase the open loop gains from Kp=565151; Ki=5407 to Kp=798291; Ki=7638. Crossover frequency is increased from 40 Hz to 52 Hz. The increase of the gains will effectively reduce the following error.
Figure 11. During motion the largest following error at the accel/decel zone is now reduced to +4 µm to -3 µm when the gains are increased.
The friction from the linear bearings reduced the low frequency response. In order to increase the low frequency response, we used the Enhanced Tracking Control (ETC) function to increase the low frequency gains, and make the system behave closer to an ideal frictional system.
Figure 12 and Figure 13 are Digital Scope 2D plots. Comparing the G-code, we found the following error is the largest in Line 60 – there is 2.9 µm error in the X direction, and 1.4 µm error in the Y direction. Activating the Enable Position Error Formatting sets the following errors that are greater than 2 µm to be shown in “Green Sections”.
Figure 12. Compare the G-code to find the largest following error location. We found the following error is the largest in Line 60.
Figure 13. 2.9 µm error in the X direction, and 1.4 µm error in the Y direction.
Use Loop Transmission to measure the XY frequency response plots. We found that when the X and Y axes are at low frequency, the friction is higher.
Figure 14. The Loop Transmission plot of the X axis.
Figure 15. The Loop Transmission plot of the Y axis.
Figure 16. The Loop Transmission plot of the X axis (red curve is without ETC; blue curve is with ETC).
When we activate the ETC function and adjust the system, the dynamic behavior of the system becomes closer to an ideal frictional system. The red curve is without ETC and the blue curve is with ETC. In the low frequency zone, the system damping is reduced and low frequency gains are increased.
Figure 17. The Loop Transmission plot of the Y axis (red curve is without ETC; blue curve is with ETC).
Figure 18. With ETC activated, the global following error is already less than 1 µm (no green sections anymore).
Figure 19. Without the activation of ETC, the X axis has 2.9 µm following error and the Y axis has 1.4 µm following error. With the activation of ETC, the following error is significantly reduced – the X axis following error is now 0.7 µm and the Y axis has 0.1 µm following error.
Advanced motion control techniques can significantly improve the throughput and quality of high-performance laser micro-drilling. From fundamental servo tuning by using frequency response tuning techniques, filtering the resonance, and increasing the servo gains to improve the throughput, to improving the roundness by limiting the acceleration and improving the following error, to improving program execution efficiency in order to quickly process bulk data points, there are a variety of control methods that can help you reach your goals.
About the Authors
William Yeh is the Branch Manager at Aerotech Taiwan, Taipei, Taiwan.
Jerry Lin is a Control Engineer at Aerotech Taiwan, Taipei, Taiwan.
« Back to White Papers