Which would you choose to test the design of a new printed-circuit board machine — traditional lab testing or text-based simulation software? Engineers at Universal Instruments chose neither. Instead they used a new type of simulation software that more closely emulates the way an engineer looks at the world.
PC-based simulation and design software proved 25% faster than building a breadboard and performing physical testing during the analysis and design of motion components in our latest machine. The software let us see the whole dynamic picture of the mechanical components. This capability helped us design components that properly support the acceleration and deceleration rates of the X and Y axes, ensuring that we always had tight control in moving the load.
Universal Instruments makes component insertion and placement machinery for printed-circuit board (PCB) manufacturers, with more than 16,000 machines installed around the world.
These machines populate PCBs automatically at speeds over several thousand components per hour. Every movement of the placement head on the General Surface Mount Application Machine (GSM1) is pre-programmed, from picking a component from a feeder to placing it in a certain location and angle on the board. Accuracy is critical in lining up a component’s leads over the board’s solder pads as the distance between leads may be only 8 to 15 mils. Therefore, a vision recognition system ensures precise component placement.
Designing speed and accuracy in a vertical axis
The head and Z-axis of the GSM1 can handle components from small 0402 capacitors to 50-mm-square quad flatpacks, Figure 1. One of our design goals was to reduce the time it takes for the head to pick up a component and place it on the board. However, we did not want the head to go up or down so fast that it would sacrifice accuracy. A settling time of 10 msec would ensure stable descent into position. Therefore, we specified a bandwidth for the axis servos of 20 to 40 Hz for positioning and 100 to 200 Hz for velocity. As accurate as the servos were, though, the performance of the system would be limited by its mechanical components — clutches, pulleys, cables, and springs. Clutches were used to reduce the complexity of controlling four vertical placement spindles. One servomotor and one encoder is required with the separate spindle clutch arrangement. Typically, only one spindle is used at a time to place a part, the clutch is applied to couple the spindle to the main drive shaft. If four parts are picked up at a time, then all four clutches are applied to connect the spindles to the drive shaft.
The servomotor for the Zaxis head drives the main shaft via a timing belt pulley assembly, Figure 3. The motor shaft is rotating five times faster than the main drive shaft. The timing belt provides a step down ratio to the main drive shaft. To move the head about the Z axis, a solenoid clutch couples an idle pulley and cable to the servo-driven shaft. Connected to the other end of the cable is the Z-axis module and a spring to keep the cable taut. If the clutch is engaged, the Z-axis will accelerate to a hard stop. If the spring does not have enough tension when the Z-axis head accelerates, the cable attached to the shaft and running through the pulley will become slack.
So our challenge was to design the mechanical elements to support the head’s acceleration and deceleration, yet maintain tight control of the load while moving it up and down. If the load was not controlled, the head could begin to bounce and lose its grip on the part.
Previous design methods involved assembling a breadboard and adding instrumentation and sensors to measure response. We also tried early versions of simulation software. Such programs were difficult to work with because they required text input, including the descriptions of nodes and interconnections. In addition, these programs did not have block diagramming capability. We had to rely on hand drawn diagrams. Therefore, because of the large time consumed in these programs, we didn’t develop a model to any depth.
Applying visual simulation software
We learned of a simulation program that let us use a visual block diagram rather than lines of text to describe a model. The software, VisSim, from Visual Solutions, Inc., Westford, Mass., graphically depicts block diagram flow on a PC screen. A Windows-based program, blocks and connections were assembled with the classical control perspective approach. By comparing several simulations with lab results, we determined that the program’s predictions matched the results enough to convince us of its accuracy.
To model the GSM1 axes in VisSim, we began by building a block diagram. Each component is represented as an icon (or block). These components are selected from a library of pre-defined blocks, including summing junctions, gains, and numerical integration methods. We model the properties specific to the servo drive, such as position, velocity, and current control. We model all the parameters affecting the cable tension, including the acceleration and deceleration rates, the mass of the load, the spring rate, and friction. Each component’s operating parameters are entered in a dialog box. To connect — or wire — the blocks together, we point the cursor between two blocks and click the mouse, Figure 4.
Continue on page 2
After the model is in flowchart form, the next step is to run simulations. The software’s interactive design environment, with plots and real-time graphs, lets engineers see the dynamics of the cable’s tension. We monitor the tension as we put in known accelerations and decelerations and the known mass. We adjust the spring rate, so the cable’s tension is positive at all times, and immediately see the effect on the system.
The whole dynamic picture
Now that we can see the dynamic picture of the cable’s tension and the control of the load, we select the mechanical elements to properly support the acceleration and deceleration rates and ensure that we always have tight control in moving the load. VisSim also allows us to design the proper gains and bandwidths to achieve the needed settling time of the servo mechanism. From the optimal settings, we built a hardware prototype of the Z-axis. After testing it, we released the unit for production.
We finished the design of the GSM1 Zaxis in about 25% less time than it would have taken using breadboards or other design methods. Also, we didn’t waste time ordering parts for a breadboard. For example, before the software, we would have ordered components, such as special springs, for the breadboard. If those springs did not work properly in the breadboard, we would have to order others. This type of wait is one of the main causes of long design cycles. But by using simulation software, we could identify the correct parts before ordering, thereby avoiding any unnecessary wait.
From the success of the GSM1 Z-axis application, we use VisSim on every new servo system design. The software also helps us make improvements to existing products. To test a proposed change, we call up that machine’s software model and run a quick simulation.
Simulation software provides several additional benefits:
• Enables engineers to better understand the workings of the system.
• Helps engineers look at signals that are normally not reviewed in breadboard design because they are difficult to monitor.
Jim York is a motion control engineer in the Engineering Div. of Universal Instruments, Binghamton, N. Y.