Mastering the dark art of screen printing: simultaneous visual inspection of the PCB and stencil improves product quality while minimizing operator intervention.

Today's printed circuit board (PCB) assemblies combine small, high-density devices and packages with more conventional components such as connectors. Increasingly complex product mixes and reduced cycle times have made accurate, reliable screen printing on shrinking PCB real estate a main concern in the modern production line. For example, automotive safety circuits are pushing the requirements of this type of build and the necessity for more inbuilt verification. Yet screen printing has always been a "dark art" process, and current trends increase the challenge of identifying solder defects correctly and quickly.

Printing defects occur early in the manufacturing process, so they must be detected immediately to achieve the high yields and high quality needed to be competitive. To meet cost requirements, printing problems must be corrected without halting the production process. Fortunately, the latest advancements in on-board machine vision and process-control tools provide accurate monitoring and control of the screen-printing process in real time, yet they require minimal operator time.

Finding the Flaws

A variety of defects can be introduced during the printing stage, including paste insufficiency, alignment problems, bridging of paste deposits between pads, stencil smearing and blockage. A system designed to monitor such problems must be capable of detecting these defects on all common PCB finishes, from copper and gold through to the poorest hot air solder level (HASL) finishes.

Complicating the issue are other variables: a variety of stencil designs and materials; pastes supplied from many different manufacturers; and various deposition methods, speeds and pressures. Regardless of these variables, the printing process must be monitored and controlled in real time.

A simple two-dimensional vision-inspection system will detect the amount of paste present on the programmed areas of the product, typically the component pads (Figure 1). As the paste is deposited through a particular aperture in the stencil, 100 percent of the expected paste would correspond to a perfect deposit for that aperture size. Of course, variations will occur over time, and tolerances can be set to maintain the optimum deposition performance. For example, on a 0.75-mm-diameter ball grid array (BGA) pad, a paste level of 75 percent might be acceptable, but anything below that level would cause some concern. A paste level below 50 percent would be a severe issue as it might lead to product defects. Therefore, setting alarms that alert the operator when the paste level is unacceptable--before a bad product gets passed--is crucial.

[FIGURE 1 OMITTED]

If making necessary corrections depends on a line engineer reviewing data before taking action, the production line will come to a halt. This situation is unacceptable. Therefore, to meet the needs of cost-effective production, a closed-loop feedback system that initially allows the printer itself to take corrective action is ideal. This kind of system can take steps such as activating the under-screen cleaner, thereby making current and future boards acceptable to the rest of the production line without stopping production.

A closed-loop feedback system can accommodate a range of print-related defects, but it should also be coupled to a statistical process control (SPC) system, giving the engineers data to note trends in the system's performance. Repeated changes over time, or a need for excessive cleaning, can indicate a need for process-related tuning that will optimize the printing operation, thus maximizing the amount of good product.

Capturing Defects at the Source

Detecting production defects is always most cost effective when done at their source. Because the vision system verifies the amount of paste on a pad as it is printed, this verification is an excellent first step in that direction. Production can be halted before value is added to defective boards.

Unfortunately, this verification does not, by itself, help the operator know what is causing the problem. For example, if inadequate paste is present on the PCB, it could be due to inadequate paste on the stencil or blocked apertures in the stencil. Thus, a vision system monitoring the pad does nothing more than identify that a problem exists.

To keep the system online and make corrections quickly in real time, the operator must know the root cause of the problem. To do this, the vision system must also inspect the stencil.

Fortunately, the vision system can be constructed so that board and stencil inspection are done in parallel. With this approach, as soon as the operator knows a problem exists, he or she also knows the condition of the stencil. Consider two examples of how this method would work.

The first example is insufficient paste on the pad. The first indication is that the paste percentage on a particular area of the board is lower than a pre-determined warning level; such as the 75 percent level for a BGA pad. At the same time that the inspection system identifies this problem, it is also inspecting the stencil and notes the blockage of the apertures. The system automatically initiates the under-stencil cleaner (USC) to clear the stencil's apertures. The printer will then automatically reprint the board and verify the paste present, which should now be correct within tolerances. In this case, no operator intervention is required.

In the second example, assume that bridging between pads is flagged as a problem at the same time that the vision system identifies under-stencil smear. The USC clears the smear, enabling the next board to be printed with good quality. If the condition occurred regularly on this site, the printer would stop and warn the operator that a more serious problem exists. Alternatively, the engineer monitoring the SPC data would observe the trend and initiate appropriate corrective action: for example, providing more tooling support under the product or reducing the print pressure.

These strategies are effective because they catch and solve the problems in real time and at the source: the screen printing process.

Maximizing Yields

The production line cannot afford to give the screen printer extra time; more than the core cycle time in printing and conveying product. Thus, a compromise is needed between the volume of products made and how much quality is built into those products.

One hundred percent real-time inspection of printing operations is not typically practical, so inspection must be selective. To get the most cost-effective inspection, consider three areas on the boards:

* positions or devices on the board that are historically challenging for screen printers

* locations where high-cost components are placed

* locations of components that are difficult and costly to rework.

Typical examples of troublesome areas are printing pads for fine-pitch quad flat packages (QFPs), where the paste volume is small and the pads are close together. Another problem area is ball grid arrays (BGAs) where, once the device is placed, rework would be difficult if the solder was insufficient. Chip-scale packages (CSPs) also require low-volume paste deposition, and the features are small.

With the ever-increasing complexity of PCB design, flexibility is required in the core inspection algorithms. For example, pads can now be printed with a series of deposits. However, each deposit must be inspected instead of just the whole pad.

The Need for Speed

On-board inspection of these key areas takes time, but the benefits that inspection provides in terms of built-in process quality far outweigh the time required. Although all key sites on every product must be inspected, the sites can be prioritized. Then, to accommodate the time constraints of production, the inspection can be structured to inspect all key sites, with further inspection of less critical areas if time permits.

The process can be built to a set level with the correct print parameters and all associated actions and consumables. Once a comfort level with the setup has been achieved, tuning down the amount of inspection will speed production. For example, instead of inspecting 40 areas of the board, one option might be to inspect 20 areas on most boards, and inspect all areas on every "n" boards. This approach reduces inspection time during production, yet monitoring the SPC data still continuously verifies the process.

Building inspection and data control into product setups will take time in the initial process setup. However, this approach will pay dividends through process maintenance during real-time inspection.

Post-print inspection is key to maximizing inspection throughput. The inspection setup is a one-time operation and, with some printers, can even be done offline, thereby minimizing the machine downtime for setup.

Pre-print inspection takes time. The factors that affect inspection speed are:

* How much data can the inspection system capture in one pass?

* How fast can the data be processed?

* How quickly can the vision system get to the next site?

* What is the field of view? Remember that the field of view plays a dual role in gaining speed. The larger the field of view, the less time required for setup as more data is available from each area at one time. And after printing, more area will be inspected in one camera move.

If the system can process data on the fly, while the camera moves from one inspection area to the next, impressive gains in inspection speed will result. A camera with simultaneous look-up/look-down capability can capture images from the board and stencil simultaneously. This approach keeps the relationship of the data constant and speeds the inspection routine as it goes from one data site to the next.

Combining these factors optimizes the inspection routine, providing the best inspection of the product and the process in the least amount of time.

Lighting and Optics

One last factor that affects inspection performance is the surface texture of the PCBs. Uneven surfaces can inhibit seeing defects clearly. For example, demand is increasing for inspection of HASL boards. However, the unevenness of the finish on HASL boards means that some areas can be missed by conventional lighting systems.

The solution is to combine two lighting sources: on axis and off axis. This lighting arrangement, if configured to the correct angles and correct choice of illumination devices, can make the poorest HASL finish appear uniform to the vision system (Figure 2). The basic tuning of a system with two lighting sources accommodates about 80 percent of all surfaces and can be adjusted further if necessary.

[FIGURE 2 OMITTED]

The location of the image and its physical size and position in the focal plane can sometimes affect the image quality. A monochrome (CCIR) camera using telecentric optics will tolerate a wider range of objects. This system provides a major benefit in the accurate and repeatable detection and reporting of inspection results.

The telecentric lens eliminates the need for a pre-align stage to bring fiducials into the center of the field of view before doing a final alignment. With conventional optics, the distance between camera and the fiducial changes the size and physical appearance of the fiducial (Figure 3). Telecentric optics ensure that board warping, the distance between the camera and the board, and the actual position of fiducials within the field of view do not affect alignment accuracy. With telecentric optics, the relative distance and sizing of objects will be the same wherever they are seen in the field of view. The net result is an effective vision system that suits the needs of production environments.

[FIGURE 3 OMITTED]

Matthew Dickinson is vision product manager with DEK Printing Machines Ltd., Weymouth, Dorset, England: e-mail: [e-mail].