Close the quality-control loop

Ask most engineers why their companies inspect products, and they'll likely answer, "To keep us from shipping defective products." They're half right. No company wants to have customers complain about a package that doesn't contain the right product, PCBs that lack components, and so on. Those companies have another reason to inspect products, but it often gets forgotten. Inspections should identify defects so someone can correct the process and prevent the manufacture of more defective products.

On a PCB manufacturing line, for example, a machine-vision system will inspect boards after a printer deposits solder paste on conductive pads. The inspection system makes it easy to pull defective boards off the line. But the inspection system also provides useful information about the solder-paste printing process. By using that information, an operator can adjust the printer, clean a stencil, replace aging solder paste, or take other corrective action.

The semiconductor industry has long realized the benefits of applying yield-management or defect-management software to its fab lines. This sophisticated software takes test results and maps them onto the physical layout of a wafer. By examining where defects occur, how often they occur, the types of defects that occur, and other factors, fab operators can determine how to adjust the wafer-processing steps. By compiling a database of such information, chip manufacturers can characterize an entire production line and optimize its output.

In the semiconductor business, a wafer of hundreds or thousands of devices passes through a fab as one lot. Unlike on a PCB production line, from which someone can remove individual defective boards, all the chips on a wafer go through all the processing steps and bad chips get weeded out at the end. Thus, yield management assumes a key role in the production of as many good devices as possible per wafer. You can't repair a defective chip or run it back through the fab a second time.

The PCB manufacturing world needs to take a careful look at the quality-control tools used by semiconductor manufacturers and learn to better use inspection as a tool to improve quality, not to simply avoid shipping bad products. Some PCB manufacturers do take advantage of statistical process-control (SPC) software tools, and they use results to adjust their processes—manually. But the electronics industry still lacks a link that connects the output of machine-vision systems with process equipment. The machine-vision system suppliers and the vendors of production equipment need to figure out how to have their equipment exchange information that corrects problems in a closed loop. Until they do, machine vision will not reach its potential as a quality-control tool. Instead, vision systems will simply continue to kick defective products off assembly lines and produce charts of defects on which subjective human operators must act.