An automotive company was experiencing field failures of an electronic module. Failure analysis on one module led to a via that, when cross-sectioned, showed a circumferential void. Predictably, the lab report regressed back through the system from the auto company to the module supplier to the contract manufacturer and finally to the PCB fabricator. The lab evidence was pretty conclusive that a defective PCB had been supplied. The requisite Corrective Action Report was produced by the supplier in record time. The report was absolutely stunning! The report's professionalism and polish were unbelievable. Up to a point.
The document had an embedded section that listed (correctly) the common root causes of this defect and then zeroed in on one as the culprit. The text and the descriptions were perfectly "textbook". There was a section to educate the reader on the technology that produced the defect, interspersed with very professional photos. The next section described the logic behind the ultimate root cause selection. This report obviously had required a great deal of time and analysis.
As the author progressed from the root cause section to the action plan, the document began to reflect on the time element. Reportedly, the fabricator in Asia created this document within 24 hours of receiving the lab report. Was this even possible? The Author concluded that such a report would have required more than 24 hours of my time and an equal amount of laboratory time (which could not overlap). Additional time would be required to add the "polish".
The summation was the final piece of the puzzle, and it seriously lacked the polish of the rest of the report. In fact, it was obvious that the summation was written by an entirely different person. It wandered, it had language errors, technical misstatements, and was, in general, poorly done.
In referring back to the action plan, the plan stated that (during the time to prepare the elegant report) that all completed PCBs in stock had been run through a 4-wire Kelvin test using a flying probe tester to determine if any further defective vias were in the unshipped stock. Without going into any detail as to how a 4-wire test works or could find a substandard via, assume two things: an open is still an open, and a discrepant via with enough remaining copper to show some continuity must be tested individually or the test is not valid. In other words, vias in an "end-to-end" test of a net cannot be characterized in this fashion (outside of a laboratory).
The action plan was so well written that (like the X-Files) you really wanted to believe. However, in the rambling summation, we learned that the fabricator had over 20,000 boards in stock. From the Gerber data, we learned that each board had over 200 vias of the size that failed. While all of the "closet statisticians" are doing the math, let's add one more small detail. All of the vias had solder mask plugs and the pads were covered. The time to test 4,000,000 vias (individually) no longer mattered because they could not be electrically accessed. The action plan said all boards had been tested and were determined to be good.
The report did not address the issue of how a circumferential void escaped detection in bare board testing (BBT) in multiple boards. The issue was glossed over with a description of residual copper in the holes which allowed all boards to pass continuity testing. No information was offered about the BBT results of the lot or any internal evaluation of BBT failures from the lot. The reader was left to conclude that all holes had some residual copper.
Re-reading the CAR from the beginning, it became obvious that other than the header information in the beginning and the summation at the end, the entire document had been prepared well in advance; it was canned! The author of the summation had only one task. Tie the canned document to the actual event by inserting facts from the actual event (such as number of PCBs in stock).
There are two shocking observations from the revelation that the report was done in advance. First, just how many requests for CARs were being processed by the fabricator that it became economically advantageous to have one ready to go for this issue? If they have one for voids in vias, do they have others for every element in their FEMA (Failure Effects and Mode Analysis)? What conclusion should we draw about how seriously they took the situation?
Second, have U.S. companies become so paper-process dependent that they are easy prey for such tactics? Have we lost the technical knowledge of PCB fabrication to the point that we accept a canned response based only on a slick professional appearance?
We have become so predictable with our paper-based quality systems that it is (obviously) easy to "game" the system, deliver no real solutions, and quickly get back to business as usual. For this fabricator, 20,000 boards got shipped; a few minutes of someone's time to draft a summation supposedly made the defects disappear. In the final analysis, who knows if the boards were any good?
Contact: DIVSYS International LLC, 8110 Zionsville Road, Indianapolis, IN 46268 317-405-9427 fax: 317-663-0729 Web: http://www.divsys.com