Researcher looks at 3.7 billion solder joints to assess DPMO
Steve Scheiber, Contributing Technical Editor -- Test & Measurement World, 7/24/2007 8:57:00 AM
In 1999, Stig Oresjo, at that time Test-Strategy Consultant for Agilent Technologies, gathered data on more than 1 billion solder joints on more than 328,000 boards of 566 types from eight OEMs and seven contract manufacturers in the US and Western Europe. He found that despite claimed quality levels of between 75 and 200 defects per million opportunities (DPMO), companies really experienced more than 1000. For nearly a decade, his analysis has represented the definitive work on the subject.
Now retired from Agilent and principal of Test Strategy Consulting, Stig has conducted a new round of research, this time examining more than 3.7 billion joints. When I asked him why he had gone back to that bottomless well, he commented, "The first study was beneficial not only to Agilent but to the rest of the industry, so an update after 8 years made a lot of sense.
"Every board manufacturer needs to know defect levels so that they can plan their test strategies. When companies saw my original results, even the ones claiming fewer than 100 DPMO nodded, saying that I had come closer to the truth. Their public claims can reflect ideal conditions—which is the way all companies represent manufacturing quality—but if they don't recognize and deal with even worst-case reality, then defects will escape to end customers, which they want to avoid at any (reasonable) cost."
He continued, "A lot has happened in the 8 years since I did the last study. For one thing, the first study included only the US and Western Europe. Much electronics manufacturing has moved to Asia, so this time we included companies there as well. Manufacturing techniques have improved, so we expect defect rates have gone down, but boards have become more complex, which works in the other direction. Also, many more companies' processes now include other types of inspection prior to x-ray."
The new study involves 14 companies—four in Europe, two in the US, and eight in Asia—and covers 1100 different board types. It incorporates high-volume as well as low-volume and high-mix.
Stig commented, "We had to allow for variations in the way companies in different parts of the world manufacture their boards. In Asia, for example, where labor tends to be less expensive than equipment, early inspection stages may include more manual inspection steps supplemented by automated techniques such as AOI [automated optical inspection], whereas companies in the US and Western Europe more commonly rely more on AOI with limited manual inspection. For consistency, we restricted our analysis to boards crossing the x-ray system as we had before.
"The results showed half the defect rate of the earlier study—500 DPMO. We have no doubt that quality levels are higher, but part of the apparent improvement comes from the greater number of defects that have been removed before boards get as far as x-ray. Still, in companies' efforts to reduce manufacturing costs, test and inspection often take a big hit. Even maintaining the same quality level under those circumstances would have been an impressive achievement."
