Bringing home the data

SIDEBAR: Adapting a commercial software system READ OTHER NOVEMBER ARTICLES:  Table of contents, Nov. 2005 NOVEMBER FEATURES: Bringing home the data Beyond at-speed Protecting the tester VoIP complicates test Sometimes, analog is better

Santa Cruz, CA—Plantronics headsets are everywhere: in homes, offices, contact centers, and airplanes. They've traveled as far as the moon, and today they're roaming wherever their wireless-phone-chattering wearers care to go. But Plantronics test engineers themselves are striving for a less peripatetic life. To that end, they're bringing up to speed software that should minimize their need to travel to far-flung manufacturing facilities. With the software's Web interface, the engineers can sit at their desks at Plantronics headquarters here and view detailed test activity at remote locations.

The test engineering team's ultimate goal is to ensure the quality of a broad mix of products produced in high volume in locations including Mexico and China. The engineers assemble rack-and-stack test systems at the company headquarters and deploy them, along with test programs, at the remote manufacturing facilities. For optimum results, the engineers must build test systems that are accurate, remain calibrated, and correlate well with each other. Further, they need to ensure that each test approach optimally balances instrumentation and labor resources and that production technicians apply the systems properly. To perform their jobs effectively, the Plantronics engineers need lots of data, which hasn't always been easy to obtain.

Responsibility for the test-engineering function ultimately rests with Rick Frier, senior manager of global test engineering. Frier got his engineering start in the medical business, where he designed fail-safe circuitry for hospital pumps that deliver fluids to patients. Then, he entered the disk-drive business, where, during a 20-year stint, he gained database experience that he's been able to apply at Plantronics.

Custom enclosures hold products under test while providing acoustic and RF isolation to eliminate the effects of ambient noise and interference from nearby test systems.

Plantronics is now focusing on wireless products that employ Bluetooth and DECT (digital enhanced cordless telecommunication) RF communications, such as the Voyager 510 Bluetooth headset and CS50 900-MHz DECT wireless office headset system. The combined audio and RF technologies of such products complement Frier's interests as a ham radio operator, but the combination also complicates the test process by adding RF functionality to the acoustic features the company has long been known for.

"Now, there are a lot more and varied devices on our boards than we had with just regular acoustics," said Frier. "So, when we start manufacturing these boards at high run rates, it's not surprising that we might initially experience higher fallout rates. Initially, it was difficult for us to understand what specific problems were occurring at our contract manufacturers, because it took us days or weeks to get the information we needed to try to identify critical failure modes. And when we did get the data, we couldn't always be certain of its accuracy."

Terry Walters, senior VP of operations, adjusts a rack of equipment that will test DECT wireless products.

And even when remote facilities did provide timely, accurate data, it was rarely in a format that engineers could easily use to visualize how well a facility was operating. "When we're ready to begin manufacturing a product, we do pilot runs to gather information," Frier said. "We often had to put data by hand into Excel to generate the charts that would let product engineering visualize upper and lower limits, for example, and look for good distribution within those limits." All in all, Frier said, it had been a very painful and expensive process.

As a consequence of this painful process, Frier said, "it became challenging to get yields to a point where we were satisfied." In addition, he said, his staff was spending too much time on planes and too little time developing and optimizing new test systems and programs. Frier explained that these issues were key to justifying the development of a database that could track each manufacturing facility, each product line, and each test system.

Rick Frier readies a test chamber and fixture for the test of a CS50 DECT headset.

The goal that Frier envisioned for a database implementation was to acquire and deliver accurate, timely data from these test systems and to make it available at headquarters in formats useful to engineers as well as managers. Frier's background in databases—he has a master's degree with an emphasis on management of engineering and technology, which included database development, and he has developed two databases for the disk-drive industry—enabled him to envision an optimal implementation for Plantronics: "We required two significantly different modes—one for engineering-level analysis and one for management. At the engineering level, we wanted to see the details of the units run to understand Cpk [process capability index] and to see the actual test limits and the spread of failure modes. Are failures catastrophic or marginal, and if marginal, what's going wrong? Is it the test limits, is it the original specification, or are we getting bad parts? We need to be able to drill down and answer these questions."

The Plantronics implementation of the SigmaQuest database presents data in a variety of formats. The top chart illustrates yield increases resulting from an April 1 engineering change order. The bottom chart shows daily production levels at a contract manufacturer in Mexico (red) and in China (blue). (The SigmaQuest charts have been altered here to disguise proprietary information.)

Frier noted, "When we actually got into reviewing different commercial databases, we found nobody had exactly what we needed to present our RF and acoustic test data in the formats we wanted to see. We had to find somebody who would be willing to work with us to develop a database that would present the data in the manner we needed to have it presented in." SigmaQuest was an attractive partner in that regard, Frier said, because the company offered price advantages in exchange for being able to integrate functions it developed for Plantronics into its standard products (see "Adapting a commercial software system").

Frier said he was pleased to help identify functions that would be applicable to firms other than one that made headsets. "It helps SigmaQuest and it helps us, because we obviously want SigmaQuest to be successful and stay there and continue to support us."

All of Plantronics' PCBAs are made at contract manufacturers. "We just build the test equipment and support the CMs with our test equipment. For final assembly into plastics, we do have a facility—called Plamex—in Tijuana, Mexico, which assembles probably about 80% of our products. We have a CM in Guadalajara, Mexico, and three in China, which build our PCBAs, so we have five facilities that I need to set up the system on. We've taken it a step at a time on the way to getting 100% of our test equipment up and running with the database—I was very hesitant to just roll it out quickly across all locations because I wanted to avoid having our system crash their systems. I wanted to make sure we did it very smartly and cautiously to make sure we got the information we needed without interfering with the CMs' operations."

Success of the database project requires acceptance on the part of Plantronics' CMs. "The CMs are supportive and not supportive. Some of the facilities have jumped into it pretty willingly; some have been hesitant, imposing delays and not returning phone calls. In those cases, we have just said, 'look I know you guys don't really want us there, but we are going to come, so you are going to have to get used to us.' I think that what they are concerned about is that we will have a lot more control and understanding of what's really happening on their manufacturing floor. It hasn't proved to be to their disadvantage yet, but I think they would prefer it if they remained totally in control."

He cited one example where a CM's control might be inappropriate: "To a manufacturer, it's a big deal in terms of money if a product isn't yielding. If yield is poor on one tester, an operator might be tempted to bump a lower limit down a tiny bit or an upper limit up a tiny bit in an effort to get the tester to correlate better with others."

He noted that there can be several causes of correlation problems: A test system's acoustic or RF shielding might be degraded by repeated opening and closing of its test chamber, for example. "If there's a correlation problem, we want to know about it and solve it ourselves," by repairing or replacing a fixture or taking other appropriate action. "We don't want the operator changing limits. But before implementing the database, we had no way to tell if he was doing that short of stationing a test engineer on the floor with him."

The CMs that are supportive, he added, realize that the timely acquisition of accurate data will ultimately help improve their production processes, allowing Plantronics engineers to respond promptly and effectively to correlation or yield problems. When asked if the CMs themselves will have direct access to the data, Frier said, "That's still in discussion. What level of access do we give CMs? I can't see any disadvantage to letting them view the data. Of course, we won't let them change anything, but if they can look, they may be able to identify problems themselves and solve them before those of us in Santa Cruz get to our desks in the morning."

The level of support is evident in the access the CMs provide to their IT infrastructure. "Some let us set up our equipment [the servers on which the data resides] in their air-conditioned IT rooms; some say we can set up on the production floor next to our test equipment— 'that's the most we will give you.' Either way works fine for us, but of course, I'd rather be in their IT room because the equipment located there should be more reliable. And in time, maybe even the reluctant CMs will see more of the benefit of the system and be more willing to work with us on that level."

No matter where the physical installation, there have been the typical IT issues to resolve: establishing firewalls that protect both Plantronics and the CMs and eliminating viruses on the test systems. Some operators do surf the Internet using the test equipment, Frier said, making anti-virus protection necessary. "We will tell the CMs that this is completely unacceptable, but what message does the operator get? Maybe, 'Just don't do this when the customer is here.'"

CMs aren't the only ones who need to accept the database—so, too, do Frier's co-workers. He said, "It's been interesting how people within Plantronics react to the system—some people have been enthusiastic from the beginning and some not. By now, though, most realize the benefits of seeing real, live data quickly. The ability to see what firmware version a product is using and to look at actual test-limit files is something that demonstrates its value very quickly."

He noted ways in which several constituencies can use the system. "Failure-analysis people can run the system from here to try to identify root causes. Management can look at data in a variety of formats. A manager can look at one day, which is useful when just starting production, or he can look at rolling weekly averages, which might be more useful for monitoring the long-term performance of a production line." Quality-control people, he added, want to look at yields on a daily basis to look for problems that might be due to different component batches.

"And software engineers can confirm that proper firmware versions are being installed. We used to e-mail firmware updates to our CMs but had no way to ensure they were updating all the test systems until products came back." In addition, he said, "Hardware engineers can get a detailed understanding of what's happening on the production floor. They can look at histograms of frequency-tuning performance or offset and drift or any other characteristics."

Of course, each person has particular ideas of what data should be available and what format it should be presented in. "I knew that was going to happen because it's happened to me in the past," Frier said. "Everyone who looks at a system will say, 'well this is great but it doesn't do exactly what I want.' So, we sat down with everyone who would use the system to compile a wish list."

Ultimately, the key selling point that got Plantronics executives on board was the money the system would save. Frier expects the system to pay for itself within eight months—mostly in engineering hours spent traveling to manufacturing facilities or entering data by hand. "Now, I can pull the data out and look at it in a format that makes sense to me in 10 or 15 minutes. That used to take as long as a week."

Frier emphasized that the savings in his calculations represent real savings and are not based on the intangibles that he said some database companies focused on in their presentations to Plantronics. His calculations, he said, are based on actual engineering time spent and production downtime avoided, all documented with accounting records. Factors like customer satisfaction, he said, "are very important, but I can't confidently assign a hard number to them. And I didn't need to in order to justify the purchase of the system. Improvements in customer satisfaction and so forth are bonuses on top of the hard savings I documented."