Cutting test times: built to be run by novices, next-generation testers are designing complexity out.

Test equipment makers have to look at what is happening in Southeast Asia as both wakeup call and opportunity. In the U.S. and Western Europe, it is not unusual to work with engineers who have decades of manufacturing experience. They understand the subtleties of the problems they face and understand the concept of eliminating the last fault during test and inspection, and they know (or can learn) how to take advantage of the advanced features of their process equipment.

[FIGURE 1 OMITTED]

[TEXT NOT REPRODUCIBLE IN ASCII]

Asians have enjoyed no such luxury. Far fewer engineers have 25 years of experience simply because 25 years ago the technology did not exist in those countries. As a result, most engineers there are young and without the benefit of that level of experience. They have no time to learn a test system's complicated and elaborate features designed to get out that last fault regardless of the effort required. Therefore they have no need for a tester that includes lots of features they will never use. They look for simple, easy-to-use, economical solutions that even someone with only weeks or months on the job can get up and running very quickly.

In-circuit testers offer an excellent case in point. On early testers, every pin could act as a driver or receiver at any time. Since there was a 1:1 correspondence between pins on the board and the drivers/receivers at the tester, program generation and fixture design proved relatively straightforward and could proceed simultaneously. As the technology began to demand more pins and greater functionality, this architecture became both too cumbersome and costly. Companies introduced multiplexing techniques (Figure 1) that permitted a single driver/receiver combination to address eight or even 16 pins at the board. Although this new approach promised more apparent pins and more sophisticated capabilities, assigning one pin on a multiplexed network as a driver prohibited any other pin on that same network from serving as receiver. Program-development software assigned nodes based on the availability of pin resources. A nearly finished program was needed before building could begin on the bed-of-nails fixture, which stretched test-development times to the breaking point. Three days of programming and test-fixture data-creation time would be followed by three to four days to build the fixture and another two to three days to debug them both together. Total test-development time for a PCBA of typical complexity could easily exceed 10 to 12 days. Also, finding enough pins on a multiplexed tester to accommodate the circuit density on today's boards can require accessing real tester pins some distance away from a node under test. To do so, fixture wires have to be longer than with an equivalent non-multiplexed solution. The longer wires, in turn, can degrade test-signal quality.

Marty Leeke is a product manager at Agilent Technologies (agilent.com); [e-mail].