Wafer-Level Burn-In Will Cut Motorola Costs

Welcome Guest. | Log In| Register | Membership Benefits

August 04, 1998 (12:46 PM EDT)

By J. Robert Lineback,

In a move to reduce the use of expensive integrated circuit (IC) testers, Motorola Monday announced plans to become the first semiconductor manufacturer to burn in chips at the wafer level, rather than waiting until they are packaged.

Wafer-level burn-in technology promises to lower chip-manufacturing costs as much as 15 percent, and shorten production cycles by 25 percent, said managers involved with the project in Motorola's Semiconductor Products sector.

For the past 18 months, Motorola (company profile) has collaborated with capital-equipment maker Tokyo Electron, and interconnect supplier W.L. Gore and Associates to create production systems that will force early failures in marginal and defective devices before silicon wafers are sent to assembly plants for final packaging and test. The first production systems will be used in a wafer-level burn-in pilot line slated to begin operations in Austin, Texas during the first quarter of 1999.

"We will get an opportunity to sort out the good products while they are still on wafers, before we have spent money packaging the parts," said Les Hazlett, manager of strategic final manufacturing operations for Motorola's semiconductor unit. "As the die becomes more complex, the number of inputs/outputs increases and the package becomes more expensive."

Wafer-level burn-in is a key step toward the ultimate goal of performing full device testing before silicon substrates are cut into individual chips and packaged. Eventually, Motorola hopes to create a wafer-level test system that will work in conjunction with the burn-in technology to dramatically reduce or even eliminate the need for automatic test equipment (ATE) of packaged parts.

"Wafer-level burn-in is an enabling technology to allow for true wafer-level tests, which will result in even greater savings," Hazlett said. He would not disclose Motorola's road map or time schedule for wafer-level tests.

"We are seriously working on it," Hazlett said. "We are still going to be a major user of testers once wafer-level test is ready, but we hope to buy fewer of them."

If successful, wafer-level testing could have a major impact on the industry's $4 billion IC test-equipment segment. Semiconductor manufacturers have seen test costs rise in recent years, as chips become more complex and devices operate at higher speeds. ATE suppliers have been asked to lower test costs, but it has not been easy.

According to studies by Austin-based Microelectronics and Computer Technology Corp. (MCC), an industry consortium, wafer-level burn-in and tests could result in cost savings of more than 50 percent when chip volumes exceed one million units a year. MCC has also been involved in organizing a collaborative effort to create commercial wafer-level burn-in and test systems with memory makers and system suppliers.

Motorola's initial wafer-level burn-in systems will be used in its bump-assembly-test BAT 1 production line in Austin to cut costs and speed completion of fast static RAMs, Power PC reduced instruction-set computer processors, and microcontrollers. To do that, Motorola teamed with Japan's TEL and W.L. Gore of Elkton, Md., to create production-worthy systems.

In the project, W.L. Gore provided the interconnect substrates and boards for wafer contacts. TEL provided equipment expertise. Motorola said it has formed a joint venture with TEL to make further improvements to the burn-in technology, so that it can be used in volume production along with wafer probers that check out each die for device failures. Over time, the capabilities of wafer probers will be enhanced to eventually perform full tests of devices.

Wafer-level burn-in is expected to give a much needed boost to good die products because it will eliminate the need for special chip carriers to heat and test bare ICs. "While our motivation is to achieve improvements over burn-in of packaged parts, wafer-level burn-in will actually allow us to create a manufacturing process that's much simpler," Hazlett said. If so, the cost of unpackaged bare-die products could also become cheaper.

Related Stories: