Author: Sudarshan Bahukudumbi

Publisher:

Published: 2008

Total Pages:

ISBN-13:

The relentless scaling of semiconductor devices and high integration levels have lead to a steady increase in the cost of manufacturing test for integrated circuits (ICs). The higher test cost leads to an increase in the product cost of ICs. Product cost is a major driver in the consumer electronics market, which is characterized by low profit margins and the use of a variety of core-based system-on-chip (SoC) designs. Packaging has also been recognized as a significant contributor to the product cost for SoCs. Packaging cost and the test cost for packaged chips can be reduced significantly by the use of effective test methods at the wafer level, also referred to as wafer sort. Test application time is a major practical constraint for wafer sort, even more than for package test. Therefore, not all the scan-based digital test patterns can be applied to the die under test. This thesis first presents a test-length selection technique for wafer-level testing of core-based SoCs. This optimization technique, which is based on a combination of statistical yield modeling and integer linear programming (ILP), provides the pattern count for each embedded core during wafer sort such that the probability of screening defective dies is maximized for a given upper limit on the SoC test time. A large number of wafer-probe contacts can potentially lead to higher yield loss during wafer sort. An optimization framework is therefore presented to address test access mechanism (TAM) optimization and test-length selection for wafer-level testing, when constraints are placed on the number of number of chip pins that can be contacted. Next, a correlation-based signature analysis technique is presented for mixed-signal test at the wafer-level using low-cost digital testers. The proposed method overcomes the limitations of measurement inaccuracies at the wafer-level. A generic cost model is developed to evaluate the effectiveness of wafer-level testing of analog and digital cores in a mixed-signal SoC, and to study its impact on test escapes, yield loss and packaging cost. Results are presented for a typical mixed-signal "big-D/small-A" SoC from industry, which contains a large section of flattened digital logic and several large mixed-signal cores. Wafer-level test during burn-in (WLTBI) is an emerging practice in the semiconductor industry that allows testing to be performed simultaneously with burn-in at the wafer-level. However, the testing of multiple cores of a SoC in parallel during WLTBI leads to constantly-varying device power during the duration of the test. This power variation adversely affects predictions of temperature and the time required for burn-in. A test-scheduling technique is presented for WLTBI of core-based SoCs, where the primary objective is to minimize the variation in power consumption during test. A secondary objective is to minimize the test application time. Finally, this thesis presents a test-pattern ordering technique for WLTBI. The objective here is to minimize the variation in power consumption during test application. The test-pattern ordering problem for WLTBI is solved using ILP and efficient heuristic techniques. The thesis also demonstrates how test-pattern manipulation and pattern-ordering can be combined for WLTBI. Test-pattern manipulation is carried out by carefully filling the don't-care (X) bits in test cubes. The X-fill problem is formulated and solved using an efficient polynomial-time algorithm. In summary, this research is targeted at cost-efficient wafer-level test and burn-in of current- and next-generation semiconductor devices. The proposed techniques are expected to bridge the gap between wafer sort and package test, by providing cost-effective wafer-scale test solutions. The results of this research will lead to higher shipped-product quality, lower product cost, and pave the way for known good die (KGD) devices, especially for emerging technologies such as three-dimensional integrated circuits.