1-3hit |
Kiyoshi OGAWA Yoshimitsu SAKAGAWA Yoshio SUNOHARA
This paper proposes, as a new LSI reliability estimation method, a procedure for estimating the failure rate of LSI by classifying failures according to failure modes and by expressing the failure rates as functions of the LSI design parameters. As a concrete example, bipolar logic LSIs and MOSRAMs were taken up. Their respective failure modes and the design parameter dependences for the failure rates were considered, in order to obtain the relations between the integration density and the failures. As a result, it was clarified that the failure rate per gate for bipolar logic LSI was inversely proportional to (the number of gates)0.4 and that the failure rate per bit for MOSRAM was inversely proportional to (the number of bits)0.7. And it was concluded that the dominant failure modes will be in the metallization for the bipolar logic LSI and in surface degradation for the MOSRAM. Thus, new LSI reliability estimation was carried out. And the technical problems to solve or new technical targets to study in order to improve the reliability of high density LSIs were clarified.
The effect of the silicone vapor on the reliability of the micro-motor was examined. Adsorbed silicone was decomposed to SiO2 by heating due to the discharge between brush and commutator surface. It was found that the operation time until the failure was extremely shortened by the formation of SiO2. The existence of the maximum operation time until the failure was found as depending on the number of revolution. For the higher revolution, many amounts of SiO2 accumulated by the decomposition of the silicone shorten the operation time. For lower revolution, as the torque of the motor reduces, the operation time also shortens. Therefore, the maximum operation time exists for optimum revolution.
A new quick response rain gauge of waterdrop counting type with a wide measurement range from 0.7 up to 300 mm/H has been developed. From the result of the filed operation for 3 years, this gauge is proven to have the accuracy better than 10%.