Increase reliability for screening by coating plastic parts

摘要

The selection of components for computer Integrated circuits (IC's) has been a major reliability problem in the past few years. What once was a 10-15- year cycle for -parts availability -- parts cycle is defined as introduction, growth, maturity and decline -- has decreased to as little as 4 or 5 years, with no maturity or decline cycle noted. Parts do not meet the equipment temperature range, which fail at the test level during the box screening, Environmental Stress Screening (ESS) or testing cycle, Qualification or First Article. Also, the number of MIL 883 or DESCI parts have greatly been reduced. The reliability engineer uses vendor-supplied numbers in the prediction that causes concern if the equipment is projected to have a long useful life; this data is supplied to the project logistics support group for equipment spares. Just a few years ago, the logistics data on our example program was stated to have a 10-year useful life but because of fielded conditions the hardware has achieved only a 7 year product life. Alliant just completed a 3-year production run on a 486-computer processor board. Three major MIL parts require an end-of-life buy this year and the processor is in its last year as a MIL screened ceramic part. Intel has stated we can still buy the commercial, unscreened part for the next few years. The present approach of using plastic IC's in MIL designs without a moisture barrier to correct possible package defects is a major risk. Vendor data for the reliability prediction value is usually overstated and may effect expected results of reliability testing. The reliability rule-of-thumb is that the hardware predicted MTBF must be at least twice the reliability test requirement. This was based on Chi-square data where the MTBF divided by the test time (length of on-time) had to equal 2.0 where at zero failures this provides a 60% confidence of passing the test. My management always wants a 90 % confidence, which would drive the MTBF to 10 times the test requirement. Given the limits which our vendors could achieve, the reliability model was based on four (4) times which provides an 82% confidence of zero failures and a 97% confidence of one or less failures. By a review of failure analyses conducted over a five-year period in the 90's, it has been determined for TTL devices that approximately 45% of all defects associated with these devices are attributed to package related defects (35% for ceramic devices and 52% for plastic packages). The screening tests for seal leakage are thermal cycling (which is part of the reliability screening and demonstration testing) and thermal shock (which is part of qualification). Ceramics parts are also screened for package defects by the vendor. Based on five years of conducting reliability demonstrations with designs using plastic or commercial parts, Paralene or X-Y coating ensures reliability improvement in the screening process and the failure rate reduction of about 40% can be used to support the hardware prediction.