Incipient Failure Detection - The Detection of Certain Contaminating Processes 670633

Three separate and distinct electrolytic and one galvanic process were identified by visual inspection, metallographic, electron microprobe, and x-ray diffraction analysis in a clocked, flip-flop integrated circuit flat pack and/or the associated printed circuit test jig (two on flat pack and two on circuit board). These four processes were all found to be detectable by the use of noise measurements in microvolts per root cycle at 1000 Hz (cycles per second). The direct current applied for noise measurement to the integrated circuit devices was 100 micro-amperes, as compared to the 6-8 milliamperes required for normal operation. After initial experimentation, the devices were caused to fail in a laboratory ambient environment, followed by an acceleration of the rate of electrolytic reaction through the use of essentially 100 percent relative humidity, versus the upper specification limit of 80 to 98% relative humidity.

In all cases, failures of the devices were preceded by a rising characteristic of high noise readings, 10 to 100 times normal (0.017 to 0.028μV = normal), which allowed predictions of failure to be made. The correlation between measured noise level and extent of contamination was calculated and was found to be high (.631). It was found that linear and polynomial regression equations could be developed to enable calculation of the degree of contamination indicated by measured noise readings.

While this paper reports on the detection of contaminating processes in integrated circuits, the ultimate objective of this research effort is to develop an incipient failure detection subsystem which can be applied to aerospace vehicles and ground equipment. Hopefully, the objective will be attained through monitoring certain characteristics present and associated with incipient failure in operational equipment.

A highly significant by-product of the research effort was the identification of at least four of the contaminating processes, the principal pre-existing and process produced contaminants, and the conditions which appear to accelerate the processes. Further work is being done to develop techniques for the control and/or elimination of these contaminants and associated destructive processes. The research clearly indicated the necessity for exercising rigid clean room controls to prevent contamination during the assembly of integrated circuits into sub-assemblies. While the processes proved to be extremely slow under ambient conditions, the processes were extremely fast where condensation of water vapor occurs (essentially 100% relative humidity). However, even in the first case, it is postulated that failures could occur as soon as six months. With actual moisture present between the leads and/or with the accumulation of water, as under essentially 100% relative humidity, the first failures could occur in eight hours with the bulk of failures falling in a 40 to 50 hour period and some not occurring for 150 hours or more.

The use of certain non-hydrogenous conformal coatings appears to eliminate or control at least three of these destructive processes. Other currently employed conformal coatings appear to have little or no effect upon the processes reported herein. This matter is being studied further and will be reported on in a later paper.

These destructive contaminating processes may very well be a greater source of integrated circuit failure than any other currently known source.