How should intermittent failures be analyzed?

EESemi.com FAQ

- questions most frequently asked by visitors of www.EESemi.com

Analyzing Intermittent Failures

Intermittent failures are failures that do not manifest themselves all the time. The fact that they are sometimes there and sometimes aren't can make them very difficult to analyze.

Just like any other FA, the first thing to do when analyzing an intermittent failure is to establish what the failure mode is. Once the failure mode has been established, the minimum set of conditions needed to make the failure mode observable over a reasonable amount of time must be determined. The more defined this set of conditions is, the easier the FA will be. The objective of this exercise is to be able to make the failure visible for a period of time that makes FA possible, since FA can not be performed on a 'good' unit. An understanding of these conditions also gives the analyst an insight as to what mechanism or physical phenomenon is responsible for the failure.

A word of caution though - in trying to make the failure mode observable for FA, one must take care not to 'aggravate' the failure mode or change it in any way. This means that the intermittency of the failure mode must be preserved - the failure should neither become permanent nor should the unit become permanently good.

A very common cause of intermittent failure is a marginally detached first or second bond that sometimes loses contact with its bonding surface. Another common cause is a broken ball bond neck or broken wedge bond heel, which can result in intermittent loss of contact between the bond and the wire. Bond-related intermittent failures such as these are relatively easy to analyze, since a simple curve tracer can verify the failure mode, i.e., the affected pin is open if the failure is there and normal if the failure isn't there. Furthermore, it is quite easy to make intermittent bond failures manifest themselves using a heat gun or, in some cases, a freeze spray can.

Changing the temperature of the package physically shifts its features relative to each other, causing the intermittent bonds to open. These bonds return to their previous state once the package has returned to room temperature. Once an intermittent failure is verified to be most likely bond-related (i.e., a pin alternates between being open and normal with the application of a high or low temperature), post-decapsulation internal inspection of the units can be performed to reveal the affected pins to have one of these bond-related defects.

Intermittent failures that require electrical bias to manifest are more difficult to analyze than bond-related issues. This is because a simple curve tracer is not enough to make the failure observable - the unit must first be set up in its normal bias condition. Furthermore, such failures are die-related most of the time, so FA requires more than just visual inspection of the device and package features.

Again, it is necessary to know the minimum set of conditions needed to make the failure mode observable for a reasonable amount of time. Many non-bonding-related intermittent failures also manifest with the application of extreme temperature (usually elevated) to the device. Other failures may become visible with the application of bias conditions bordering on datasheet spec limits, application of bias in the presence of a high temperature, or exposure of the die to light or darkness.

Once the analyst has determined these conditions and has set them up, the device should exhibit its failure and make the conduct of failure analysis on it possible. Failure analysis from thereon will simply be the same as conventional failure analysis for a permanent failure.