The importance of Test Strength Modelling in defining a new Reliability Stress Test profile

The importance of Test Strength Modelling in defining a new Reliability Stress Test profile Engineers are too often ‘forced’ to apply Mil-Std 217 or similar to define simple Accelerated Life tests with a target of 0 defects to minimise test time. We all know this is quite a ridiculous and meaningless approach as the engineer’s role is to stimulate Latent defects with an organised style of Accelerated Stress test and be able to then assess / predict failure level. To achieve this aim in electronic product testing we of course must consider the Early Life and Longer Life failure distributions and apply different styles of stress testing to stimulate quite different failure mechanisms. Engineers often struggle with this unless they have sufficient experience of doing this and designing different forms of stress tests For the Early life Stress Test we are looking to find latent defects which are most often process related and are result of escapes, while longer term life defects are generally more design and component quality related. Hence the stress tests to stimulate the defects need to be very different, but how long should our accelerated stress tests last ? The best way to answer this question is to consider using the Test Strength modelling , equations are available from Mil-Hdbk-344a, but are often considered to provide too high a Test Strength for the stresses imposed which therefore leads the inexperienced engineer to believe 2 or 3 fast ramp thermal cycles can detect 90% of the latent Early Life defects !! Applying the more conservative ‘Hughes’ Test Strength equations limits the Test Strength levels and makes their calculation much more realistic. While with IBM we refined and used this model to successfully plan Early Life Reliability Tests and create a benchmark for all test planning / execution. This enabled setting a Test Strength target (generally 80-90%) to define the optimum Reliability Stress Test that would have an 80-90% probability of ‘stimulating’ a latent defect . hence providing strong confidence in the Early Life Reliability test. For example; 30 thermal cycles between -20 and +70 degC, with average temperature change rate of 8 deg C per minute, Test Strength = 0.68 If only 10 thermal cycles with a 5 deg C change rate, Test Strength = 0.26 Test Strength modelling should NOT be confused with Acceleration Modelling as we are not trying to predict failure rates, simply stimulate defects. To go one step further and predict failure rates from the test would of course require application of appropriate acceleration model such as Coiffon Manson, Arrhenius, etc In summary, the Test Strength approach is an excellent way to ensure all Early Life stress tests are planned and evaluated in similar manner. Hence comparison of results becomes possible and the reliability engineer can have confidence in the test, whether it finds defects or not If you wish to find out more, contact Martin Shaw of Reliability solutions at reliabilitysolutions@yahoo.co.uk or visit the website; www.reliabilitysolutions.co.uk

For more information on The importance of Test Strength Modelling in defining a new Reliability Stress Test profile talk to Reliability Solutions