800 surface-mount solder joints were test with repeated thermal shock cycles. The manufacturer’s field failure data indicates that the B(1) life* is 25 months. Estimate the duration (in days) each thermal cycle can simulate with 80% confidence level.
A solder joint is considered failed if its resistance exceeded a certain value (The scenario is simplified for the purpose of this presentation). The measurements for all the solder joints were done every 25 thermal cycles. The experiment stopped at 150 cycles and 780 joints were found working.
Following was the test results.
*Note: B(1) life => The time by which 1% of the population fails.
The 1st row in Figure 1 indicated that the 1st joint failed between 25 and 50 cycles. We only know the cycle interval in which it failed, but not the exact cycle. This is known as Interval Data Type.
To enter Interval censored data, from the main panel, set the Data Type to Interval Time. An additional column is added in the worksheet to accept Interval censored data.
The analysis assumes time-to-failure of the solder joints obey Lognormal distribution.
Distribution was set to Lognormal.
For dataset with censoring the preferred Analysis method is MLE.
Note that the Unit remains in Hour, which mean 1 hour is equivalent to 1 cycle.
To add an 80% Two-Sided-bounds on the plot, click on Confidence Bounds button.
Redraw the plot.
The 80% 2-Sided bounds B(1) Life (also correspond to 25 months) are 92.3 and 121.5 cycles respectively. Hence, 1 thermal cycle is equivalent to, 6.2 days (25*30/121.5) to 8.1 days in the field with 80% confidence level.
By comparing the field reliability data and lab test life data, the manufacturer quantified the effect of thermal cycle to field operating time ratio. The manufacturer could use this information to estimate the product field life through thermal cycle test, and hence avoid the much more expensive Accelerated Life Test (ALT).