Parametric Reliability Analysis In DfRSoft

 Return Home Page

Parametric Reliability AnalysisParametric Reliability Analysis

Exclusive from DfRSoft - Parametric Reliability Analysis. Assess device degradation failure rates over time without having to test to failure and with more accuracy saving hours of test time, at lower stress and with smaller sample sizes. Here is how:

There are three ways to assess the failure rate of a device that drifts out of specification. As an example we will do Power loss for a device.

Each way requires a Parameter Limit, such as Power Limit. Let say 20% Power loss.

Method 1: Consider a power limit of 20%. Devices are on life test and when they reach 20% Power Loss, this is treated as the time to failure. In this method a catastrophic analysis is used. Use for example the Rel plts page to do a Weibull assessment. Put the times to failure for each device and perform the analysis similar to catastrophic data analysis. In some cases this may be the best method when an aging law is not known.

Method 2: Initial and Final Parametric Measurement Method:
In this method we require less test samples since we can characterize the devices distribution with say 20 or 30 samples. Typically device may be on life test at an elevated temperature and we simply take two measurements at room temperature - before and after the test. We have a population of samples and we find its mean and sigma. At this point we can use the parametric method where we can often prove a more accurate failure rate at a specific confidence level than say a catastrophic analysis, even when no failures occur. In addition we can get a point estimate (i.e., no confidence bound data estimate) which cannot be done when 0 failures occurs in the catastrophic analysis method. To do this follow the hyperlink below for Initial and Final drift measurement analysis.

Method 3: In this method, if a general aging model is known i.e. how the drift for power loss occurs over time, then Parametric reliability analysis makes the most sense and has tremendous advantages over Method 1 & 2.

For example, the three aging laws in DfRSoft are:
Parameter Shift=A (Time)^B ,
Parameter Shift = A Log(1+B Time)
or Parameter Shift = User defined.
DfRSoft allows time to be accelerated with an acceleration factor. An Arrhenius calculator is available to determine this factor if temperature is the acceleration stress.

If the Aging Law is known then there are three major advantages over Method 1:
1. We require less test samples since we can characterize the devices distribution with say 20 or 30 samples.
2. We do not have to test to failure only over a sufficient time to fit the data to the aging law. With knowledge of the ageing law and its distribution, the reliability can quickly be predicted at any temperature.
3. We can test at a lower stress like a lower temperature since we do not have to test to failure. Often the closer the temperature to the use conditions for example, the more accurate will be the prediction to the extrapolated temperature of interest.