In order to conduct quantitative failure analysis, estimates of the random failure rates for the components that will be considered in the analysis must be generated. There are many different models and techniques that can be used for failure rate estimation. Neither IEC 61508 nor ISO 26262 mandate the use of a particular failure estimation methodology. Estimation methods commonly used include:

• IEC/TR 62380:2004, "Reliability Data Handbook - Universal Model for Reliability Prediction of Electronics, PCBs, and Equipment"
• Siemens Norm SN29500:2010, "Failure Rates of Components"
• IEC 61709:2017, "Electric components - Reliability - Reference conditions for failure rates and stress models for conversion"
• Supplier reliability data from similar products already in production and deployed under similar operating conditions
• Targeted studies and experiments that seek to induce failures on silicon under conditions that simulate accelerate lifespan (such as temperature, voltage, frequency, vibration, humidity, or radiation exposure).

Estimations of failure rate are often defined in terms of Failures In Time (FIT). TI's data respects FIT in terms of failures per 10^9 hours of operation, as is consistent with most handbooks. However, certain handbooks, such as those for military applications, may refer to FIT based on failures per 10^6 hours of operation. Take care when using such data to respect a common definition of FIT in all calculations.

In TI's experience, all of the models generate estimations of failure rate that are not consistent with failure rates which are observed and reported in the field or predicted based on data generated from targeted experiments. The models consistently predict higher failure rates than those observed in the field or predicted via targeted experiments. One possible reason for this discrepancy is that these standards consider reliability data that does not make a distinction between random and systematic failure. In both IEC 61508 and ISO 26262, the focus for quantitative analysis is on random failure rate. TI's data indicates that the vast majority of field failure issues seen in semiconductors are due to systematic failures, whether traced to semiconductor supplier, system integrator, or end user. TI has quality and reliability programs in place that constantly improve our products and processes to reduce these systematic failures.

The failure rates derived from SN29500 tend to be conservative as compared to TI product field failure rate data or TI accelerated lifetime testing. TI considers the IEC 61709 to be similar to the SN29500 and we refer to this model as the IEC 61709/SN29500 model in the FMEDA. The IEC/TR 62380, while still conservative, provides the closest match available to TI product data. Although this standard has been formally withdrawn, the equations have been incorporated inside ISO 26262-11:2018 section 4.6.2. As such, TI has used IEC/TR 62380 as the basis for our random failure rate estimation, augmented with data from targeted studies for failure modes not considered in the base model.

When considering failure rates for semiconductors, TI applies the following partition and methodology: