Both MTTF and MTBF are standards for predicting reliability life. They are different in a way that MTBF is for products that are recoverable or expensive in the unit cost (e.g., computers, airplanes, cars), while MTTF is used for products that are irrecoverable or cheaper in the unit cost (e.g., ICs, semiconductor components)

• MTBF (Mean Time Between Failure) refers to the average interval between failures. It is often used for system products and is usually applied to find the average operating time between two consecutive failures.
• For example, a computer server is a system integrating multiple products, including the power supply unit (PSU), motherboard (MB), hard disk drive (HDD), graphics card (VGA), memories (RAM) and others. If the PSU fails, the computer server (system) will also fail ( stop working). After restoration and operating for some time, the system server can fail again due to a RAM anomaly. In this case, the interval between the two failures (PSU and RAM) can be called the MTBF of this computer server.
• MTTF (Mean Time To Failure) refers to the average time to failure. It is different from MTBF in that it is often used for irrecoverable products, such as ICs, whose designated functions after failure are irrecoverable through maintenance and repair.
• For example, an IC will fail after long-time use due to electromigration and ageing and this IC cannot be restored through maintenance and repair and must be replaced directly.

After the iST lab helps you run HTOL, your end-customers often ask you to calculate the MTTF and λ values. How to calculate them?

We can calculate these values with reference to the formula in the JEP001-3A provided by the Joint Electron Device Engineering Council (JEDEC).

In the following example, the lab conditions are as follows:

After the experiment, if the sample’s failure count is 0 with the confidence level of 90%, the average failure rate λ and reciprocal MTTF are:

Continuing with the above calculations, how to use them? Are they good or bad?

First, we need to understand the definitions of λ and MTTF. It is very difficult to calculate the short-time failure rate and it is also less meaningful. Hence, we prefer to calculate the average failure rate in a given time interval, e.g., 1,000 hours/one month or one year as the time unit. In general, the results are presented this way: Average failure level (λ) = Total failures in the interval / Total operating hours (Total operating hours = total samples x operating hours). As the failure rate is very low, we will express it in failures in time ( FIT = 10-9). MTTF is the reciprocal of λ, i.e., the average time leading to the failure. However, please note that instead of hours, Device Hours is the unit for MTTF.

Then, with reference to the calculation of cumulative distribution function (CDF) in JESD74A:

After inputting the required prediction time in this equation, we can calculate the predicted group cumulative failure rate at different times.

Figure 1 below shows the predicted group cumulative failure rate at different periods. For example, the group cumulative failure rate is 1.2% in 5 years and 2.39% in 10 years. With these values, you can compare if a product meets the requirements of end-customers or the level of acceptance of your company.

After understanding the conversion among MTTF, λ and cumulative failure rate and receiving the testing requirements of end-users, we can design suitable testing conditions based on these requirements.

In the following example, the required result is a failure rate below 1% for 10 years (24Hrs a day).

First, after calculating CDF,

we can see that λ≦114 FITs will meet the requirement.

Based on this target, we can match with other experiment conditions, such as how long the test takes, how many samples the test needs, the voltage and temperature required to boost stress to what extent, the number of accepted failures, the targeted failed model…

Table 1 below shows the three sets of reliability test conditions. The outcomes of each set can meet the requirement λ ≦ 114 FITs, i.e., the initial requirement of end-customers. The adjustable parameters are composite. Fine-tuning based on the ambient conditions is advised in designing reliability experiment conditions to design the experiment conditions that meet the ambient requirements.