Standards, failures and reliability
A 200-component device that at 40 °C has a 92 % probability of working five years without failing drops to 66 % if it runs at 60 °C. Temperature is the first enemy of reliability.
01Standards and specifications
An electronic product is designed against a specification: what it has to do, under what conditions and for how long. A good part of it is set by standards, which make it possible to compare products and meet legal requirements:
| Body | Examples of standards |
|---|---|
| IEC | Electrical safety (IEC 62368-1), electromagnetic compatibility (61000 series), components |
| IPC | Printed circuit board design (IPC-2221), acceptability of assemblies (IPC-A-610), soldering (J-STD-001) |
| ISO | Quality management (ISO 9001), environmental management (ISO 14001) |
| JEDEC | Packages, moisture sensitivity and reliability tests of semiconductors |
| IRAM | Argentine standards, many adopted from international ones |
The specifications of a component distinguish the absolute maximum ratings, which must never be exceeded, from the recommended operating conditions and the guaranteed characteristics. Testing and certification were covered in testing and standards, in Electronic Measurements I.
02Failures and the bathtub curve
The failure rate of a population of components follows the bathtub curve: high at the start, because of manufacturing defects (infant mortality), low and almost constant during the useful life (random failures), and rising at the end, because of wear-out. Manufacturers eliminate the first part with burn-in tests (burn-in). The most common failure mechanisms are electromigration in metallizations, oxide breakdown, thermal fatigue of solder joints, corrosion and the drying out of electrolytic capacitors.
03Failure rate, MTBF and reliability
In the flat region, the failure rate \( \lambda \) is constant and the probability of working without failing up to \( t \) is exponential. In a series system, where any failure stops it, the rates add up:
With redundancy, two units in parallel where one is enough, the probability of failure is the product: \( R_{par} = 1 - (1 - R)^2 \). The failure rate depends on temperature according to the Arrhenius model:
A piece of equipment with N components of 10 FIT each at 40 °C and an activation energy of 0.7 eV. The curve is the probability of working without failing as a function of the years, without redundancy (blue) and with two redundant units (green).
The failure rates of each type of component are estimated with prediction handbooks, such as MIL-HDBK-217 or Siemens SN 29500, or with the manufacturer's field data.
04In the lab
Estimate the MTBF of a lab board by counting its components by type and using failure rates from a prediction handbook.
Measure with a thermal camera or thermocouples the temperature of the hottest components of a power supply and recompute the reliability.
In the datasheet of a regulator, separate the absolute maximum ratings, the recommended conditions and the guaranteed characteristics.
05Common mistakes
- Reading the MTBF as the useful life. It is the inverse of the random failure rate, not the time until wear-out.
- Designing at the absolute maximum rating. It is a damage limit; work with margins, derating voltage, current and temperature.
- Forgetting temperature in reliability. A few degrees more multiply the failure rate.
- Believing that redundancy solves everything. Common-cause failures affect both units at once.
06Self-assessment
What MTBF does a 5000 FIT piece of equipment have?
200,000 hours.
What probability does it have of working one year without failing?
\( e^{-8760/200\,000} \approx 0.957 \).
With Ea = 0.7 eV, how much does the failure rate increase going from 40 to 60 °C?
About 4.7 times.
What is infant mortality?
The high failure rate at the beginning of life, because of manufacturing defects; it is eliminated with burn-in tests.
07Further reading
- Patrick D. T. O'Connor and Andre Kleyner. Practical Reliability Engineering. 5th ed., Wiley, 2012. The practical reference on reliability, prediction and accelerated testing.
- United States Department of Defense. MIL-HDBK-217F: Reliability Prediction of Electronic Equipment. 1991. The classic handbook of failure rates by component type.
- Milton Ohring and Lucian Kasprzak. Reliability and Failure of Electronic Materials and Devices. 2nd ed., Academic Press, 2014. The physical failure mechanisms of electronic materials and devices.