Burn-In and Stress Screening
What Burn-In Is For
Burn-in subjects every unit to a period of operation at an elevated stress, usually temperature, in order to precipitate the defects that would otherwise appear in the first weeks of the customer’s use. It works on the premise that a population has a higher failure rate early in its life, and that the weak units can be removed before they are shipped. It is an expensive step, since it occupies time, equipment and floor space, and the decision to use it should follow from a reliability requirement rather than from a habit or a customer’s default expectation.
The Bathtub Curve and Its Assumptions
The justification for burn-in is the early failure region of the reliability curve, where the failure rate is higher than in the middle of the life. The mechanisms are usually manufacturing or material related: a marginal joint, a particle in a device, a contamination, a component that was weakened during handling. Burn-in precipitates those, so that the unit fails on the line rather than in the field. The assumption is that the population has such defects and that the stress accelerates them; where the process is mature and the defects have been eliminated, the burn-in may find nothing and its value falls.
Choosing the Stress
The stress is chosen so that it accelerates the relevant mechanisms without introducing new ones. Temperature is the most common, and it is usually applied with the unit powered and cycling so that the thermal expansion and the electrical load act together. Voltage stress is used for a component that has a wear out mechanism related to the field. Vibration and thermal cycling are used where the mechanism is mechanical. The level has to be within the product’s design limits, because a burn-in that damages the product is worse than no burn-in. The duration follows from the acceleration factor and the intended screening strength rather than from a round number.
Monitoring During Burn-In
A burn-in that does not monitor the unit is a burn-in that can only find the units that fail completely. Continuous monitoring of the critical parameters shows a unit that drifts or a function that stops intermittently, which is often the sign of the defect that the screening is meant to find. The monitoring should cover the functions whose failure would matter, and the data should be recorded so that a drift can be seen rather than only a binary pass or fail. Where the monitoring is a simple power check, the screening is weaker than it appears.
Screening and Its Economics
The economics are a balance between the cost of the screening and the cost of a field failure. A screening that removes a few units per thousand has a value that depends on what a field failure costs, which for an unreachable installation or a safety related product is very high and for a consumer accessory is low. The calculation should be explicit rather than assumed, and it should be revisited as the process matures, because a screening that was justified when the product was new may not be justified after the causes of the early failures have been eliminated. Removing an unnecessary screening step is a legitimate improvement.
Alternatives and Complements
Burn-in is not the only way to reduce early failures. A process that is under control, a supplier that is qualified, a design with margin and a test that covers the critical functions all reduce the population of weak units before the screening step. Where the failures come from a specific mechanism, the correct answer is often to remove the mechanism rather than to screen it out. A high temperature operating life test on a sample is the way to identify whether the mechanism exists, and the screening is then a decision about how to handle the units that may contain it.
Records and Traceability
The burn-in’s result is part of the unit’s record. The conditions, the duration, the monitoring data and the failures should be recorded against the unit or the lot, and the failures should be analysed rather than simply discarded, since the pattern across a batch is what identifies the mechanism. Where a unit fails during the screening, its failure should be investigated in the same way as a field failure, because the cause is the same and the diagnosis is easier while the unit and its process data are at hand.
Screening Without a Burn-In Chamber
Where a full burn-in is not justified, other steps can capture part of its value. An extended functional test that exercises the unit for a period under load, a thermal soak at the high end of the operating range, and a cycling of the interfaces all precipitate a proportion of the early defects at a lower cost. The value of those steps comes from the same principle, which is that a unit which is going to fail early can be encouraged to fail before it is shipped. The choice between them is a matter of what the failure mechanism needs, and the evidence should come from an analysis of the failures that the product has actually experienced.
FAQ
Why burn in a product? To precipitate the early failures that would otherwise occur in the customer’s first weeks of use.
How is the stress chosen? From the mechanism to be precipitated, at a level within the product’s design limits and for a duration derived from the acceleration factor.
Why monitor during burn-in? A simple power check only finds complete failures; monitoring the functions finds the drifting and the intermittent defects.
Is burn-in always justified? No. Its value depends on the cost of a field failure, and it should be reviewed as the process matures.
What should happen to a unit that fails? Its failure should be analysed like a field failure, since the cause is the same and the data is at hand.
Conclusion
Burn-in precipitates early failures at a cost, so choose the stress from the mechanism and review the screening as the process matures. Analyse the failures. Reliability screening belongs to PCBA testing, the process it screens is described in PCB assembly, and the acceptance is part of quality management. Screening for a new product is defined during prototype PCB assembly in 2026.
