burn-in
The process by which components of a system are exercised prior to being placed in service (and sometimes before assembly), forcing certain failures to occur under supervised conditions.
burn-in: controlled failure under load before shipment
Burn-in is the deliberate operation of equipment under specified conditions, often elevated temperature and stress, to expose latent defects before the product reaches a customer. A newly manufactured electronic device, mechanical assembly, or subsystem is powered up, loaded, and run continuously for a period ranging from hours to days. Failures that would otherwise occur in the field are forced to happen in the factory, where they can be fixed or the component discarded.
The practice rests on a fundamental principle of failure physics: defective units often fail early, within a characteristic "infant mortality" window. A semiconductor with a shorted junction, a bearing with a surface defect, or a solder joint with a cold connection will typically fail under thermal and electrical stress much sooner than during normal operation. By compressing that failure timeline into the controlled environment of manufacturing, a producer eliminates the cost and reputation damage of warranty returns.
Temperature and electrical stress are the workhorses of burn-in
Most burn-in procedures elevate the ambient temperature to between 50 and 75 degrees Celsius, sometimes higher, while the device operates at or near full rated load. Electrical components may be cycled on and off, or exercise test software may be executed to stress processors and memory. Mechanical equipment might run at rated speed or torque. The duration varies: consumer electronics often undergo 24 to 48 hours; high-reliability military or aerospace hardware may burn in for 168 hours or more. Some manufacturers perform burn-in twice, before and after final assembly, to catch assembly-induced failures.
The cost of burn-in is real. It consumes energy, ties up capital in inventory, requires dedicated facilities and monitoring, and extends lead time. When a unit fails during burn-in, diagnosis and rework add further expense. This tension shapes decisions about where burn-in occurs in the supply chain. A semiconductor manufacturer typically burns in wafers or packaged dies; an integrator may burn in subassemblies; an OEM may perform system-level burn-in. The choice depends on the defect rate history of supplied components, the cost of failure in use, and contractual obligations.
Burn-in is most common in electronics manufacturing but also appears in the production of compressors, pumps, gear drives, and other components where early-life failure risk is significant and replacement cost is high. The alternative to burn-in is field support and warranty management, acceptable only when the expected defect rate is very low or the cost of failure in service is tolerable. A light bulb manufacturer does not burn in every lamp; a semiconductor supplier serving automotive customers often does, because an infant-mortality failure in an engine control unit can disable a vehicle in traffic.