It is generally accepted that over 97 percent of all product failures will occur in the first 24-hours of operation. Product burn-in “infant mortality” in finished goods is one of the biggest problems facing electronic product manufacturers. Burn-in plays a critical part in ensuring products delivered to the market are reliable and will not fail.

200,000 of a given product manufactured with an infant mortality rate of 0,5 percent means that 1,000 units could be expected to fail in the field.  Burning in for 24 hours reduces this figure by 97 percent or more giving an expected field failure of only 30.  “This is a vast improvement from 1,000 failed units,” explains Murton System Technologies MD Brent Maurer, Turnkey Manufacturers of electronic products, sub-assemblies and customized development projects. After this, manufacturers experience diminishing returns for the length of burn in time.

Semiconductor manufacturers, independent test labs, end-users and Turnkey Manufacturers normally perform burn-in.  Sample units of new devices or new lots are burned in to provide reliable data for the remaining devices.

At the component level, burn-in is done by loading devices into high temperature sockets, which make temporary electrical contact with the device leads, and are mounted on high temperature circuit boards with circuitry to provide the proper voltages and stimuli to the device.  The devices are isolated from one another with passive components, which limit the current each device can draw.

The boards containing the devices are then loaded into a convection oven which elevates the temperature of the device and provides an electrical interconnect to the power supplies and signal generators.  The devices remain in the oven for an amount of time determined to induce failure.  The rate that these devices fail is called the infant mortality rate.

Most ICs consist of ion junctions implanted on silicon substrate.  The junctions are connected with small traces of conductive oxides.  The ambient heat and the heat caused by current flow causes the junction temperature to rise.  This causes ion 'clouds' which surround the junctions which dissipate from their original locations.

These stresses can lead to premature failure of weaker devices.  They can also lead to the failure of non-marginal devices, if the junction temperature exceeds the manufacturers’ maximum rating.  At the "end of operational life", the heat of burn-in has caused the ions to dissipate to the point where the devices no longer functions.

There are two types of burn-in, lot sampling and operational life testing.  In burn-in, all of the devices of a particular type are burned in to 'weed out' the weak devices. Lot sampling is a technique where a small control sample is burned in and the statistical data derived from testing those devices is used for the remaining devices of the lot.

“Different types of burn-in are designed for different types of devices.  We use static-burn-in as well as dynamic burn-in, both during the design phase and production phases, and test during as well as after burn-in for device system deviation tolerance and failure.  This enables us to ensure that we are able to ship goods of consistently high export quality and that potential design flaws are detected during burn-in and rectified,” said Maurer