IEC 63287-1-2021 pdf download.Semiconductor devices – Generic semiconductor qualification guidelines – Part 1: Guidelines for IC reliability qualification.
Since ICs contain very small feature sizes and are dense with complex geometry, they are susceptible to defects generated in manufacturing processes. Therefore, some lots classified as ‘good devices’ contain potential failures with a minor intrinsic defect which does not influence measured electrical characteristics.
This allows the device to operate to specification during the sorting process. Removing potential failures before shipment and thus reducing or eliminating the early field failure rate is called screening. General screening methods to remove devices containing such minor intrinsic defects include the application of stricter stresses such as voltage and temperature than those under actual use conditions, combining a product test and burn-in, methods to remove initial defects in packages using X-ray inspections and visual inspections and combining mounting stress and stresses under a temperature cycle test, etc.
Screening conditions need to be adjusted when necessary depending on the target early failure rate. They should be examined in a manner that the screening itself will not influence the useful life significantly, i.e. reduce the time to wear out. By stabilizing the product manufacturing lines and manufacturing the product carefully to reduce intrinsic manufacturing defects, the product early failure rate can also be reduced. Burn-in has generally been conducted after packaging, but recently, it is common to conduct it to die on wafers. Both methods will produce the same effect in terms of the purpose of removing manufacturing defects in device die. Figure 5 shows the relationship between the screening and the early failure rate which can be statistically estimated.
Figure 5 shows the verification method for the early failure rate after screening which can be estimated statistically. In some cases, burn-in is repeated until no actual burn-in failure occurs to confirm the convergence of the screening but, as shown in Figure 5, the early failure always occurs even at a slight rate when it is statistically estimated in the Weibull distribution. . For this reason, the sampling becomes meaningless unless data is statistically analysed taking account of the experimental parameter, acceleration factor, and failure mode even when the burn-in is unreasonably repeated until no burn-in failure occurs. The m value calculated with the failure probability of the burn-in process does not reflect the failures which occurred and were removed in the test processes before the burn-in process. If the m value is calculated without considering failures in the test processes before burn-in, it . becomes larger.
Therefore, the early failure rate estimated from the m value tends to become a poorer estimate than the actual quality. In general, it is necessary to calculate the m -value taking account of the failure probability before and after burn-in as shown in Figure 5 to estimate the early failure. Verification of the early failure distribution and rate based on this concept should be incorporated into the verification of the product reliability, which is also the most effective means to improve the actual delivery quality. Since the early failure depends on the intrinsic manufacturing defect rate of the manufacturing line, the early failure rate of other products in the same manufacturing line and with the same design standard can also be estimated easily as long as the manufacturing defect rate and the convergent characteristics (m -value of the Weibull distribution) of the early failure are understood.IEC 63287-1 pdf download.