Time-Varying Multi-Stress ALT for Modeling Life of Outdoor Optical Products

The long-term performance prediction of optical products deployed outdoors is very important. The accuracy of the prediction directly affects the warranty cost calculations. In the indoor environment, products are usually operated under the controlled temperature and humidity. However, in the outdoors, products usually experience multiple stresses that continuously vary with time. An outdoor use optical product, such as a fiber distribution hub cabinet with splitters, connectors and fan-outs built-in, will experience temperature cycling effects plus varying humidity. In unusual circumstances, it will also experience vibration and shock. Therefore, the time varying temperature and humidity are considered as the major stresses acting on the outdoor use optical products. In this paper, a practical method which uses design of experiment (DOE) techniques and generalized log-linear (GLL) life-stress relationship to predict the life of outdoor optical products is proposed. Considerations are made to include the mating and un-mating of connectors, and environmental vibration. The temperature effect is defined by the following settings: 1) High temperature, 2) Low temperature, 3) Ramp rates, 4) Cycle length and 5) Dwell time at high and low temperatures. These settings are used in the accelerated testing to mimic the temperature cycle in the outdoor environment. Humidity is also applied during the test. The individual effect of each stress and their interactions is studied by using multiple levels of each stress, analyzed using DOE methods. Using the accelerated testing results obtained from the laboratory, the relationship between stresses and product life is established, and the life distribution model can be constructed. Subsequently, the 20-year reliability of the products under different outdoor environments is predicted. Since the temperature and humidity change seasonally, the weighted probabilities are used in the calculation of the 20-year reliability. Based on the predicted reliability, the warranty costs can be easily calculated