38 The measured pressure sensitivity depends on input polarization of light and examined output polarization. It changes from 0,45rad/MPa*m to 4,2rad/MPa*m for two different fibers with change of input and output light polarization. Measured temperature sensitivity depends on input polarization of light and examined output polarization too. It change from 3rad/K*m to 12 rad/K*m for two exemplary fibers. The change of K T for different h indicates the existence of maximum. Its determination requires of further experimental investigations. Probably the shape of dependence of K T on h is caused by stress profile inside optical fiber in spite of small numerical aperture. We hope that the planed experimental investigation of the analogous dependences in the dual core photonic crystal fibers may explain this problem. 1. Woliński T.R. Polarimetric Optical Fibers and Sensors in Progress in Optics Vol.XL, ed. E.Wolf (2000), 1-75. 2. Friberg S.R., Silberberg Y., Oliver M.K., Andrejco M.J., Saifi M.A., W.Smith P. Ultrafast All- Optical Switching in a Dual-Core Fibre Nonlinear Coupler, Appl. Phys. Lett. 51, (1987), 1135-1137. 3. Gang-Ding Peng, T.Tjugiarto, P.L.Chu. Twin-core optical fiber with large core ellipticity, Appl. Opt. Vol.30, No6, (1991), 632-634. UDC 621.382 A.S.Andonova, N.G. Atanasova FETT, Technical University of Sofia, Sofia, Bulgaria TESTING ALGORITHMS FOR SCREENING OF LARGE ELECTRONIC SYSTEMS  Andonova A.S., Atanasova N.G., 2004 When a hardware system is screening, a problem is when to stop the test and accept the system. Based on this these, the paper describes and evaluates seven possible algorithms. Three of these algorithms as most promising are tested with simulated data. Different systems are simulated, and 50 Monte Carlo simulations made on each system. The stop times generated by the algorithm is compared with the known perfect stop time. Of the three algorithms two is selected as good. These two algorithms are then tested on real data. The algorithms are tested with three different levels of confidence. The number of correct and wrong stop decisions are counted. The conclusion is that the Weibull algorithm with 90% confidence level takes the right decision in every one of the cases. 1. Introduction When performing a run-in or acceptance testing on a large hardware system, it is often a problem to decide when to stop testing [1,2]. If it is stop too early the system will be delivered to the customer with too many early failures. On the other hand testing is very expensive, and the test can delay the delivery. For hardware screening exist the same problem. A stress screening process will in the beginning precipitate many early failures per hour, but the last failures take a lot longer to precipitate. In the IEC standard IEC 61163-1 “Reliability stress screening of repairable items produced in lots” [3], the problem is solved by accepting that a sample of the product is run for an extended screening period, in order to find the optimum duration of the screening process. This duration is expressed as a failure free period. The aim is to stop the screening process as soon as the curve of the accumulated failures per 100 items levels out i.e. the curve converges towards a straight line as can be seen in fig. 1. Lviv Polytechnic National University Institutional Repository http://ena.lp.edu.ua Lviv Polytechnic National University Institutional Repository http://ena.lp.edu.ua