Fast Simulations Of Turbo Codes Based On Importance Sampling

In1993, the French scholars, C.Berrou, put forward anew kind of codes called Turbo codes which was produced by using two Recursion systematical convolution encoders connecting with an interleaver. Owing to its superior performance closed to Shannon limitation, Turbo codes were paid attention to widely. From then on, estimating the performance of the Turbo codes becomes a very important research direction. However, Owing to interleaver is used in Turbo encoders, the closed form analysis of Turbo coder systems is intractable. Their performance evaluation is traditionally relied upon upper bounds which are quite loose in some cases, which makes the mathematical analysis of bit error rate (BER) of Turbo codes very difficult. The computer simulation becomes an important way of estimating the performance of Turbo codes.Good results of estimating BER of communication can be obtained under the condition of low signal-to-noise ratio (SNR) by using traditional Monte Carlo method. But under the condition of high signal-to-noise ratio and requirements of certain accuracy, Monte Carlo simulation requires a great amount of data. Under the base of the study importance sampling simulation technology of communication system, according to the characteristics of Turbo codes, a method pointed to specific decoding errors in Importance Sampling simulations is used to estimate BER of Turbo codes. the key of the method is how to design the probability density function(pdf) used in simulation according specific decoding errors, in order to decreasing the variance of simulations, as a result,to reduce the amount of data required in simulation at the same precision.Finally, for some Turbo codes systems, respectively on AWGN and on Rayleigh fading channel, respectively different interleavers, results from MC simulation and IS simulation pointed to set including specific decoding error set are compared. So as to demonstrate that time consumption of IS simulation pointed to specific decoding errors under high signal-to-noise ratio and at the same accuracy is far less than the time consumption of MC simulation.