Turbo Code Interleaver And Decoding Structure

The thesis researchs into the interleaver and iterative decoding structure of Turbo code. The Turbo code is the channel coding of third-generation mobile communications (3G). In Turbo code system,the interleaver can decrease the pertinency between information bits, so that the outburst errors are separated. The encoding performance of Turbo code can be grately improved by the existence of interleaver. The iterative decoding structure is a characteristic of Turbo decoding. However, Turbo decoding algorithm and long time delay of iteration process restricted its application to the high rate communication systems which require high level of real time performance. Therefor, the improvement of Turbo code focuses on the interleaver and iterative structure.Firstly, the thesis presents a symmetrical interleaver algorithm which adopts T-Chaotic mapping and odd-even protection (TCOP). The odd-even protection and symmetrical characteristic are emphasized, which are principles of interleaver design. Three kinds of symmetrical methods are presented. The odd-even protection and symmetrical characteristic are used in T-Chaotic interleaver. Chaotic mapping is used to improve capability of controllable randomicity and reduce the relativity of the algorithm based on odd-even protection. The symmetrical design based on odd-even protection improves the equal protection of check sequence and predigests the system structure.Three kinds of TCOP interleavers are compared with S-Chaotic interleaver and SCOP in delay and BER performance. The S-Chaotic interleaver can increase the distance between the bits and bits interleavered, so that the weight of code is increased. However, the disadvantage of S-Chaotic interleaver is very obvious. The position map of S-Chaotic interleaver is selected artificially. So the random of code is destroyed. Otherwise, the restriction distance of S-Chaotic interleaver presented is. But it is nearly impossible that S reach. Even if the S is comparatively small, the time of generating position map{α_i} is more than method of compositor. In the relay experiments, the S-Chaotic interleaver is 8 to 676 times longer than TCOP.2 interleaver. And the time of SCOP interleaver is 8 to 118 times longer than TCOP.2 interleaver. In the BER experiments, SCOP interleaver compares with S-Chaotic interleaver at the condition that the frame is 1024; the iterative times is 5; the number of error frame is 15. The results indicate that odd-even protection improves the equal protection of check sequence and the performance of SCOP interleaver improves about 0.2dB. In the comparement experiments of three TCOP interleaver, the bits distance of TCOP.2 is equal. The results indicate that the performance of TCOP.1 interleaver and TCOP.3 interleaver are close. And TCOP.2 interleaver is better than TCOP.1 interleaver and TCOP.3 interleaver. The performance of the TCOP.2 interleaver is close to SCOP interleaver and improves about 0.3dB to S-Chaotic interleaver.To sum up, TCOP.2 interleaver can reduce the time delay on 1-2 orders of magnitude. While the BER proformence of TCOP.2 is not down. At the same time, the symmetry structure can reduce the complexity of system implementation. Moreover, while TCOP.2 interleaver fixes some variable, such as 2 chaotic initial value, 3 the selected value of symmetric algorithms, the frame can be changed and symmetric algorithm at interleaver can be switched. So it is easy to achieve variable-depth interleaver.Secondly, the iterative decoding structure of Turbo codes is researched in this paper. The iterative decoding algorithm is the most characteristic of Turbo code decoding. The estimative value of exterior information could be more reliable after each iterative process. Therefore, the decoding performance could be improved greatly. On improvements of iterative structure, the sliding window[27] as well as the parallel decoding structure[44] are proposed to reduce hardware storage requirements and to improve the decoding efficiency. This paper introduces the linear prediction control algorithm, to improve the iterative structure of Turbo code decoding. A decoding structure of Turbo code based on parallel prediction control (PPC-Turbo code) is presented. According to H.E. Gamal and A.R. Hammons theory[60], the component decoder n-order polynomial linear model has been set up. The information of front n iterations are saved in the linear model. The value of№(n+1) exterior information of the component decoder can be predicted using the polynomial curve fitting or index fitting algorithm. The№(n+1) iterative decoding process will be replaced by the prediction value of exterior information. Compared to the traditional decoding algorithm, the computation quantity and computing time of the Parallel Prediction Control Algorithm (PPCA) will be smaller. Therefore, the complexity and delay of the decoding algorithm can be reduced by the new decoding scheme presented in the paper.For linear prediction control module, simulation tests of prediction values, prediction time and BER performance of PPC-Turbo were made.Each external information of decoding is the most representative of the effect of iterative decoding, so the external information at different iterations are chosen. The single prediction 6⁹th and compound prediction 6+8th,7+9th exterior information are taken to compare with the external information of component decoder. Adopting single prediction, exterior information value can show a perfect regression. As to compound prediction, along with a comparatively large fluctuation, there is a regression tendency. It has been proved that the linear prediction control module can replace the traditional structure of the component decoder.For the time delay compare examinations, the results indicate that when the frames are 1024, 512, 256, 128, t_dec1is higher than t_pc. When the frame is 64, t_dec1and t_pc are in a dynamic range. Therefore, although the linear prediction control module has not manifested the superiority for the short frame, it can show obvious superiority for the long frames.For the examinations, that the 6-9th prediction of PPC-Turbo compare with the traditional Turbo code, the results indicate that the change tendency between PPC-Turbo and Turbo curve is closer and BER curve is coincidence while predicting the 7,9th . Totally, the decoding structure of Turbo code based on parallel prediction control is 0.1dB lower than traditional Turbo code in BER performance. However, the delay has been reduced.Lots of research has been done in application of Turbo code. The scheme of symmetrical interleaver based on T-Chaotic mapping and odd-even protection and decoding structure of Turbo code based on parallel prediction control could reduce the delay in application to Turbo code. They enlarge the range of Turbo code application and accelerate the speed of Turbo code application.