Research On ICI Forming And Cancellation In OFDM Systems

Orthogonal frequency division multiplexing (OFDM) is one kind of multi-carrier modulation (MCM), and has been used in many fields. Its top advantage is the much higher frequency spectrum efficiency than the conventional frequency division multiplexing, which benefit from the orthogonality between the sub-carriers. The different sub-carriers overlap each other, and don't affect the receiving respectively. But the advantage also brings a hidden trouble, i.e., inter-carrier interference (ICI). With the development of the communication requirement, the number of sub-carriers in OFDM becomes bigger and bigger; the carrier frequency becomes higher and the moving speed becomes more rapid. These result in the interval of sub-carriers narrower and the level of ICI more serious. In this case, the orthogonality in OFDM becomes quite fragile. So, the effect of ICI cannot be neglected.This paper researches on ICI cancellation in OFDM system of mobile environment. The contents involve ICI self-cancellation, time-varying channel estimation, system equalization and joint multi-origin ICI cancellation. Concretely, the author's research and contributions are listed as follows.(1) First, the current ICI self-cancellation coding scheme based on polynomial is not optimal and lack of flexibility in coding efficiency. An ICI self-cancellation coding scheme with variable coding efficiency based on generator matrix is presented. And, at the instance with given coding efficiency, an optimal generator matrix in ICI canceling is designed by a new algorithm.(2) Two standards of designing optimum pulse in OFDM systems are presented: the effect of ICI reducing and the out-of-band leakage power. An optimum pulse model is established with given roll-off factor, and a pulse which has good performance in reducing the out-of-band leakage power and ICI simultaneously, is designed.(3) A time-varying channel estimation scheme based on particle filter is presented here. Firstly, channel impulse responses at the middle time of the adjacent OFDM symbol period are expressed by an AR model. Then, a state space model is established with aid of the pilot symbols, in which the ICI term is pointed out to be non-Gaussian variable after statistical analysis. So, the particle filter is used to estimate the channel in a linear and non-Gaussian state space model. Finally, cubic spline interpolation method is adopted to approximate the whole time-varying channel.(4) The idea of ICI self-cancellation is used to improve the performance of time-varying channel estimation. Firstly, the problem on pilot-aided channel estimation with BEM modeling is pointed out that ICI term of adjacent non-pilot symbols will decrease the accuracy of estimation. And then, these ICI coefficients from the symbols on the adjacent sub-carriers are found to be approximate. So, an ICI self-cancellation operation is presented to reduce the ICI terms at the receiver, in order to improve the effect of estimation.(5) A time-varying channel estimation scheme using superimposed training in OFDM systems is presented. The superimposed training sequence has the advantage of saving spectrum resources. But this scheme requires that the estimated parameter must be constants due to the utilizing of first-order statistics, which is conflict with the feature of time-varying channel. So, the time-varying channel during an OFDM symbol period is transformed to a series of time-invariant BEM coefficients by BEM modeling, which makes it possible to estimate time-varying channel using superimposed training.(6) A low-complexity decision feedback equalizer (DFE) is presented based on CE-BEM modeling. This equalizer directly constructs DFE utilizing the estimated CE-BEM coefficients. It successfully avoids the computation of recovering the time-varying channel from the CE-BEM coefficients and inverting a large-scale matrix in equalization. So, it decreases the computational complexity greatly.(7) A scheme of reducing ICI generated by time-varying channels and phase noise together is presented. When the time-varying channel is mixed with the phase noise, it is still a stationary narrowband random process. So, KL-BEM is adopted to modeling the mixed channels and phase noise, which makes it possible to estimate the time-varying channel and the phase noise jointly.