Study Of Channel Estimation And Synchronization Technology In Filtered Multitone Modulation

A future wireless broadband communication system will necessarily be a high speed and large capacity system that can not be supported by all existing communication technologies. In this case, this paper researched FMT technology, one of multicarrier modulation technologies, where channel estimation technology and synchronization technology are two mainly researched aspects.Firstly, this paper researches an efficient implementation of FMT systems and analyzed FMT signal in detail. A method with mathematics reasoning and graph analyzing is first used to show how to transform an irrealizable theory structure into an efficient implementation structure. The efficient structure is based on IFFT/FFT and a polyphase structure of a prototype filter. And then FMT signal is first analyzed in time domain and frequency domain, which explains the relations between a prototype filter, a multipath channel and three components of FMT signal including the useful signal, ISI and ICI.Secondly, this paper researches FMT channel estimation methods in time domain. Because the existing time-domain estimation algorithms are biased, a floor effect appears when SNR is high. Therefore, two algorithms are presented in this paper. The one is a modified unbiased estimation algorithm based on PN code, and the other is a estimation algorithm based on a new sequence whose correlation with PN code is an uniform function. Compared with the existing algorithms, these two algorithms are both unbiased and have performances independent of channel characteristics. Their estimation precisions are higher. And the second algorithm can be realized more easiler for its lower computing cost. Because the limited amount of PN code results in a small selected range, this paper presented an averaged short period channel estimation algorithm based on OCI (Optimum Channel Independent) sequence. OCI sequence is periodic and orthogonal with an adjustable period. The computing cost only approximates the value of the sequence's period by virtue of IFFT/FFT, which is lower than the first above algorithm. This algorithm also gains better performance than the above two algorithms. To avoid a negative effect on estimation performance from some small ray gains, this paper also presented methods for estimating the noise power. A method of restraining noise, based on the estimated noise power, is used to improve the performance of channel estimations.. These methods set the small ray gains undistinguishable from noise to zeros. They reduce the effect of the small gains on estimations and improve estimation performance when SNR is low.Thirdly, this paper researches a estimation method and a channel equalization method in FMT frequency domain. When a channel estimation is implemented in time domain, time pilot data will be frequently inserted. And when space-time joint channel estimation is adopted, a high complexity is also inevitable. To overcome the above two defaults, FMT frequency pilot assisted channel estimation algorithm is presented. When pilots are inserted in frequency domain, signals in corresponding subcarriers are expressed as a weighed frequency response of channel where weights are determined only by the prototype filter and the matched filter. This algorithm shows a good estimation performance that helps FMT to get a good system performance. It is known that operations including matrix inverse or pseudoinverse are inevitable when the FMT signal is equalized by the estimated channel in time domain. That will bring a very high computing cost especially when the amount of subcarriers is great. Thereby a frequency-domain single tapped equalization method is presented. When the channel is not frequency selective, the performance of this algorithm is better than that of a time-domain equalization. And it has a lower complexity and a lower computing cost. This paper also researches estimation and equalization methods of CP-FMT (Cyclic Prefix FMT). In a frequency selective channel, CP-FMT performs better than FMT.Lastly,this paper researches FMT synchronization technologies. In FMT systems, the high-order prototype filter causes long memory between successive transmitted FMT symbols. Because of this, the period of time-domain training data is long that increases synchronization overheads, and shrinks a range of the captured fractional carrier frequency offset. To settle the above questions, this paper presented a shortened cyclic synchronization algorithm. It inserts nonzero training data in subcarriers with a same interval. Compared with other FMT synchronization algorithms, it decreases the period of time-domain training data and shrinks the synchronization window. Therefore, it extends the range of the captured frequency offset and improves the estimation precision of the time offset. In order to further synchronization performance and reduce synchronization overheads, this paper presented a synchronization algorithm with an adjustable window. It estimates the time offset and the frequency offset based on the correlativity between the received signal and the local training data. It requires only one training symbol and estimates the time offset with a probability as 100%. It also extends the range of the frequency offset to maximum and improves the estimation precision approximating 100%. Carrier frequency offset includes a fractional offset and an integer offset relative to a frequency interval between subcarriers. The above synchronization algorithms only estimate the fractional offset. Therefore, in order to determine the integer carrier frequency offset,this paper presents energy detecting and deciding algorithm,post-equalizer and pre-equalizer PN correlative detecting algorithms. The first one detects an energy distribution of all subcarriers that is based on the energy distribution rule of training symbols used in the shortened cyclic synchronization algorithm. The offset of the energy peak equals to the normalized integer carrier frequency offset with the frequency interval. Its precision is so high that the probability can be larger than 99% even if SNR is 0dB. However, this algorithm only determines the integer carrier frequency offset smaller than the interval between nonzero values in training symbols. The second one searches the position of a PN code in signals from equalizers, which depends on a synchronous code before the PN code. Then the offset is determined by detecting the position of the correlative peak of the found and a local PN code in frequency domain. The default is that the performance of this algorithm depends on FMT equalizers. However, this algorithm shows a high probability of determining the offset correctly that exceeds 94%. The third one inserts two same PN codes into the transmitted FMT symbols in frequency domain. The offset is determined by detecting the correlative peak between the two PN codes and a local PN code. The peak is unique in time domain and frequency domain that results in a probability 100% of capturing the offset correctly.