The Demultiplexing And DSP Realization Of Frequency Division Multiplexing Signal

Massive Machine Type Communication(mMTC)is one the most important applications of 5G.In the usage scenarios of mMTC,massive frequency channels and narrowband signals are included,which requires channelization algorithm to be done at very low cost.This article is aimed at finding some channelization with low complexity,meaning minimize the DSP resources and time occupied under the premise of ensuring performance.Large numbers of frequency channels and high order filters are regarded to be the cause of high complexity of channelization algorithm.For performance reasons,traditional channelization scheme choses FPGA,which costs more and is difficult to exploit.At present,with the development of DSP Technology,DSP chips' s performance improve significantly,DSP has become a better option for mMTC application which costs less and is easier to exploit.Due to the transform of hardware platform,the algorithm benchmark need to be reconsidered.Two algorithms are evaluated from a new angle.The first algorithm is called fast convolution algorithm.We focus on the improved overlap saving(IOS)method,which is one of fast convolution algorithm.This method comes from OFDM study and can not be used directly in channelization,so the paper discussed what details need to be changed when it is applied in channelization.The IOS method has following characteristics: higher computational complexity compared with traditional channelization method,higher data transferring efficiency which makes it suitable to the DSP application,each channel can be filtered separated and has no correlation with each other which makes it more flexible.The second algorithm is called polyphase decomposition(PD)method,which is a traditional channelization algorithm.The article separately deduced the computational complexity in uniformly distributed channel and ununiformly distributed channel,discussed the relation between deviation and filter.The PD method has following characteristics: in uniformly distributed channel,lower computational complexity compared with IOS method,lower data transferring efficiency which makes it not suitable to the DSP application,in ununiformly distributed channel,the computational complexity increased significantly,and the filters of different channels need to meet certain conditions,which makes it not so flexible.This article also provides the MATLAB simulation to compare the SNR of IOS and PD,both of the two methods are realized on DSP.Through the compare of the actual instruction cycle numbers of these two methods on the DSP,we demonstrates that IOS is a better channelization method on DSP,which makes it more suitable to mMTC application.