请问各位大大...有没有人有模拟过下面这篇paper... J.J. van de Beek, M. Sandell and P. O. Borjesson, “ML estimation of time and frequency offset in OFDM systems,” IEEE Trans. Signal Processing, vol. 45, no. 7, pp. 1800-1805, July 1997. 我在写程式时把杂讯和频率偏移都设为0...可是估测出来还是有错误... 可是找了半天也找不出错误...可不可以请有写过的大大帮我找一下错误... 下面是我的程式码... clear; clc; format long g N=256; % The size of an OFDM symbol CP=15; % Cyclic prefix interval timer=100;%10^6; % Per SNR runs times packetnum=3; snr_db=0:10; % SNR range MSE_ML=zeros(1,length(snr_db)); % Mean-square error in ML algorithm MSE_NA=zeros(1,length(snr_db)); % Mean-square error in new algorithm freq_offset=0;%0.25; % frequency offset range=(N/2)+CP+1:(N)+(N/2)+(2*CP); % Observation interval Pos=(N/2); % The actual symbol timing in observation interval for snr=1:length(snr_db) for ttt=1:timer %%% TX %%% clear data_tx; clear out_ifft; clear ps_out; data_tx=sign(unidrnd(2,packetnum,N)-1.5); % BPSK signal generator out_ifft(1,:)=sqrt(N)*ifft(data_tx(1,:)); out_ifft(2,:)=sqrt(N)*ifft(data_tx(2,:)); out_ifft(3,:)=sqrt(N)*ifft(data_tx(3,:)); ps_out=[out_ifft(1,N-CP+1:N),out_ifft(1,:),out_ifft(2,N-CP+1:N), out_ifft(2,:),out_ifft(3,N-CP+1:N),out_ifft(3,:)]; % Transmitted signal %%% Channel %%% clear tx_power; clear noise_std; clear rx1; clear noise; tx_power=sum(ps_out.*conj(ps_out))/length(ps_out); % transmitted power noise_std=sqrt(tx_power/(2*10^(snr_db(snr)/10))); % noise standard deviation %%% Frequency offset %%% rx1=ps_out; noise=noise_std*(randn(1,length(rx1))+(j*randn(1,length(rx1)))); clear freq_time; clear rx; freq_time=1:length(rx1); rx=rx1.*exp(j*((2*pi*freq_offset*mod(freq_time,N+CP))/N));%+noise; % Received signal LO=snr/(snr+1); %%% RX %%% clear symboltiming; clear NL; clear Distance; [NL]=NewAlgorithm(N,CP,rx,snr); [rrr,symboltiming]=max(NL(range)); a(snr,ttt)=symboltiming; Distance=abs(Pos-symboltiming); MSE_NA(snr)=MSE_NA(snr)+Distance^2; clear SymbolTiming; clear RL; clear Dis; [RL]=ML(N,CP,rx,snr); [ggg,SymbolTiming]=max(RL(range)); b(snr,ttt)=SymbolTiming; Dis=abs(Pos-SymbolTiming); MSE_ML(snr)=MSE_ML(snr)+Dis^2; end tim=round(clock); snr_db(snr); MSE_NA(snr)=MSE_NA(snr)/timer MSE_ML(snr)=MSE_ML(snr)/timer end tim=round(clock) figure semilogy(snr_db,MSE_NA,'-g*',snr_db,MSE_ML,'-ro') legend('NA(CP=15)','ML(CP=15)') xlabel('SNR(dB)') ylabel('Mean-squared error') grid ====================================================== 下面是NewAlgorithm 和 ML两个function 的程式 function [refer_line]=NewAlgorithm(N,CP,rx,snr) LO=snr/(snr+1); for n=1:length(rx)-N-CP % time offset r(n)=(1/(1+LO))*sum(conj(rx(n+(2*CP)+2:n+(N/2)+CP)).* conj((rx(n+N:(-1):n+(N/2)+CP+2))))+ (1/(1+(2*LO)))*(sum(rx(n+1:n+CP).*conj(rx(n+N+1:n+N+CP)))+ sum(conj(rx(n+CP+2:n+(2*CP)+1)).*conj(rx(n+N+CP:(-1):n+N+1)))); IO1(n)=sum(abs(rx(n+(2*CP)+2:n+(N/2)+CP)).^2)+ sum(abs(rx(n+(N/2)+CP+2:n+N)).^2); r1(n)=real(1/(1+(2*LO))*sum(conj(rx(n+1:n+CP)).* conj(rx(n+(2*CP)+1:(-1):n+CP+2)))); IO2(n)=sum(abs(rx(n+1:n+CP)).^2)+sum(abs(rx(n+CP+2:n+(2*CP)+1)).^2)+ sum(abs(rx(n+N+1:n+N+CP)).^2); refer_line=abs(r(n))+r1(n)-((LO/(2*(1+LO)))*IO1(n))- ((LO/(1+(2*LO)))*IO2(n)); end function [refer_line]=ML(N,CP,rx,snr) LO=snr/(snr+1); for n=1:(length(rx)-N-CP) R(n)=sum(rx(n+1:n+CP).*conj(rx(n+N+1:n+N+CP))); IO(n)=(1/2)*(sum(abs(rx(n+1:n+CP)).^2)+sum(abs(rx(n+N+1:n+N+CP)).^2)); refer_line=abs(R(n))-(LO*IO(n)); end --

※ 发信站: 批踢踢实业坊(ptt.cc) ◆ From: 163.22.16.42