【含仿真录像】大规模MIMO系统下量化ML信道估计matlab仿真

clc; clear; close all; warning off; addpath(genpath(pwd)); Nr = 64; %% receive antenna number Nt = 32; %% transmit antenna number K = 64; %% pilot length SNR_dB = -10:5:15; %% SNR SNR = 10.^(SNR_dB / 10); Pnt = 1 ./ SNR; %% noise power (the transmit power is normalized as 1) MonteC = 10; %% Monte Carlo trail number %% Rayleigh Fading Channel Model H = randn(Nr,Nt) + 1i* randn(Nr,Nt); H = H/sqrt(2); h_true = H(:); h_true = [real(h_true); imag(h_true)]; H_fro = norm(H,'fro')^2; %% MSE vector for different quantizaton bit number mse_1bit_mc = zeros(length(SNR), MonteC); mse_2bit_mc = zeros(length(SNR), MonteC); mse_3bit_mc = zeros(length(SNR), MonteC); mse_4bit_mc = zeros(length(SNR), MonteC); mse_unqt_mc = zeros(length(SNR), MonteC); %% Zadoff-Zhu sequence / Orthogonal Pilots x = zadoff_chu(K); S0 = zeros(K,K); ind = 1:K; for jj = 1:K ind = mod(ind,K) + 1; S0(:,jj) = x(ind); end for i = 1 : length(SNR) Pn = Pnt(i); %% Thresholds used for 1-bit quantization h_max = sqrt(1 + Pn)/sqrt(2); ht = -h_max : 2*h_max/7 : h_max; ht = ht.'; for j = 1: MonteC % [i, j] %% Thresholds used for 1-bit quantization t_bar = zeros(2*K*Nr,1); kro = ones(K,1); tem = ht(randi(8, Nr, 1)); t_bar(1:K*Nr) = kron(kro, tem); tem = ht(randi(8,Nr,1)); t_bar(K*Nr+1:2*K*Nr) = kron(kro, tem); %% Received Signal Model % qpsk_signal = randi(4,Nt,K); % X = pskmod(qpsk_signal - 1, 4, pi/4); %% Random QPSK pilots X = S0(randperm(Nt),:); %% Random Orthogonal Pilots X = X / sqrt(Nt); %% normalize the transmit power to 1 N = random('norm',0,sqrt(Pn/2),Nr,K) + 1i * random('norm',0,sqrt(Pn/2),Nr,K); %% Noise matrix Y_unqt = H * X + N; y_unqt = Y_unqt(:); y_bar = [real(y_unqt); imag(y_unqt)]; %% 1-bit z_bar = sign(y_bar - t_bar); y_lowb = z_bar(1 : K*Nr) + 1i * z_bar(K*Nr+1 : 2*K*Nr); Y_lowb = reshape(y_lowb, Nr, K); h_1bit = func_1bMM_ML(z_bar, X, Nr, t_bar); mse_1bit_mc(i, j) = sum((h_1bit - h_true).^2); %% 2-bit [y_lowb, thres_r, thres_l] = func_lowbSamp(y_bar, 2); h_2bit = func_CQML(thres_l, thres_r, y_lowb, X, Nr); mse_2bit_mc(i, j) = sum((h_2bit - h_true).^2); %% 3-bit [y_lowb, thres_r, thres_l] = func_lowbSamp(y_bar, 3); h_3bit = func_CQML(thres_l, thres_r, y_lowb, X, Nr); mse_3bit_mc(i, j) = sum((h_3bit - h_true).^2); %% 4-bit [y_lowb, thres_r, thres_l] = func_lowbSamp(y_bar, 4); h_4bit = func_CQML(thres_l, thres_r, y_lowb, X, Nr); mse_4bit_mc(i, j) = sum((h_4bit - h_true).^2); %% Unquantized case h_unqt = func_unqtML(Y_unqt, X); mse_unqt_mc(i, j) = sum((h_unqt - h_true).^2); end end %% Compute the NMSE mse_1bit = mean(mse_1bit_mc, 2)/H_fro; mse_2bit = mean(mse_2bit_mc, 2)/H_fro; mse_3bit = mean(mse_3bit_mc, 2)/H_fro; mse_4bit = mean(mse_4bit_mc, 2)/H_fro; mse_unqt = mean(mse_unqt_mc, 2)/H_fro; %% Plot figure fx=@(x) 10*log10(x); A=[0.13 0.55 0.13;0.60 0.20 0.98]; figure(1); hold on; plot(SNR_dB,fx(mse_1bit),'-*','COLOR',A(1,:),'LineWidth',1.5,'MarkerSize',14); plot(SNR_dB,fx(mse_2bit),'-ks','LineWidth',1.5,'MarkerSize',14); plot(SNR_dB,fx(mse_3bit),'-kd','LineWidth',1.5,'MarkerSize',14); plot(SNR_dB,fx(mse_4bit),'-ko','LineWidth',1.5,'MarkerSize',14); plot(SNR_dB,fx(mse_unqt),'-r>','LineWidth',1.5,'MarkerSize',14); grid on; set(gca, 'FontSize', 14); lg = legend('1bMM-ML','CQML:2-bit','CQML:3-bit','CQML:4-bit','UnqtML'); lg.FontSize = 14; xlabel('SNR (dB)','FontSize',14); ylabel('MSE (dB)','FontSize',14);