Sd.rar_sorry

clc;clear; %% Parameters M = 128; % N = 20; % G = 64; M1 = 16; M2 = 8; K = 4; P = 4; b = 2; B0 = 8; loop_num = 100; loop_num1 = loop_num; SNR_table = 0:3:12; B_table = zeros(size(SNR_table)); d = 0.5; A = zeros(P,M,K); A_hat = zeros(P,M,K); Rs0=zeros(loop_num,length(SNR_table)); Rs1=zeros(loop_num,length(SNR_table)); Rs2=zeros(loop_num,length(SNR_table)); Rs3=zeros(loop_num,length(SNR_table)); % Rs4=zeros(loop_num,length(SNR_table)); R_t_sum= zeros(M,M); U = zeros(M,M,K); R_sqt =zeros(M,M,K); sin_theta_table = zeros(P,K); cos_theta_table = zeros(P,K); sin_psi_table = zeros(P,K); %% Channel generation for i_K=1:1:K theta = random('unif',-1/2,1/2,P,1)*pi; psi = random('unif',-1/2,1/2,P,1)*pi; sin_theta_table(:,i_K)=sin(theta); sin_psi_table(:,i_K)=sin(psi); cos_theta_table(:,i_K)=cos(theta); end for i_loop=1:loop_num1 H=zeros(K,M); for i_K=1:1:K for i_P=1:1:P a1 = exp(cos_theta_table(i_P,i_K)*sin_psi_table(i_P,i_K)*(-1j*2*pi*d*[0:1:M1-1])); a2 = exp(sin_theta_table(i_P,i_K)*(-1j*2*pi*d*[0:1:M2-1])); A(i_P,:,i_K) = kron(a1,a2); end g = randn(1,P) + 1j*randn(1,P); h = g*A(:,:,i_K); H(i_K,:) = h; end R_t_sum(:,:) = R_t_sum(:,:) + H' * H; end R_t = R_t_sum/loop_num1; for i_K=1:1:K [U(:,:,i_K),S,V] = svd(R_t(:,:)); R_sqt(:,:,i_K) = U(:,:,i_K)*sqrt(S)*U(:,:,i_K)'; end for i_loop=1:loop_num %% Channel generation H=zeros(K,M); for i_K=1:1:K sin_theta = sin_theta_table(:,i_K); sin_theta_quant = UQ(sin_theta,B0,-1,1); sin_psi = sin_psi_table(:,i_K); sin_psi_quant = UQ(sin_psi,B0,-1,1); cos_theta = cos_theta_table(:,i_K); cos_theta_quant = UQ(cos_theta,B0,-1,1); for i_P=1:1:P a1 = exp(cos_theta_table(i_P,i_K)*sin_psi_table(i_P,i_K)*(-1j*2*pi*d*[0:1:M1-1])); a2 = exp(sin_theta_table(i_P,i_K)*(-1j*2*pi*d*[0:1:M2-1])); A(i_P,:,i_K) = kron(a1,a2); end for i_P=1:1:P a1_hat = exp(cos_theta_quant(i_P)*sin_psi_quant(i_P)*(-1j*2*pi*d*[0:1:M1-1])); a2_hat = exp(sin_theta_quant(i_P)*(-1j*2*pi*d*[0:1:M2-1])); A_hat(i_P,:,i_K) = kron(a1_hat,a2_hat); end g = randn(1,P) + 1j*randn(1,P); h= g*A(:,:,i_K); H(i_K,:) = h; end for i_SNR=1:1:length(SNR_table) SNR = SNR_table(i_SNR); snr=10^(SNR/10); norm_sq_sum=0; for i_K=1:1:K norm_sq_sum = norm_sq_sum + norm(H(i_K,:))^2; end norm_h_sq=norm_sq_sum/K; % norm_h_sq=M; Gamma = snr*K/norm_h_sq; % B = round((P-1)/3*SNR+(P-1)*log2((K-1)/(b-1))) - B_delta; B = 8; B_table(i_SNR) = B; %% RVQ [H_RVQ,err_RVQ]= quantiz_RVQ_real(H, B); %% SC [H_AC,err_AC] = quantiz_SC_real(H,A_hat,B-3); %% CRVQ [H_CRVQ,err_CRVQ1] = quantiz_CRVQ_real(H,R_sqt,B); % %% CS % Phi = randn(M,N); % H_til = H * Phi; % [H_til_CS,err_CS] = quantiz_CS_real(H_til,R_sqt,Phi,B); % H_CS = zeros(K,M); % for i_K=1:1:K % h_til_CS = H_til_CS(i_K,:); % s_ori = H(i_K,:) * U(:,:,i_K); % s = OMP(h_til_CS',(U(:,:,i_K)'*Phi)',P); % H_CS(i_K,:) = s' * U(:,:,i_K)'; % end % %% AG % GP = zeros(M,G); % for i=1:1:G % GP(2*i-1,i)=1; % GP(2*i,i)=1; % end % EP=GP'/2; % Hr = H*GP; % for i=1:1:K % AGP(:,:,i) = A(:,:,i)*GP; % end % Hr_AG = quantiz_AC_real(Hr,AGP,B); % % Hr_AG = quantiz_RVQ_real(Hr,B); % H_CS = Hr_AG * EP; %% RS HK=H; HK1= H_RVQ; HK2 = H_AC; HK3 = H_CRVQ; % HK4 = H_CS; % HK5(i_K,:) = H_squant5; % precode % W0=randn(M,K) + j*randn(M,K); W0 = HK'/(HK*HK'); % W0=HK'; W0 = W0./repmat(sqrt(sum(abs(W0).^2,1)),M,1); W1 = HK1'/(HK1*HK1'); % W1=HK1'; W1 = W1./repmat(sqrt(sum(abs(W1).^2,1)),M,1); W2 = HK2'/(HK2*HK2'); % W2=HK2'; W2 = W2./repmat(sqrt(sum(abs(W2).^2,1)),M,1); W3 = HK3'/(HK3*HK3'); % W3=HK3'; W3 = W3./repmat(sqrt(sum(abs(W3).^2,1)),M,1); % W4 = HK4'/(HK4*HK4'); % % W4=HK4'; % W4 = W4./repmat(sqrt(sum(abs(W4).^2,1)),M,1); % W5 = HK5'/(HK5*HK5'); % W5 = W5./repmat(sqrt(sum(abs(W5).^2,1)),M,1); p=Gamma/K; for i_K = 1:K SINR0 = p*norm(HK(i_K,:)*W0(:,i_K))^2/(1 + p*HK(i_K,:)*W0*(HK(i_K,:)*W0)'-p*norm(HK(i_K,:)*W0(:,i_K))^2); SINR1 = p*norm(HK(i_K,:)*W1(:,i_K))^2/(1 + p*HK(i_K,:)*W1*(HK(i_K,:)*W1)'-p*norm(HK(i_K,:)*W1(:,i_K))^2); SINR2 = p*norm(HK(i_K,:)*W2(:,i_K))^2/(1 + p*HK(i_K,:)*W2*(HK(i_K,:)*W2)'-p*norm(HK(i_K,:)*W2(:,i_K))^2); SINR3 = p*norm(HK(i_K,:)*W3(:,i_K))^2/(1 + p*HK(i_K,:)*W3*(HK(i_K,:)*W3)'-p*norm(HK(i_K,:)*W3(:,i_K))^2); % SINR4 = p*norm(HK(i_K,:)*W4(:,i_K))^2/(1 + p*HK(i_K,:)*W4*(HK(i_K,:)*W4)'-p*norm(HK(i_K,:)*W4(:,i_K))^2); % SINR5 = p*norm(HK(i_K,:)*W5(:,i_K))^2/(1 + p*HK(i_K,:)*W5*(HK(i_K,:)*W5)'-p*norm(HK(i_K,:)*W5(:,i_K))^2); Rs0(i_loop,i_SNR) = Rs0(i_loop,i_SNR)+log2(1+SINR0); Rs1(i_loop,i_SNR) = Rs1(i_loop,i_SNR)+log2(1+SINR1); Rs2(i_loop,i_SNR) = Rs2(i_loop,i_SNR)+log2(1+SINR2); Rs3(i_loop,i_SNR) = Rs3(i_loop,i_SNR)+log2(1+SINR3); % Rs4(i_loop,i_SNR) = Rs4(i_loop,i_SNR)+log2(1+SINR4); end fprintf('i_loop=%d,i_SNR=%d\n',i_loop,i_SNR); end end mRs0=mean(Rs0,1)/K; mRs1=mean(Rs1,1)/K; mRs2=mean(Rs2,1)/K; mRs3=mean(Rs3,1)/K; % mRs4=mean(Rs4,1)/K; % mRs5=mean(Rs5,1)/K; %% plot figure; plot(SNR_table,mRs0,'k--','LineWidth',1.5); hold on; plot(SNR_table,mRs2,'rd-','LineWidth',1.5); hold on; plot(SNR_table,mRs3,'bs-','LineWidth',1.5); % hold on; % plot(SNR_table,mRs4,'g^-','LineWidth',1.5); % hold on; % plot(SNR_table,mRs1,'mo-','LineWidth',1.5); % hold on; % plot(B_table,mRs5,'ms-','LineWidth',1.5); % had=legend('$Perfect$','$RVQ$','$DC(\mathbf{A}=\hat{\mathbf{A}})$','$DC(\mathbf{A}\neq\hat{\mathbf{A}})$'); % set(had,'Interpreter','latex') legend('Perfect CSIT','Proposed AoD-adaptive subspace codebook','Conventional channel statistics-based codebook [12]'); % axis([3 15 0 4]); grid on; xlabel('SNR (dB)'); ylabel('Per-user rate (bps/Hz)');