DPD预失真算法的matlab仿真实现

clc; close all; %% 信号产生 simout=load('E:\预失真\仿真\zjp_pre-distoration simulation\dpdsimulation\qpsk_8000.mat'); simout=simout.simout; fs=2*10^8;%采样率200Mhz st=0:length(simout)-1; s_qpsk=(simout.').*exp(j*2*pi*20000000*st/fs);%取实部为QPSK调制 %% 滤波器系数设置 N=50;%滤波器阶数 Wn1=[0.75,0.85];%1为fs的一半 Wn2=[0.15,0.25]; A=fir1(N,Wn1,'bandpass'); B=fir1(N,Wn2,'bandpass'); %% 功率放大 fc=6*10^7;%载波60MHZ t=1:length(s_qpsk); s_carri=s_qpsk.*exp(j*2*pi*fc*(t-1)/fs);%上变频 s_carri_b=filter(A,1,s_carri);%带通滤波 h = spectrum.welch; hpsd_carri_b=psd(h,s_carri_b,'fs',fs); figure(1); plot(hpsd_carri_b);%功率放大前的功率谱密度 a=[1.0513+0.0904j,-0.068-0.0023j,0.0289+0.0054j,0.0542-0.29j,0.2234+0.2317j,-0.0621-0.0932j,-0.9657-0.7028j,-0.2451-0.3735j,0.1229+0.1508j]; %a=[2.3,4.2,1.3,-1.2,-3.2,9.1,0.5,2.67,1.7]; HPA_s=volterra(a,s_carri_b); % h=spectrum.welch; hpsd=psd(h,HPA_s,'fs',fs); figure(2); plot(hpsd); %% 预失真+功放-----多项式法 b=[1.0513+0.0904j,-0.068-0.0023j,0.0289+0.0054j,0.0542-0.29j,0.2234+0.2317j,-0.0621-0.0932j,-0.9657-0.7028j,-0.2451-0.3735j,0.1229+0.1508j]; %b=[2.3,4.2,1.3,-1.2,-3.2,9.1,0.5,2.67,1.7]; w=zeros(1,length(b)); w=[0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j,0.01+0.01j]; %w=[0.00679765478699548-0.000259498756269653i,0.00837971394159005-0.000111941227697152i,0.0114287412467534+0.000135048972435035i,0.00999572083062263-3.11814652641192e-07i,0.00999748855999973-1.58555811669834e-07i,0.0100016305410513+1.47719343296075e-07i,0.00999993931875462-4.06509747496053e-09i,0.00999995731939103-2.60788444989852e-09i,0.0100000155833265+1.43812621935027e-09i]; s_qpsk=[0,0,s_qpsk]; u=0.05;%LMS算法的参数 DPD_s0=zeros(1,length(HPA_s)+2); DPD_s1=zeros(1,length(HPA_s)+2); DPD_s=zeros(1,length(HPA_s)+2); LVB1=zeros(1,N+1); LVB2=zeros(1,N+1); HPA_s_p=zeros(1,length(HPA_s)+2); e=zeros(1,length(HPA_s)); y=zeros(1,length(HPA_s)+2); lamda=0.99;%QRD-RLS算法的参数 y_q=zeros(1,length(HPA_s)+2); Y_q=ones(9,9); X_q=zeros(1,9); for n=3:length(HPA_s)+2, %预失真 S_qpsk=[s_qpsk(n),s_qpsk(n-1),s_qpsk(n-2),abs(s_qpsk(n))^2*s_qpsk(n),abs(s_qpsk(n-1))^2*s_qpsk(n-1),abs(s_qpsk(n-2))^2*s_qpsk(n-2),abs(s_qpsk(n))^4*s_qpsk(n),abs(s_qpsk(n-1))^4*s_qpsk(n-1),abs(s_qpsk(n-2))^4*s_qpsk(n-2)]; DPD_s0(n)=w*S_qpsk.'; %上变频 DPD_s1(n)=DPD_s0(n)*exp(j*2*pi*fc*(n-3)/fs); %滤波 LVB1=[DPD_s1(n),LVB1(1:N)]; DPD_s(n)=A*LVB1.'; %功放 DPD_S=[DPD_s(n),DPD_s(n-1),DPD_s(n-2),abs(DPD_s(n))^2*DPD_s(n),abs(DPD_s(n-1))^2*DPD_s(n-1),abs(DPD_s(n-2))^2*DPD_s(n-2),abs(DPD_s(n))^4*DPD_s(n),abs(DPD_s(n-1))^4*DPD_s(n-1),abs(DPD_s(n-2))^4*DPD_s(n-2)]; HPA_s_p(n)=b*DPD_S.'; %下变频 DPD_s2(n)=HPA_s_p(n)*exp(j*2*pi*fc*(n-3)/fs); %滤波 LVB2=[DPD_s2(n),LVB2(1:N)]; DPD_s3(n)= B*LVB2.'; %LMS自适应算法 y(n)=DPD_s3(n)/2;%假设功率放大4倍 Y=[y(n),y(n-1),y(n-2),abs(y(n))^2*y(n),abs(y(n-1))^2*y(n-1),abs(y(n-2))^2*y(n-2),abs(y(n))^4*y(n),abs(y(n-1))^4*y(n-1),abs(y(n-2))^4*y(n-2)]; e(n-2)=DPD_s0(n)-w*Y.'; w=w-u*Y*e(n-2); %自适应算法---QRD-RLS % X_q=[DPD_s0(n),X_q(1),X_q(2),X_q(3),X_q(4),X_q(5),X_q(6),X_q(7),X_q(8)];%期望信号 % y_q(n)=DPD_s3(n)/2;%假设功放放大4倍 % Y_q_t=[y_q(n),y_q(n-1),y_q(n-2),abs(y_q(n))^2*y_q(n),abs(y_q(n-1))^2*y_q(n-1),abs(y_q(n-2))^2*y_q(n-2),abs(y_q(n))^4*y_q(n),abs(y_q(n-1))^4*y_q(n-1),abs(y_q(n-2))^4*y_q(n-2)]; % Y_q=[Y_q_t;Y_q(1,:);Y_q(2,:);Y_q(3,:);Y_q(4,:);Y_q(5,:);Y_q(6,:);Y_q(7,:);Y_q(8,:)];%预失真训练器的输入信号 % Am=[1,0,0,0,0,0,0,0,0;0,lamda,0,0,0,0,0,0,0;0,0,lamda^2,0,0,0,0,0,0;0,0,0,lamda^3,0,0,0,0,0;0,0,0,0,lamda^4,0,0,0,0;0,0,0,0,0,lamda^5,0,0,0;0,0,0,0,0,0,lamda^6,0,0;0,0,0,0,0,0,0,lamda^7,0;0,0,0,0,0,0,0,0,lamda^8];%权重矩阵 % [Q,R]=qr(Am*Y_q); % U=Q'*(Am*X_q.'); % w=(U\R);%R是上三角矩阵，逐个回代即可解w end % h=spectrum.welch; hpsd_HPA_s_p=psd(h,HPA_s_p(2000:end),'fs',fs); figure(3); plot(hpsd_HPA_s_p); %% 预失真+功放-----查找表法 ram1=ones(1,80000); ram2=zeros(1,80000); ram3=zeros(1,10000);%考虑记忆效应的作用时要初始化这个表 %b_c=[2.3,4.2,1.3,-1.2,-3.2,9.1,0.5,2.67,1.7]; b_c=[1.0513+0.0904j,-0.068-0.0023j,0.0289+0.0054j,0.0542-0.29j,0.2234+0.2317j,-0.0621-0.0932j,-0.9657-0.7028j,-0.2451-0.3735j,0.1229+0.1508j]; % b_c=[1.0513+0.0904j,0,0,0.0542-0.29j,0,0,-0.9657-0.7028j,0,0]; DPD_c_s0=zeros(1,length(HPA_s)+2); DPD_c_s1=zeros(1,length(HPA_s)+2); s_qpsk_c=zeros(1,length(HPA_s)+2); LVB_c1=zeros(1,N+1); LVB_c2=zeros(1,N+1); HPA_s_c=zeros(1,length(HPA_s)+2); s_abs_square=ones(1,length(HPA_s)+2); u_p=-0.001; u_s=0.98; v_max_square=0.6; R_max=10000; %预失真 for i=3:length(HPA_s)+2, s_abs(i)=abs(s_qpsk(i)); s_abs_square(i)=s_abs(i)^2; if(s_abs_square(i)<=v_max_square), n_s=floor(s_abs_square(i)/v_max_square*length(ram1)); else n_s=length(ram1)-1; end; b_s_a=ram1(n_s+1); b_s_f=ram2(n_s+1); R_y=s_abs_square(i-1)/s_abs_square(i); if(R_y<=R_max), n_y=floor(R_y/R_max*length(ram3)); else n_y=length(ram3)-1; end; DPD_c_s0(i)=(s_qpsk(i)*b_s_a+abs(ram3(n_y+1)))*exp(j*(b_s_f+angle(ram3(n_y+1)))); % DPD_c_s0(i)=(s_qpsk(i)*b_s_a)*exp(j*(b_s_f)); %上变频 DPD_c_s1(i)=DPD_c_s0(i)*exp(j*2*pi*fc*(i-3)/fs); %滤波 LVB_c1=[DPD_c_s1(i),LVB_c1(1:N)]; s_qpsk_c(i)=A*LVB_c1.'; %功率放大 DPD_S_C=[s_qpsk_c(i),s_qpsk_c(i-1),s_qpsk_c(i-2),abs(s_qpsk_c(i))^2*s_qpsk_c(i),abs(s_qpsk_c(i-1))^2*s_qpsk_c(i-1),abs(s_qpsk_c(i-2))^2*s_qpsk_c(i-2),abs(s_qpsk_c(i))^4*s_qpsk_c(i),abs(s_qpsk_c(i-1))^4*s_qpsk_c(i-1),abs(s_qpsk_c(i-2))^4*s_qpsk_c(i-2)]; % DPD_S_C=[s_qpsk_c(i),abs(s_qpsk_c(i))^2*s_qpsk_c(i),abs(s_qpsk_c(i))^4*s_qpsk_c(i)]; HPA_s_c(i)=b_c*DPD_S_C.'; %下变频 DPD_c_s2(i)=HPA_s_c(i)*exp(j*2*pi*fc*(i-3)/fs); %滤波 LVB_c2=[DPD_c_s2(i),LVB_c2(1:N)]; DPD_c_s3(i)= B*LVB_c2.'; %自适应算法---LMS e_p=2*s_abs(i)-abs(DPD_c_s3(i));%功率放大4倍 e_s=angle(s_qpsk(i))-angle(DPD_c_s3(i)); ram1(n_s+1)=ram1(n_s+1)+e_p*u_p; ram2(n_s+1)=ram2(n_s+1)+e_s*u_s; end % h = spectrum.welch; hpsd_HPA_s_c=psd(h,HPA_s_c(2000:end),'fs',fs); figure(4); plot(hpsd_HPA_s_c);