maki/openairinterface5g
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Files
1661294de53675cc11305a15fbbd5f453750222c
openairinterface5g/openair1/SIMULATION/LTE_PHY/Abstraction/Training/polyfit_beta_training.m
T
Cedric Roux 56c66c09aa - Fix: avoid infinite loop on mulitcast read thread if no data to read 
Pre-ci test passed

git-svn-id: http://svn.eurecom.fr/openair4G/trunk@4000 818b1a75-f10b-46b9-bf7c-635c3b92a50f
2013-06-25 09:05:02 +00:00

348 lines
14 KiB
Matlab
Raw Blame History

%***************************************************************************************
% Calibration of Adjustment Factors for Exponential Effective SINR Mapping Methdology
% For 3GPP Long Term Evolution (LTE)
% Eurecom Institute
% (c) Imran Latif, Communications Mobile, 2012
%***************************************************************************************
%close all
%clear all
%***************************************************************************************
% Some basic initializations
%***************************************************************************************
global SINR_p BLER_meas abs_mode modu Pm tx_mode alpha seta
warning on
set(0, 'DefaultLineMarkerSize', 10);
set(0, 'Defaultaxesfontsize', 10);
set(0, 'DefaultLineLineWidth', 2);
set(0, 'DefaultAxesFontName', 'Helvetica');
set(0, 'DefaultTextFontName', 'Helvetica');
plot_style = {'b*';'r*';'g*';'y*';'k*';...
'bo';'ro';'go';'yo';'ko';...
'bs';'rs';'gs';'ys';'ks';...
'bd';'rd';'gd';'yd';'kd';...
'bx';'rx';'gx';'yx';'kx';...
'b+';'r+';'g+';'y+';'k+'};
echo off;
pathname = '/homes/latif/devel/trunk/openair1/SIMULATION/LTE_PHY/TM5_ideal_CE/MIESM/M1/Polyfit';
%root_path = '/homes/latif/OpenAirPlatform/test_devel/openair1_multi_eNb_UE/SIMULATION/LTE_PHY/BLER_SIMULATIONS/awgn_bler_new/';
%%
%***************************************************************************************
% Input Parameters used for the training of adjustment Factors (betas)
%***************************************************************************************
mcs = [0 2 4 6 7 8 9];
% An array which hold the number of LTE MCS which is to be calibrated for
% abstraction
s2=[];
s3=[];
% String arrays used for printing in plots
tx_mode=5;
%1--> SISO, 2 --> Alamouti, 5 --> MU-MIMO ,6 --> TX Beamforming
% tells which transmission mode is to be abstracted.
abs_mode=2; % 0:MIESM (M2-MIESM), 1:EESM, 2:IA-MIESM (M1-MIESM)
%tells which abstraction methodology is to be used.
Pm=100; %2*50 QPSK
% Used for calculating RBIR
modu=4;
% Modulation used.
p=50;
save_out=1;
if(save_out)
fid = fopen('IA_polyfit_M1_MIESM.tex','w');
fprintf(fid,'\\documentclass[english,12pt,a4paper]{article}\n\n');
fprintf(fid,'\\usepackage{times}\n');
fprintf(fid,'\\usepackage{babel}\n');
fprintf(fid,'\\usepackage{graphicx}\n');
fprintf(fid,'\\usepackage{hyperref}\n\n');
fprintf(fid,'\\begin{document}\n\n');
fprintf(fid,'\\title{%s}\n\n','MU-MIMO with IA: Results of Abstraction using Polynomial fitting for M1-MIESM');
fprintf(fid,'\\maketitle\n\n');
fprintf(fid,'\\newpage\n\n');
fprintf(fid,'\\listoffigures\n\n');
fprintf(fid,'\\listoftables\n\n');
fprintf(fid,'\\newpage\n\n');
end
disp('*******************************************************************\n')
%printf('TX Mode = %d, Abstraction Mode = %d\n',tx_mode,abs_mode);
disp('*******************************************************************\n')
%***************************************************************************************
% Main Execution Loop starts from here:
%***************************************************************************************
for m=1:1:length(mcs)
ss = strcat('LTE TM5 M1 MIESM MCS ', num2str(mcs(m)));
if(mcs(m) >9 && mcs(m) < 17)
modu=16;
Pm=200; %4*50 16QAM
else if(mcs(m) > 16 && mcs(m) < 28)
modu = 64;
Pm=300; %6 *50 64QAM
end
end
%%
% %***************************************************************************************
% %Reference AWGN Curves
% %***************************************************************************************
% data = dlmread(sprintf('/homes/latif/devel/trunk/openair1/SIMULATION/LTE_PHY/awgn_bler_tx1_mcs%d.csv',mcs(m)),';',1,0);
% snr = data(:,1);
% ble = data(:,5)./data(:,6); % round 1
%
% blerr = ble(ble<0.99);
% bler = blerr(blerr>0.001);
%
% snr = snr(ble<0.99);
% snr = snr(blerr>0.001);
%
% % snr_interp = snr(1):0.001:snr(end);
% % bler_interp = interp1(snr,bler,snr_interp,'linear');
% % snr = snr_interp;
% % bler = bler_interp;
%
% %***************************************************************************************
% % Data Extraction on Sub carrier basis data file saved using dlsim and reference
% % SINR Calculation based on Reference AWGN Curve
% %***************************************************************************************
eval(['data_all = data_all' num2str(mcs(m)) ';']);
BLER_meas = data_all(:,end);
%
% SINR_awgn = interp1(bler, snr, BLER_meas,'linear','extrap');
% SINR_awgn = SINR_awgn';
%%
%***************************************************************************************
% Calculation of SINR per subcarrier
%***************************************************************************************
%for SISO
switch(tx_mode)
case 1
ch_tmp = data_all(:,2:51);
abs_channel = (abs(ch_tmp).^2);
input_snr = 10.^(data_all(:,1)/10);
input_snr = repmat(input_snr,1,50);
SINR_p = 10*log10(input_snr.*abs_channel);
case 2
ch1_tmp = data_all(:,2:2:100);
ch2_tmp = data_all(:,3:2:101);
abs_channel1 = (abs(ch1_tmp).^2)/2;
abs_channel2 = (abs(ch2_tmp).^2)/2;
input_snr = 10.^(data_all(:,1)/10);
input_snr = repmat(input_snr,1,50);
SINR_p = 10*log10(input_snr.*abs_channel1 + input_snr.*abs_channel2);%- 10*log10(2);
case 5
prec1_tmp = data_all(:,203:end-1);
prec2_tmp = arrayfun(@opposite_q,prec1_tmp);
q1 = arrayfun(@demap_q,prec1_tmp);
%q2 = q1;
q2 = arrayfun(@demap_q,prec2_tmp);
ch1_tmp = data_all(:,2:2:100);
ch2_tmp = data_all(:,3:2:101);
snr_tmp = 10.^(data_all(:,1)/10);
snr_tmp = repmat(snr_tmp,1,50);
alpha =(1/2)*abs(ch1_tmp + q1.*ch2_tmp);
seta = (1/2)*((abs(ch1_tmp + q2.*ch2_tmp)));
if (abs_mode==2 )
SINR_p = 10*log10(snr_tmp.*(alpha.^2));%-10*log10(2);
else
SINR_p = 10*log10((snr_tmp.*(alpha.^2)) ./ ((snr_tmp.*(seta.^2)) + 1));
end
case 6
prec_tmp = data_all(:,203:end-1);
q = arrayfun(@demap_q,prec_tmp);
ch1_tmp = data_all(:,2:2:100);
ch2_tmp = data_all(:,3:2:101);
snr_tmp = 10.^(data_all(:,1)/10);
snr_tmp = repmat(snr_tmp,1,50);
ch_composite = (abs(ch1_tmp + q.*ch2_tmp))/sqrt(2);
SINR_p = 10*log10((snr_tmp.*((ch_composite).^2)));
end
%***************************************************************************************
% for coarse beta calculation:
%***************************************************************************************
% beta_vec = 1:0.1:30;
% delta = zeros(size(beta_vec));
% for beta=1:length(beta_vec)
% delta(beta)=delta_BLER_1((beta_vec(beta)));
% end
% [val ind] = min(delta);
% opt_beta_new(mcs(m)) = beta_vec(ind);
%***************************************************************************************
% for optmimized betas calculation using fminsearch
%***************************************************************************************
%beta_out=[1.0151 1.0565];
beta_out=zeros(1,2);
%[opt_beta(mcs(m)), MSE] = fminsearch(@delta_BLER_1,opt_beta_new(mcs(m)))
[beta_out, std_deviat] = fminsearch(@polyfit_delta_BLER,[1 1])
%***************************************************************************************
% Calculation of Effective SINR based on optimized beta values
%***************************************************************************************
if(abs_mode==0) %MIESM Mapping
eval(['load ' '/homes/latif/devel/trunk/openair1/SIMULATION/LTE_PHY/Abstraction/siso_MI_abs_' num2str(modu) 'Qam.mat'])
[xize y]= size(SINR_p);
RBIR = [];
for t=1:1:xize
s = SINR_p(t,:);
s(s<-10)=-10;
s(s>49)=49;
eval(['SI_p = interp1(newSNR,newC_siso_' num2str(modu) 'QAM,s, ''linear'' , ''extrap'');']);
RBIR(t) = (sum(SI_p/beta_out(1))/Pm);
%RBIR(t) = (sum(SI_p)/Pm);
end
SINR_eff = interp1(newRBIR, newSNR, RBIR,'linear');
SINR_eff = SINR_eff .* beta_out(2);
elseif(abs_mode==2)
eval(['load ' '/homes/latif/devel/trunk/openair1/SIMULATION/LTE_PHY/Abstraction/siso_MI_abs_' num2str(modu) 'Qam.mat'])
if(tx_mode~=5)
error('IA-MIESM only for LTE TM5!!!')
end
if(modu==4)
eval(['load ' '/homes/latif/devel/trunk/openair1/EMOS/LTE/POST_PROCESSING/Mode_5/mat_Files/data_mode5_4_' num2str(modu) '_morerealizations.mat'])
else
if(modu==16)
eval(['load ' '/homes/latif/devel/trunk/openair1/EMOS/LTE/POST_PROCESSING/Mode_5/mat_Files/data_mode5_16_16.mat'])
end
end
[xize y]= size(SINR_p);
RBIR = [];
for t=1:1:xize
s = SINR_p(t,:);
s(s<-10)=-10;
s(s>49)=49;
a = round(alpha(t,:)*100);
a(a==0) = 1;
a(a>200)=200;
b = round(seta(t,:)*100);
b(b==0) = 1;
b(b>200)=200;
for u=1:length(a)
if(modu==16)
SI_p(u) = data_mode5_16_16(b(u) ,a(u), round(s(u) + 11));
else
SI_p(u) = data_mode5_4_4(b(u) ,a(u), round(s(u) + 11));
end
end
RBIR(t) = (sum(SI_p/beta_out(1))/Pm);
end
SINR_eff = interp1(newRBIR, newSNR, RBIR,'linear');
SINR_eff = SINR_eff .* beta_out(2);
%SINR_eff = 10*log10((10.^(SINR_eff/10)) .* beta_out(2));
else %% EESM Mapping
SINR_eff= 10*log10(-1*beta_out(2)*log((1/p)*sum(exp((-10.^(SINR_p'./10))./beta_out(1)))));
end
%***************************************************************************************
% PolyNomial Calculation
%***************************************************************************************
[po s mu] = polyfit(SINR_eff,((BLER_meas))',3);
[f] = polyval(po,SINR_eff,s,mu);
%***************************************************************************************
%MSE Calculation
%***************************************************************************************
% MSE_final(mcs(m)+1) = mean((SINR_awgn - SINR_eff).^2);
Optimum_beta1(mcs(m)+1) = beta_out(1);
Optimum_beta2(mcs(m)+1) = beta_out(2);
%***************************************************************************************
% BLER Plot w.r.t Mean Effective SINR
%***************************************************************************************
% figure;
% avg_SINR_p = mean(SINR_p');
% hist(avg_SINR_p)
% s = strcat('Average_SINR_p, MCS = ',num2str(mcs(m)));
% title(2)
% figure;
% grid on
% semilogy(avg_SINR_p,BLER_meas,plot_style{m+1})
% xlim([-8 15])
% ylim([1e-3 1])
% s = strcat('LTE TM1 Avergae Abstraction MCS ', num2str(mcs(m)));
% title(s);
% ylabel 'BLER'
% xlabel 'SINR_{average}'
% s2 = strcat('BLER_{meas} MCS ', num2str(mcs(m)));
% legend(s2, 'Location', 'Best');
%***************************************************************************************
% BLER Plot w.r.t Effective SINR
%***************************************************************************************
% figure;
% hist(SINR_eff)
% s = strcat('SINR_eff, MCS = ',num2str(mcs(m)));
% title(s)
h_fig = figure;
semilogy(SINR_eff,BLER_meas,plot_style{m+1})
hold on
grid on
semilogy(SINR_eff,f,'mx');
xlim([-20 20])
ylim([1e-3 1])
s = strcat(ss, ',beta1= ', num2str(Optimum_beta1(mcs(m)+1)),', beta2= ',num2str(Optimum_beta2(mcs(m)+1)));
title(s);
ylabel 'BLER'
xlabel 'SINR_{effective}'
s2 = strcat('BLER_{meas} MCS ', num2str(mcs(m)));
s3 = strcat('BLER_{AWGN} MCS ', num2str(mcs(m)));
legend(s2,s3, 'Location', 'Best');
if(save_out==1)
saveas(h_fig,fullfile(pathname,strcat('IA_RX_M1-MIESM_TX5_mcs',num2str(mcs(m)) ,'.eps')),'epsc2');
saveas(h_fig,fullfile(pathname,strcat('IA_RX_M1-MIESM_TX5_mcs',num2str(mcs(m)) ,'.fig')),'fig');
saveas(h_fig,fullfile(pathname,strcat('IA_RX_M1-MIESM_TX5_mcs',num2str(mcs(m)) ,'.jpg')),'jpg');
fprintf(fid,'\\begin{figure}[!h]\n');
fprintf(fid,'\\centering\n');
fprintf(fid,'\\includegraphics[scale=0.5]{%s}\n',fullfile(pathname,strcat('IA_RX_M1_MIESM_TX5_mcs', num2str(mcs(m)),'.eps')));
fprintf(fid,'\\caption{%s}\n',s);
fprintf(fid,'\\label{fig:mcs%d}\n',mcs(m));
fprintf(fid,'\\end{figure}\n\n');
end
end
if(save_out)
fprintf(fid,'\\end{document}\n');
fclose(fid);
end
%***************************************************************************************
%For saving the files
%***************************************************************************************
% if(save_out)
% save(strcat('opt_beta_mcs',num2str(mcs(m)),'.dat'),'Optimum_beta*')
% save(strcat('MSE_mcs',num2str(mcs(m)),'.dat'),'MSE_final')
% end
Reference in New Issue View Git Blame Copy Permalink