csdr/grc_tests/bpsk31_tedvar.m
2017-04-30 13:51:12 +02:00

109 lines
3.4 KiB
Matlab
Executable file

#!/usr/bin/octave
%you need to first install the parallel and struct packages:
%pkg install -forge struct
%pkg install -forge parallel
pkg load parallel
function y=inarg(x)
for i=1:length(argv())
if strcmp(argv(){i},x)
y=1;
return
end
end
y=0;
end
if !inarg('--nogen')
fwrite(stdout, "===========================================\nGenerating baseband signal from random data\n===========================================\n");
system('cat /dev/urandom | csdr pack_bits_8to1_u8_u8 | csdr psk_modulator_u8_c 2 | csdr gain_ff 0.25 | csdr psk31_interpolate_sine_cc 256 | csdr add_n_zero_samples_at_beginning_f 170 | pv -ps 2g | dd iflag=fullblock bs=128M count=16 of=/tmp/psk31-raw-data');
fwrite(stdout, "===========================================\nGenerating Gaussian white noise for agwn_cc\n===========================================\n");
system('csdr gaussian_noise_c | pv -ps 256m | dd of=/tmp/psk31-gaussian-noise iflag=fullblock bs=256M count=1');
end
if inarg('--onlygen')
exit(0)
end
fwrite(stdout, "===========================================\nCalculating variance graph data \n===========================================\n");
function output=shrun(cmd, type, minsize)
SIGTERM=15;
output=[];
cmd
[pin, pout, pid]=popen2('bash',{'-c', cmd});
%fclose(pin);
do
sleep(0.3)
fwrite(stdout,'.');
%size(output)
%output
current_output=fread(pout, Inf, type);
frewind(pout);
output=[output; current_output];
until(size(output)(1)>=minsize)
waitpid(pid);
kill(pid, SIGTERM);
fclose(pin);
fclose(pout);
end
function variance=run_var(snr, which_ted)
disp('ran a command')
out_vect=shrun(sprintf('cat /tmp/psk31-raw-data | csdr awgn_cc %d --awgnfile /tmp/psk31-gaussian-noise | csdr timing_recovery_cc %s 256 --add_q --output_indexes | CSDR_FIXED_BUFSIZE=1048576 csdr normalized_timing_variance_u32_f 256 85', snr, which_ted), 'float32', 1);
disp('run_var output:');
out_vect'
variance=out_vect(1);
end
function variances=mkvarplot(which_ted, snrs)
fun = @(x) run_var(x, which_ted);
variances=pararrayfun(nproc, fun, snrs);
%{
variances=[]
for snr=snrs
snr
variances=[variances run_var(snr, which_ted)];
end
%}
end
function fmtplot(h)
FN = findall(h,'-property','FontName');
set(FN,'FontName','/usr/share/fonts/truetype/ttf-dejavu/DejaVuSerifCondensed.ttf');
set(FN,'FontName','times');
FS = findall(h,'-property','FontSize');
set(FS,'FontSize',18);
xlabel('E_b/N_0 [dB]');
ylabel('Phase error variance [rad^2]');
end
snrs=-5:5:30
%snrs=[10]
error_values_gardner=mkvarplot('GARDNER',snrs);
%{
snrs_earlylate=0:256
error_values_earlylate=mkvarplot('EARLYLATE',snrs_earlylate);
%}
%graphics_toolkit("gnuplot")
h=figure(1);
ebn0=snrs-13.26-10*log10(1/256.)
%13.56 dB is the difference between the real (measured) SNR and the number input to awgn_cc.
%This is because agwn_cc assumes a signal with 0dB power at te input, while our BPSK31 baseband signal is of -13.26 dB.
semilogy(ebn0, error_values_gardner, 'linewidth', 2);
title('Estimation variance');
fmtplot(h)
pause
%{
semilogy(snrs_earlylate, error_values_earlylate, 'linewidth', 2);
title('S-curve for early-late TED');
fmtplot(h)
pause
%}
if !inarg('--nogen')
system('rm /tmp/psk31-raw-data /tmp/psk31-gaussian-noise');
end