Mirror/aicodix___modem
1
0
Fork 0
mirror of https://github.com/aicodix/modem.git synced 2026-04-27 14:30:34 +00:00
Code Issues Projects Releases Packages Wiki Activity

estimate SFO and CFO via simple linear regression

Browse source 
and compensate by correcting constellation points
This commit is contained in:
Ahmet Inan 2021-09-02 22:23:27 +02:00
commit 19535a7344
1 changed files with 30 additions and 6 deletions
Download patch file Download diff file Expand all files Collapse all files

38
decode.cc
View file

@ -9,6 +9,7 @@ Copyright 2021 Ahmet Inan <inan@aicodix.de>
#include <cassert>
#include <cmath>
namespace DSP { using std::abs; using std::min; using std::cos; using std::sin; }
#include "regression.hh"
#include "bip_buffer.hh"
#include "xorshift.hh"
#include "trigger.hh"
@ -201,6 +202,7 @@ struct Decoder
code_type code[65536];
cmplx head[symbol_len], tail[symbol_len], cons[cons_max];
cmplx fdom[symbol_len], tdom[buffer_len];
value index[cons_max], phase[cons_max];
value cfo_rad, sfo_rad;
const uint32_t *frozen_bits;
int code_order;
@ -465,18 +467,40 @@ struct Decoder
cons[cons_cols*j+i] = demod_or_erase(fdom[bin(i+code_off)], head[bin(i+code_off)]);
}
if (1) {
value sum = 0;
for (int i = 0; i < cons_cnt; ++i) {
code_type tmp[mod_max];
mod_hard(tmp, cons[i]);
sum += arg(cons[i] * conj(mod_map(tmp)));
index[i] = (i % cons_cols) + code_off;
phase[i] = arg(cons[i] * conj(mod_map(tmp)));
}
value avg = sum / cons_cnt;
cfo_rad += avg / (symbol_len+guard_len);
std::cerr << "finer cfo: " << cfo_rad * (rate / Const::TwoPi()) << " Hz " << std::endl;
cmplx comp = DSP::polar<value>(1, -avg);
DSP::SimpleLinearRegression<value> dirty(index, phase, cons_cnt);
//std::cerr << "dirty slope = " << dirty.slope() << std::endl;
//std::cerr << "dirty yint = " << dirty.yint() << std::endl;
value avg_diff = 0;
for (int i = 0; i < cons_cnt; ++i)
cons[i] *= comp;
avg_diff += dirty(index[i]) - phase[i];
avg_diff /= value(cons_cnt);
value var_diff = 0;
for (int i = 0; i < cons_cnt; ++i)
var_diff += (dirty(index[i]) - phase[i] - avg_diff) * (dirty(index[i]) - phase[i] - avg_diff);
value std_dev = std::sqrt(var_diff/(cons_cnt-1));
int count = 0;
for (int i = 0; i < cons_cnt; ++i) {
if (std::abs(dirty(index[i])-phase[i]) <= std_dev) {
index[count] = index[i];
phase[count] = phase[i];
++count;
}
}
DSP::SimpleLinearRegression<value> pruned(index, phase, count);
//std::cerr << "pruned slope = " << pruned.slope() << std::endl;
//std::cerr << "pruned yint = " << pruned.yint() << std::endl;
sfo_rad -= pruned.slope() * symbol_len / value(symbol_len+guard_len);
cfo_rad += pruned.yint() / (symbol_len+guard_len);
std::cerr << "coarse sfo: " << 1000000 * sfo_rad / Const::TwoPi() << " ppm" << std::endl;
std::cerr << "finer cfo: " << cfo_rad * (rate / Const::TwoPi()) << " Hz " << std::endl;
for (int i = 0; i < cons_cnt; ++i)
cons[i] *= DSP::polar<value>(1, -pruned((i % cons_cols) + code_off));
}
value precision = 16;
if (1) {