mirror
/
OpenRTX
kopia lustrzana https://github.com/OpenRTX/OpenRTX
1
0
Forkuj 0
Kod Zgłoszenia Packages Projekty Wydania Wiki Aktywność

Rewrite quantization algorithm 

Quantization algorithm is now based on separate rolling averages for
positive and negative samples.
Merged csv outputs from syncword search and quantization on Linux.

TG-81
pull/68/head
Niccolò Izzo 2022-04-14 10:51:27 +02:00 zatwierdzone przez Silvano Seva
rodzic aa9fcc26b7
commit 7d1b89550e
5 zmienionych plików z 55 dodań i 66 usunięć
Pokaż statystyki Ściągnij plik aktualizacji Ściągnij plik porównania Expand all files Collapse all files

8
openrtx/include/protocols/M17/M17Demodulator.h
Wyświetl plik

@ -124,8 +124,8 @@ private:
static constexpr size_t M17_BRIDGE_SIZE = M17_SAMPLES_PER_SYMBOL * M17_SYNCWORD_SYMBOLS;
static constexpr float CONV_STATS_ALPHA = 0.001f;
static constexpr float QNT_STATS_ALPHA = 0.999f;
static constexpr float CONV_THRESHOLD_FACTOR = 3.40;
static constexpr int16_t QNT_SMA_WINDOW = 40;
/**
* M17 syncwords;
@ -164,8 +164,10 @@ private:
/*
* Quantization statistics computation
*/
float qnt_max = 0.0f; ///< Max hold of the sliced samples
float qnt_min = 0.0f; ///< Min hold of the sliced samples.
std::deque<int16_t> qnt_pos_fifo;
std::deque<int16_t> qnt_neg_fifo;
float qnt_pos_avg = 0.0f; ///< Rolling average of positive samples
float qnt_neg_avg = 0.0f; ///< Rolling average of negative samples
/*
* DSP filter state

67
openrtx/src/protocols/M17/M17Demodulator.cpp
Wyświetl plik

@ -66,11 +66,8 @@ void M17Demodulator::init()
newFrame = false;
#ifdef PLATFORM_LINUX
FILE *csv_log = fopen("demod_log_1.csv", "w");
fprintf(csv_log, "Signal,Convolution,Threshold,Offset\n");
fclose(csv_log);
csv_log = fopen("demod_log_2.csv", "w");
fprintf(csv_log, "Sample,Max,Min,Symbol,I\n");
FILE *csv_log = fopen("demod_log.csv", "w");
fprintf(csv_log, "Signal,Convolution,Threshold,Offset,Sample,Max,Min,Symbol,I\n");
fclose(csv_log);
#endif // PLATFORM_MOD17
}
@ -124,8 +121,8 @@ float M17Demodulator::getCorrelationStddev()
void M17Demodulator::resetQuantizationStats()
{
qnt_max = 0.0f;
qnt_min = 0.0f;
qnt_pos_avg = 0.0f;
qnt_neg_avg = 0.0f;
}
void M17Demodulator::updateQuantizationStats(int32_t offset)
@ -136,14 +133,28 @@ void M17Demodulator::updateQuantizationStats(int32_t offset)
sample = basebandBridge[M17_BRIDGE_SIZE + offset];
else // Otherwise use regular data buffer
sample = baseband.data[offset];
if (sample > qnt_max)
qnt_max = sample;
if (sample > 0)
{
// If the FIFO is not full just push a new sample
if (qnt_pos_fifo.size() >= QNT_SMA_WINDOW)
{
qnt_pos_avg += 1 / static_cast<float>(QNT_SMA_WINDOW) *
(sample - qnt_pos_fifo.back());
qnt_pos_fifo.pop_back();
}
qnt_pos_fifo.push_front(sample);
}
else
qnt_max *= QNT_STATS_ALPHA;
if (sample < qnt_min)
qnt_min = sample;
else
qnt_min *= QNT_STATS_ALPHA;
{
// If the FIFO is not full just push a new sample
if (qnt_neg_fifo.size() >= QNT_SMA_WINDOW)
{
qnt_neg_avg += 1 / static_cast<float>(QNT_SMA_WINDOW) *
(sample - qnt_neg_fifo.back());
qnt_neg_fifo.pop_back();
}
qnt_neg_fifo.push_front(sample);
}
}
int32_t M17Demodulator::convolution(int32_t offset,
@ -169,7 +180,7 @@ int32_t M17Demodulator::convolution(int32_t offset,
sync_t M17Demodulator::nextFrameSync(int32_t offset)
{
#ifdef PLATFORM_LINUX
FILE *csv_log = fopen("demod_log_1.csv", "a");
FILE *csv_log = fopen("demod_log.csv", "a");
#endif
sync_t syncword = { -1, false };
@ -192,11 +203,12 @@ sync_t M17Demodulator::nextFrameSync(int32_t offset)
else // Otherwise use regular data buffer
sample = baseband.data[i];
fprintf(csv_log, "%" PRId16 ",%d,%f,%d\n",
fprintf(csv_log, "%" PRId16 ",%d,%f,%d,%" PRId16 ",%f,%f,%d,%d\n",
sample,
conv,
CONV_THRESHOLD_FACTOR * getCorrelationStddev(),
i);
conv / 10,
CONV_THRESHOLD_FACTOR * getCorrelationStddev() / 10,
i,
0,0.0,0.0,0,0);
#endif
// Positive correlation peak -> frame syncword
@ -226,9 +238,9 @@ int8_t M17Demodulator::quantize(int32_t offset)
sample = basebandBridge[M17_BRIDGE_SIZE + offset];
else // Otherwise use regular data buffer
sample = baseband.data[offset];
if (sample > static_cast< int16_t >(qnt_max) / 2)
if (sample > static_cast< int16_t >(qnt_pos_avg))
return +3;
else if (sample < static_cast< int16_t >(qnt_min) / 2)
else if (sample < static_cast< int16_t >(qnt_neg_avg))
return -3;
else if (sample > 0)
return +1;
@ -271,7 +283,7 @@ bool M17Demodulator::update()
dsp_dcRemoval(&dsp_state, baseband.data, baseband.len);
#ifdef PLATFORM_LINUX
FILE *csv_log = fopen("demod_log_2.csv", "a");
FILE *csv_log = fopen("demod_log.csv", "a");
#endif
if(baseband.data != NULL)
@ -314,12 +326,13 @@ bool M17Demodulator::update()
int8_t symbol = quantize(symbol_index);
#ifdef PLATFORM_LINUX
fprintf(csv_log, "%" PRId16 ",%f,%f,%d,%d\n",
baseband.data[symbol_index] - (int16_t) qnt_ema,
qnt_max / 2,
qnt_min / 2,
fprintf(csv_log, "%" PRId16 ",%d,%f,%d,%" PRId16 ",%f,%f,%d,%d\n",
0,0,0.0,0,
baseband.data[symbol_index],
qnt_pos_avg,
qnt_neg_avg,
symbol * 666,
frameIndex == 0 ? 2300 : symbol_index);
frameIndex);
#endif
setSymbol<M17_FRAME_BYTES>(*activeFrame, frameIndex, symbol);

16
scripts/plot_m17_demod_csv.py
Wyświetl plik

@ -6,11 +6,15 @@ from sys import argv
plt.rcParams["figure.autolayout"] = True
df = pd.read_csv(argv[1])
#df = df.head(n=10000)
print("Contents in csv file:\n", df)
#plt.plot(df.index, df.Input)
#plt.plot(df.index, df.RRCSignal)
plt.plot(df.index, df.LSFConvolution)
plt.plot(df.index, df.FrameConvolution)
plt.plot(df.index, df.Stddev)
plt.plot(df.index, df.Signal)
plt.plot(df.index, df.Convolution)
plt.plot(df.index, df.Threshold)
plt.plot(df.index, df.Threshold * -1)
plt.plot(df.index, df.Offset)
plt.plot(df.index, df.Sample)
plt.plot(df.index, df.Max)
plt.plot(df.index, df.Min)
#plt.plot(df.index, df.Symbol)
plt.plot(df.index, df.I)
plt.show()

15
scripts/plot_m17_demod_csv3.py
Wyświetl plik

@ -1,15 +0,0 @@
#! /usr/bin/env python3
import pandas as pd
from matplotlib import pyplot as plt
from sys import argv
plt.rcParams["figure.autolayout"] = True
df = pd.read_csv(argv[1])
print("Contents in csv file:\n", df)
plt.plot(df.index, df.Signal)
plt.plot(df.index, df.Convolution)
plt.plot(df.index, df.Threshold)
plt.plot(df.index, df.Threshold * -1)
plt.plot(df.index, df.Offset)
plt.show()

15
scripts/plot_m17_demod_csv4.py
Wyświetl plik

@ -1,15 +0,0 @@
#! /usr/bin/env python3
import pandas as pd
from matplotlib import pyplot as plt
from sys import argv
plt.rcParams["figure.autolayout"] = True
df = pd.read_csv(argv[1])
print("Contents in csv file:\n", df)
plt.plot(df.index, df.Sample)
plt.plot(df.index, df.Max)
plt.plot(df.index, df.Min)
plt.plot(df.index, df.Symbol)
plt.plot(df.index, df.I)
plt.show()