/*
* aptdec - A lightweight FOSS (NOAA) APT decoder
* Copyright (C) 2004-2009 Thierry Leconte (F4DWV) 2019-2022 Xerbo (xerbo@protonmail.com)
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see .
*/
#include
#include
#include
#include
#include "apt.h"
#include "filter.h"
#include "taps.h"
#include "util.h"
// Block sizes
#define BLKAMP 32768
#define BLKIN 32768
#define CARRIER_FREQ 2400.0
#define MAX_CARRIER_OFFSET 20.0
#define RSMULT 15
#define Fi (APT_IMG_WIDTH*2 * RSMULT)
static float _sample_rate;
static float offset = 0.0;
static float FreqLine = 1.0;
static float oscillator_freq;
static float pll_alpha;
static float pll_beta;
// Initalise and configure PLL
int apt_init(double sample_rate) {
if(sample_rate > Fi) return 1;
if(sample_rate < APT_IMG_WIDTH*2) return -1;
_sample_rate = sample_rate;
// Pll configuration
pll_alpha = 50 / _sample_rate;
pll_beta = pll_alpha * pll_alpha / 2.0;
oscillator_freq = CARRIER_FREQ/sample_rate;
return 0;
}
static float pll(complexf_t in) {
static float oscillator_phase = 0.0;
// Internal oscillator
#ifdef _MSC_VER
complexf_t osc = _FCbuild(cos(oscillator_phase), -sin(oscillator_phase));
in = _FCmulcc(in, osc);
#else
complexf_t osc = cos(oscillator_phase) + -sin(oscillator_phase)*I;
in *= osc;
#endif
// Error detector
float error = cargf(in);
// Adjust frequency and phase
oscillator_freq += pll_beta*error;
oscillator_freq = clamp_half(oscillator_freq, (CARRIER_FREQ + MAX_CARRIER_OFFSET) / _sample_rate);
oscillator_phase += M_TAUf * (pll_alpha*error + oscillator_freq);
oscillator_phase = remainderf(oscillator_phase, M_TAUf);
return crealf(in);
}
// Convert samples into pixels
static int getamp(float *ampbuff, int count, apt_getsamples_t getsamples, void *context) {
static float inbuff[BLKIN];
static int idxin = 0;
static int nin = 0;
for (int n = 0; n < count; n++) {
float I2, Q;
// Get some more samples when needed
if (nin < HILBERT_FILTER_SIZE * 2 + 2) {
// Number of samples read
int res;
memmove(inbuff, &(inbuff[idxin]), nin * sizeof(float));
idxin = 0;
// Read some samples
res = getsamples(context, &(inbuff[nin]), BLKIN - nin);
nin += res;
// Make sure there is enough samples to continue
if (nin < HILBERT_FILTER_SIZE * 2 + 2)
return n;
}
// Process read samples into a brightness value
complexf_t sample = hilbert_transform(&inbuff[idxin], hilbert_filter, HILBERT_FILTER_SIZE);
ampbuff[n] = pll(sample);
// Increment current sample
idxin++;
nin--;
}
return count;
}
// Sub-pixel offsetting
int getpixelv(float *pvbuff, int count, apt_getsamples_t getsamples, void *context) {
// Amplitude buffer
static float ampbuff[BLKAMP];
static int nam = 0;
static int idxam = 0;
float mult;
// Gaussian resampling factor
mult = (float) Fi / _sample_rate * FreqLine;
int m = (int)(LOW_PASS_SIZE / mult + 1);
for (int n = 0; n < count; n++) {
int shift;
if (nam < m) {
int res;
memmove(ampbuff, &(ampbuff[idxam]), nam * sizeof(float));
idxam = 0;
res = getamp(&(ampbuff[nam]), BLKAMP - nam, getsamples, context);
nam += res;
if (nam < m)
return n;
}
pvbuff[n] = interpolating_convolve(&(ampbuff[idxam]), low_pass, LOW_PASS_SIZE, offset, mult) * mult * 256.0;
shift = ((int) floor((RSMULT - offset) / mult)) + 1;
offset = shift * mult + offset - RSMULT;
idxam += shift;
nam -= shift;
}
return count;
}
// Get an entire row of pixels, aligned with sync markers
int apt_getpixelrow(float *pixelv, int nrow, int *zenith, int reset, apt_getsamples_t getsamples, void *context) {
static float pixels[APT_IMG_WIDTH + SYNC_PATTERN_SIZE];
static size_t npv;
static int synced = 0;
static float max = 0.0;
static float minDoppler = 1000000000, previous = 0;
if(reset) synced = 0;
float corr, ecorr, lcorr;
int res;
// Move the row buffer into the the image buffer
if (npv > 0)
memmove(pixelv, pixels, npv * sizeof(float));
// Get the sync line
if (npv < SYNC_PATTERN_SIZE + 2) {
res = getpixelv(&(pixelv[npv]), SYNC_PATTERN_SIZE + 2 - npv, getsamples, context);
npv += res;
if (npv < SYNC_PATTERN_SIZE + 2)
return 0;
}
// Calculate the frequency offset
ecorr = convolve(pixelv, sync_pattern, SYNC_PATTERN_SIZE);
corr = convolve(&pixelv[1], sync_pattern, SYNC_PATTERN_SIZE - 1);
lcorr = convolve(&pixelv[2], sync_pattern, SYNC_PATTERN_SIZE - 2);
FreqLine = 1.0+((ecorr-lcorr) / corr / APT_IMG_WIDTH / 4.0);
float val = fabs(lcorr - ecorr)*0.25 + previous*0.75;
if(val < minDoppler && nrow > 10){
minDoppler = val;
*zenith = nrow;
}
previous = fabs(lcorr - ecorr);
// The point in which the pixel offset is recalculated
if (corr < 0.75 * max) {
synced = 0;
FreqLine = 1.0;
}
max = corr;
if (synced < 8) {
int mshift;
static int lastmshift;
if (npv < APT_IMG_WIDTH + SYNC_PATTERN_SIZE) {
res = getpixelv(&(pixelv[npv]), APT_IMG_WIDTH + SYNC_PATTERN_SIZE - npv, getsamples, context);
npv += res;
if (npv < APT_IMG_WIDTH + SYNC_PATTERN_SIZE)
return 0;
}
// Test every possible position until we get the best result
mshift = 0;
for (int shift = 0; shift < APT_IMG_WIDTH; shift++) {
float corr;
corr = convolve(&(pixelv[shift + 1]), sync_pattern, SYNC_PATTERN_SIZE);
if (corr > max) {
mshift = shift;
max = corr;
}
}
// Stop rows dissapearing into the void
int mshiftOrig = mshift;
if(abs(lastmshift-mshift) > 3 && nrow != 0){
mshift = 0;
}
lastmshift = mshiftOrig;
// If we are already as aligned as we can get, just continue
if (mshift == 0) {
synced++;
} else {
memmove(pixelv, &(pixelv[mshift]), (npv - mshift) * sizeof(float));
npv -= mshift;
synced = 0;
FreqLine = 1.0;
}
}
// Get the rest of this row
if (npv < APT_IMG_WIDTH) {
res = getpixelv(&(pixelv[npv]), APT_IMG_WIDTH - npv, getsamples, context);
npv += res;
if (npv < APT_IMG_WIDTH)
return 0;
}
// Move the sync lines into the output buffer with the calculated offset
if (npv == APT_IMG_WIDTH) {
npv = 0;
} else {
memmove(pixels, &(pixelv[APT_IMG_WIDTH]), (npv - APT_IMG_WIDTH) * sizeof(float));
npv -= APT_IMG_WIDTH;
}
return 1;
}