504 lines
17 KiB
C
504 lines
17 KiB
C
/* FSK decoder of carrier FSK signals received by simple FM receiver
|
|
*
|
|
* (C) 2016 by Andreas Eversberg <jolly@eversberg.eu>
|
|
* All Rights Reserved
|
|
*
|
|
* 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 3 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 <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
/* How does it work:
|
|
* -----------------
|
|
*
|
|
* C-Netz modulates the carrier frequency. If it is 2.4 kHz above, it is high
|
|
* level, if it is 2.4 kHz below, it is low level. Look at FTZ 171 TR 60
|
|
* Chapter 5 (data exchange) for closer information.
|
|
*
|
|
* Detect level change:
|
|
*
|
|
* We don't just look for high/low level, because we don't know what the actual
|
|
* 0-level of the phone's transmitter is. (level of carrier frequency) Also we
|
|
* use receiver and sound card that cause any level to return to 0 after some
|
|
* time, even if the transmitter still transmits a level above or below the
|
|
* carrier frequnecy. Insted we look at the change of the received signal. An
|
|
* upward change indicates 1. An downward change indicates 0. (This may also be
|
|
* reversed, it we find out, that we received a sync sequence in received
|
|
* polarity.) If there is no significant change in level, we keep the value of
|
|
* last change, regardless of what level we actually receive.
|
|
*
|
|
* To determine a change from noise, we use a theshold. This is set to half of
|
|
* the level of last received change. This means that the next change may be
|
|
* down to a half lower. There is a special case during distributed signaling.
|
|
* The first level change of each data chunk raises or falls from 0-level
|
|
* (unmodulated carrier), so the threshold for this bit is only a quarter of the
|
|
* last received change.
|
|
*
|
|
* While searching for a sync sequence, the threshold for the next change is set
|
|
* after each change. After synchronization, the the threshold is locked to half
|
|
* of the average change level of the sync sequence.
|
|
*
|
|
* Search window
|
|
*
|
|
* We use a window of one bit length (9 samples at 48 kHz sample rate) and look
|
|
* for a change that is higher than the threshold and has its highest slope in
|
|
* the middle of the window. To determine the level, the min and max value
|
|
* inside the window is searched. The differece is the change level. To
|
|
* determine the highest slope, the highest difference between subsequent
|
|
* samples is used. For every sample we move the window one bit to the right
|
|
* (next sample), check if change level matches the threshold and highest slope
|
|
* is in the middle and so forth. Only if the highes slope is exactly in the
|
|
* middle, we declare a change. This means that we detect a slope about half of
|
|
* a bit duration later.
|
|
*
|
|
* When we are not synced:
|
|
*
|
|
* For every change we record a bit. A positive change is 1 and a negative 0. If
|
|
* it turns out that the receiver or sound card is reversed, we reverse bits.
|
|
* After every change we wait up to 1.5 bit duration for next change. If there
|
|
* is a change, we record our next bit. If there is no change, we record the
|
|
* state of the last bit. After we had no change, we wait 1 bit duration, since
|
|
* we already 0.5 behind the start of the recently recorded bit.
|
|
*
|
|
* When we are synced:
|
|
*
|
|
* After we recorded the time of all level changes during the sync sequence, we
|
|
* calulate an average and use it as a time base for sampling the subsequent 150
|
|
* bit of a message. From now on, a bit change does not cause any resync. We
|
|
* just remember what change we received. Later we use it for sampling the 150
|
|
* bits.
|
|
*
|
|
* We wait a duration of 1.5 bits after the sync sequence and the start of the
|
|
* bit that follows the sync sequence. We record what we received as last
|
|
* change. For all following 149 bits we wait 1 bit duration and record what we
|
|
* received as last change.
|
|
*
|
|
* Sync clock
|
|
*
|
|
* Because we transmit and receive chunks of sample from buffers of different
|
|
* drivers, we cannot determine the exact latency between received and
|
|
* transmitted samples. Also some sound cards may have different RX and TX
|
|
* speed. One (pure software) solution is to sync ourself to the mobile phone,
|
|
* since the mobile phone is perfectly synced to use.
|
|
*
|
|
* After receiving and decording of a frame, we use the time of received sync
|
|
* sequence to synchronize the reciever to the mobile phone. If we receive a
|
|
* message on the OgK (control channel), we know that this is a response to a
|
|
* message of a specific time slot we recently sent. Then we can fully sync the
|
|
* receiver's clock. For any other frame, we cannot determine the absolute
|
|
* clock. We just correct the receiver's clock, as the clock differs only
|
|
* slightly from the time the message was received.
|
|
*
|
|
*/
|
|
|
|
#include <stdio.h>
|
|
#include <stdint.h>
|
|
#include <string.h>
|
|
#include <math.h>
|
|
#include "../common/timer.h"
|
|
#include "../common/debug.h"
|
|
#include "cnetz.h"
|
|
#include "dsp.h"
|
|
#include "telegramm.h"
|
|
|
|
/* use to debug decoder */
|
|
//#define DEBUG_DECODER if (1)
|
|
//#define DEBUG_DECODER if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V)
|
|
//#define DEBUG_DECODER if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V && sync)
|
|
|
|
static int len, half;
|
|
static int16_t *spl;
|
|
static int pos;
|
|
static double bits_per_sample, next_bit;
|
|
static int level_threshold;
|
|
static double bit_time, bit_time_uncorrected;
|
|
static enum fsk_sync sync;
|
|
static int last_change_positive;
|
|
static double sync_level;
|
|
static double sync_time;
|
|
static double sync_jitter;
|
|
static int bit_count;
|
|
static int16_t *speech_buffer;
|
|
static int speech_size, speech_count;
|
|
|
|
int fsk_fm_init(fsk_fm_demod_t *fsk, cnetz_t *cnetz, int samplerate, double bitrate)
|
|
{
|
|
memset(fsk, 0, sizeof(*fsk));
|
|
if (samplerate < 48000) {
|
|
PDEBUG(DDSP, DEBUG_ERROR, "Sample rate must be at least 48000 Hz!\n");
|
|
return -1;
|
|
}
|
|
|
|
fsk->cnetz = cnetz;
|
|
|
|
len = (int)((double)samplerate / bitrate + 0.5);
|
|
half = (int)((double)samplerate / bitrate / 2.0 + 0.5);
|
|
if (len > sizeof(fsk->bit_buffer_spl) / sizeof(fsk->bit_buffer_spl[0])) {
|
|
PDEBUG(DDSP, DEBUG_ERROR, "Sample rate too high for buffer, please use lower rate, like 192000 Hz!\n");
|
|
return -1;
|
|
}
|
|
|
|
fsk->bit_buffer_len = len;
|
|
fsk->bit_buffer_half = half;
|
|
fsk->bits_per_sample = bitrate / (double)samplerate;
|
|
|
|
fsk->speech_size = sizeof(fsk->speech_buffer) / sizeof(fsk->speech_buffer[0]);
|
|
|
|
fsk->level_threshold = 655;
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* get levels, sync time and jitter from sync sequence or frame data */
|
|
static inline void get_levels(fsk_fm_demod_t *fsk, int *_min, int *_max, int *_avg, int *_probes, int num, double *_time, double *_jitter)
|
|
{
|
|
int min = 32767, max = -32768, avg = 0, count = 0, level;
|
|
double time = 0, t, sync_average, sync_time, jitter = 0;
|
|
int bit_offset;
|
|
int i;
|
|
|
|
/* get levels an the average receive time */
|
|
for (i = 0; i < num; i++) {
|
|
level = fsk->change_levels[(fsk->change_pos - 1 - i) & 0xff];
|
|
if (level <= 0)
|
|
continue;
|
|
|
|
/* in spk mode, we skip the voice part (62 bits) */
|
|
if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V)
|
|
bit_offset = i + ((i + 2) >> 2) * 62;
|
|
else
|
|
bit_offset = i;
|
|
t = fmod(fsk->change_when[(fsk->change_pos - 1 - i) & 0xff] - bit_time + (double)bit_offset + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
if (t > BITS_PER_SUPERFRAME / 2)
|
|
t -= BITS_PER_SUPERFRAME;
|
|
//if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V)
|
|
// printf("%d: level=%d%% @%.2f difference=%.2f\n", bit_offset, level * 100 / 65536, fsk->change_when[(fsk->change_pos - 1 - i) & 0xff], t);
|
|
time += t;
|
|
|
|
if (level < min)
|
|
min = level;
|
|
if (level > max)
|
|
max = level;
|
|
avg += level;
|
|
count++;
|
|
}
|
|
|
|
if (!count) {
|
|
*_min = *_max = *_avg = 0;
|
|
return;
|
|
}
|
|
|
|
/* when did we received the sync?
|
|
* sync_average is the average about how early (negative) or
|
|
* late (positive) we received the sync relative to current bit_time.
|
|
* sync_time is the absolute time within the super frame.
|
|
*/
|
|
sync_average = time / (double)count;
|
|
sync_time = fmod(sync_average + bit_time + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
|
|
*_probes = count;
|
|
*_min = min;
|
|
*_max = max;
|
|
*_avg = avg / count;
|
|
|
|
if (_time) {
|
|
// if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V)
|
|
// printf("sync at distributed mode\n");
|
|
// printf("sync at bit_time=%.2f (sync_average = %.2f)\n", sync_time, sync_average);
|
|
/* if our average sync is later (greater) than the current
|
|
* bit_time, we must wait longer (next_bit above 1.5)
|
|
* for the time to sample the bit.
|
|
* if sync is earlier, bit_time is already too late, so
|
|
* we must wait less than 1.5 bits */
|
|
next_bit = 1.5 + sync_average;
|
|
*_time = sync_time;
|
|
}
|
|
if (_jitter) {
|
|
/* get jitter of received changes */
|
|
for (i = 0; i < num; i++) {
|
|
level = fsk->change_levels[(fsk->change_pos - 1 - i) & 0xff];
|
|
if (level <= 0)
|
|
continue;
|
|
|
|
/* in spk mode, we skip the voice part (62 bits) */
|
|
if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V)
|
|
bit_offset = i + ((i + 2) >> 2) * 62;
|
|
else
|
|
bit_offset = i;
|
|
t = fmod(fsk->change_when[(fsk->change_pos - 1 - i) & 0xff] - sync_time + (double)bit_offset + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
if (t > BITS_PER_SUPERFRAME / 2)
|
|
t = BITS_PER_SUPERFRAME - t; /* turn negative into positive */
|
|
jitter += t;
|
|
}
|
|
*_jitter = jitter / (double)count;
|
|
}
|
|
}
|
|
|
|
static inline void got_bit(fsk_fm_demod_t *fsk, int bit, int change_level)
|
|
{
|
|
int min, max, avg, probes;
|
|
|
|
/* count bits, but do not exceed 4 bits per SPK block */
|
|
if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V) {
|
|
/* for first bit, we have only half of the modulation deviation, so we multiply level by two */
|
|
if (bit_count == 0)
|
|
change_level *= 2;
|
|
if (bit_count == 4)
|
|
return;
|
|
}
|
|
bit_count++;
|
|
|
|
//printf("bit %d\n", bit);
|
|
fsk->change_levels[fsk->change_pos] = change_level;
|
|
fsk->change_when[fsk->change_pos++] = bit_time;
|
|
|
|
|
|
switch (sync) {
|
|
case FSK_SYNC_NONE:
|
|
fsk->rx_sync = (fsk->rx_sync << 1) | bit;
|
|
/* use half level of last change for threshold change detection.
|
|
* if there is no change detected for 5 bits, set theshold to
|
|
* 1 percent, so the 7 pause bits before a frame will make sure
|
|
* that the change is below noise level, so the first sync
|
|
* bit is detected. then the change is set and adjusted
|
|
* for all other bits in the sync sequence.
|
|
* after sync, the theshold is set to half of the average of
|
|
* all changes in the sync sequence */
|
|
if (change_level) {
|
|
level_threshold = (double)change_level / 2.0;
|
|
} else if ((fsk->rx_sync & 0x1f) == 0x00 || (fsk->rx_sync & 0x1f) == 0x1f) {
|
|
if (fsk->cnetz->dsp_mode != DSP_MODE_SPK_V)
|
|
level_threshold = 655;
|
|
}
|
|
if (detect_sync(fsk->rx_sync)) {
|
|
sync = FSK_SYNC_POSITIVE;
|
|
got_sync:
|
|
get_levels(fsk, &min, &max, &avg, &probes, 30, &sync_time, &sync_jitter);
|
|
sync_level = (double)avg / 65535.0;
|
|
if (sync == FSK_SYNC_NEGATIVE)
|
|
sync_level = -sync_level;
|
|
// printf("sync (change min=%d%% max=%d%% avg=%d%% sync_time=%.2f jitter=%.2f probes=%d)\n", min * 100 / 65535, max * 100 / 65535, avg * 100 / 65535, sync_time, sync_jitter, probes);
|
|
level_threshold = (double)avg / 2.0;
|
|
fsk->rx_sync = 0;
|
|
fsk->rx_buffer_count = 0;
|
|
break;
|
|
}
|
|
if (detect_sync(fsk->rx_sync ^ 0xfffffffff)) {
|
|
sync = FSK_SYNC_NEGATIVE;
|
|
goto got_sync;
|
|
}
|
|
break;
|
|
case FSK_SYNC_NEGATIVE:
|
|
bit = 1 - bit;
|
|
/* fall through */
|
|
case FSK_SYNC_POSITIVE:
|
|
fsk->rx_buffer[fsk->rx_buffer_count] = bit + '0';
|
|
if (++fsk->rx_buffer_count == 150) {
|
|
sync = FSK_SYNC_NONE;
|
|
if (fsk->cnetz->dsp_mode != DSP_MODE_SPK_V) {
|
|
/* received 40 bits after start of block */
|
|
sync_time = fmod(sync_time - (7+33) + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
} else {
|
|
/* received 662 bits after start of block (10 SPK blocks + 1 bit (== 2 level changes)) */
|
|
sync_time = fmod(sync_time - (66*10+2) + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
}
|
|
cnetz_decode_telegramm(fsk->cnetz, fsk->rx_buffer, sync_level, sync_time, sync_jitter);
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* DOC TBD: find change for bit change */
|
|
static inline void find_change(fsk_fm_demod_t *fsk)
|
|
{
|
|
int32_t level_min, level_max, change_max;
|
|
int change_at, change_positive;
|
|
int16_t s, last_s = 0;
|
|
int threshold;
|
|
int i;
|
|
|
|
/* levels at total reverse */
|
|
level_min = 32767;
|
|
level_max = -32768;
|
|
change_max = -1;
|
|
change_at = -1;
|
|
change_positive = -1;
|
|
|
|
for (i = 0; i < len; i++) {
|
|
last_s = s;
|
|
s = spl[pos++];
|
|
if (pos == len)
|
|
pos = 0;
|
|
if (i > 0) {
|
|
if (s - last_s > change_max) {
|
|
change_max = s - last_s;
|
|
change_at = i;
|
|
change_positive = 1;
|
|
} else if (last_s - s > change_max) {
|
|
change_max = last_s - s;
|
|
change_at = i;
|
|
change_positive = 0;
|
|
}
|
|
}
|
|
if (s > level_max)
|
|
level_max = s;
|
|
if (s < level_min)
|
|
level_min = s;
|
|
}
|
|
/* for first bit, we have only half of the modulation deviation, so we divide the threshold by two */
|
|
if (fsk->cnetz->dsp_mode == DSP_MODE_SPK_V && bit_count == 0)
|
|
threshold = level_threshold / 2;
|
|
else
|
|
threshold = level_threshold;
|
|
/* if we are not in sync, for every detected change we set
|
|
* next_bit to 1.5, so we wait 1.5 bits for next change
|
|
* if it is not received within this time, there is no change,
|
|
* so the bit does not change.
|
|
* if we are in sync, we remember last change. after 1.5
|
|
* bits after sync average, we measure the first bit
|
|
* and then all subsequent bits after 1.0 bits */
|
|
//DEBUG_DECODER printf("next_bit=%.4f\n", next_bit);
|
|
if (level_max - level_min > threshold && change_at == half) {
|
|
#ifdef DEBUG_DECODER
|
|
DEBUG_DECODER
|
|
printf("receive bit change to %d (level=%d, threshold=%d)\n", change_positive, level_max - level_min, threshold);
|
|
#endif
|
|
last_change_positive = change_positive;
|
|
if (!sync) {
|
|
next_bit = 1.5;
|
|
got_bit(fsk, change_positive, level_max - level_min);
|
|
}
|
|
}
|
|
if (next_bit <= 0.0) {
|
|
#ifdef DEBUG_DECODER
|
|
DEBUG_DECODER {
|
|
if (sync)
|
|
printf("sampling here bit %d\n", last_change_positive);
|
|
else
|
|
printf("no bit change\n");
|
|
}
|
|
#endif
|
|
next_bit += 1.0;
|
|
got_bit(fsk, last_change_positive, 0);
|
|
}
|
|
next_bit -= bits_per_sample;
|
|
}
|
|
|
|
/* receive FM signal from receiver */
|
|
void fsk_fm_demod(fsk_fm_demod_t *fsk, int16_t *samples, int length)
|
|
{
|
|
int i;
|
|
double t;
|
|
|
|
len = fsk->bit_buffer_len;
|
|
half = fsk->bit_buffer_half;
|
|
spl = fsk->bit_buffer_spl;
|
|
speech_buffer = fsk->speech_buffer;
|
|
speech_size = fsk->speech_size;
|
|
speech_count = fsk->speech_count;
|
|
bits_per_sample = fsk->bits_per_sample;
|
|
level_threshold = fsk->level_threshold;
|
|
pos = fsk->bit_buffer_pos;
|
|
next_bit = fsk->next_bit;
|
|
sync = fsk->sync;
|
|
last_change_positive = fsk->last_change_positive;
|
|
sync_level = fsk->sync_level;
|
|
sync_time = fsk->sync_time;
|
|
sync_jitter = fsk->sync_jitter;
|
|
bit_time = fsk->bit_time;
|
|
bit_time_uncorrected = fsk->bit_time_uncorrected;
|
|
bit_count = fsk->bit_count;
|
|
|
|
/* process signaling block, sample by sample */
|
|
for (i = 0; i < length; i++) {
|
|
spl[pos++] = samples[i];
|
|
if (pos == len)
|
|
pos = 0;
|
|
/* for each sample process buffer */
|
|
if (fsk->cnetz->dsp_mode != DSP_MODE_SPK_V) {
|
|
#ifdef DEBUG_DECODER
|
|
DEBUG_DECODER
|
|
puts(debug_amplitude((double)samples[i] / 32768.0));
|
|
#endif
|
|
find_change(fsk);
|
|
} else {
|
|
/* in distributed signaling, measure over 5 bits, but ignore 5th bit.
|
|
* also reset next_bit, as soon as we reach the window */
|
|
|
|
/* note that we start from 0.5, because we detect change 0.5 bits later,
|
|
* because the detector of the change is in the middle of the 1 bit
|
|
* search window */
|
|
t = fmod(bit_time, BITS_PER_SPK_BLOCK);
|
|
if (t < 0.5) {
|
|
next_bit = 1.0 - bits_per_sample;
|
|
#ifdef DEBUG_DECODER
|
|
if (bit_count) {
|
|
DEBUG_DECODER
|
|
printf("start spk_block bit count=%d\n", bit_count);
|
|
}
|
|
#endif
|
|
bit_count = 0;
|
|
} else
|
|
if (t >= 0.5 && t < 5.5) {
|
|
#ifdef DEBUG_DECODER
|
|
DEBUG_DECODER
|
|
puts(debug_amplitude((double)samples[i] / 32768.0));
|
|
#endif
|
|
find_change(fsk);
|
|
} else
|
|
if (t >= 5.5 && t < 65.5) {
|
|
/* get audio for the duration of 60 bits */
|
|
/* prevent overflow, if speech_size != 0 and SPK_V
|
|
* has been restarted. */
|
|
if (speech_count <= speech_size)
|
|
speech_buffer[speech_count++] = samples[i];
|
|
} else
|
|
if (t >= 65.5) {
|
|
if (speech_count) {
|
|
unshrink_speech(fsk->cnetz, speech_buffer, speech_count);
|
|
speech_count = 0;
|
|
}
|
|
}
|
|
|
|
}
|
|
bit_time += bits_per_sample;
|
|
if (bit_time >= BITS_PER_SUPERFRAME) {
|
|
bit_time -= BITS_PER_SUPERFRAME;
|
|
}
|
|
/* another clock is used to measure actual super frame time */
|
|
bit_time_uncorrected += bits_per_sample;
|
|
if (bit_time_uncorrected >= BITS_PER_SUPERFRAME) {
|
|
bit_time_uncorrected -= BITS_PER_SUPERFRAME;
|
|
calc_clock_speed(fsk->cnetz, fsk->cnetz->sender.samplerate * 24 / 10, 0, 1);
|
|
}
|
|
}
|
|
|
|
fsk->level_threshold = level_threshold;
|
|
fsk->bit_buffer_pos = pos;
|
|
fsk->speech_count = speech_count;
|
|
fsk->next_bit = next_bit;
|
|
fsk->sync = sync;
|
|
fsk->last_change_positive = last_change_positive;
|
|
fsk->sync_level = sync_level;
|
|
fsk->sync_time = sync_time;
|
|
fsk->sync_jitter = sync_jitter;
|
|
fsk->bit_time = bit_time;
|
|
fsk->bit_time_uncorrected = bit_time_uncorrected;
|
|
fsk->bit_count = bit_count;
|
|
}
|
|
|
|
void fsk_correct_sync(cnetz_t *cnetz, double offset)
|
|
{
|
|
bit_time = fmod(bit_time - offset + BITS_PER_SUPERFRAME, BITS_PER_SUPERFRAME);
|
|
}
|
|
|