Files
VasylSamoilov 396f8aef41 POCSAG: show '?' for characters from uncorrectable codewords (#3138)
Use per-character error tracking to substitute '?' when a
character's bits came from an uncorrectable codeword (BCH error_count >= 3). 
Previously these displayed as garbage.
2026-04-18 18:56:33 +02:00

722 lines
24 KiB
C++

/*
* Copyright (C) 2015 Jared Boone, ShareBrained Technology, Inc.
* Copyright (C) 2016 Furrtek
*
* This file is part of PortaPack.
*
* 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, 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; see the file COPYING. If not, write to
* the Free Software Foundation, Inc., 51 Franklin Street,
* Boston, MA 02110-1301, USA.
*/
#include "pocsag.hpp"
#include "baseband_api.hpp"
#include "portapack.hpp"
using namespace portapack;
#include "string_format.hpp"
#include "utility.hpp"
namespace pocsag {
std::string bitrate_str(BitRate bitrate) {
switch (bitrate) {
case BitRate::FSK512:
return "512bps ";
case BitRate::FSK1200:
return "1200bps";
case BitRate::FSK2400:
return "2400bps";
default:
return "????";
}
}
std::string flag_str(PacketFlag packetflag) {
switch (packetflag) {
case PacketFlag::NORMAL:
return "OK";
case PacketFlag::TIMED_OUT:
return "TIMED OUT";
default:
return "";
}
}
void insert_BCH(BCHCode& BCH_code, uint32_t* codeword) {
uint32_t parity = 0;
int data[21];
int bit;
int* bb;
size_t c;
for (c = 0; c < 21; c++) {
bit = (((*codeword) << c) & 0x80000000U) ? 1 : 0;
if (bit) parity++;
data[c] = bit;
}
bb = BCH_code.encode(data);
// Make sure ECC bits are cleared
(*codeword) &= 0xFFFFF801;
for (c = 0; c < 10; c++) {
bit = bb[c];
(*codeword) |= (bit << (10 - c));
if (bit) parity++;
}
// Even parity
(*codeword) |= (parity & 1);
}
uint32_t get_digit_code(char code) {
if ((code >= '0') && (code <= '9')) {
code -= '0';
} else {
if (code == 'S')
code = 10;
else if (code == 'U')
code = 11;
else if (code == ' ')
code = 12;
else if (code == '-')
code = 13;
else if (code == ']')
code = 14;
else if (code == '[')
code = 15;
else
code = 12;
}
code = ((code & 0x0C) >> 2) | ((code & 0x03) << 2); // ----3210 -> ----1032
code = ((code & 0x0A) >> 1) | ((code & 0x05) << 1); // ----1032 -> ----0123
return code;
}
void pocsag_encode(const MessageType type, BCHCode& BCH_code, const uint32_t function, const std::string message, const uint32_t address, std::vector<uint32_t>& codewords) {
size_t b, c, address_slot;
size_t bit_idx, char_idx = 0;
uint32_t codeword, digit_code;
char ascii_char = 0;
size_t message_size = message.size();
// Preamble
for (b = 0; b < (POCSAG_PREAMBLE_LENGTH / 32); b++) {
codewords.push_back(0xAAAAAAAA);
}
// Address
codeword = (address & 0x1FFFF8U) << 10;
address_slot = (address & 7) * 2;
// Function
codeword |= (function << 11);
insert_BCH(BCH_code, &codeword);
// Address batch
codewords.push_back(POCSAG_SYNCWORD);
for (c = 0; c < 16; c++) {
if (c == address_slot) {
codewords.push_back(codeword);
if (type != MessageType::ADDRESS_ONLY) break;
} else
codewords.push_back(POCSAG_IDLEWORD);
}
if (type == MessageType::ADDRESS_ONLY) return; // Done.
c++;
codeword = 0;
bit_idx = 20 + 11;
// Messages batch(es)
do {
if (c == 0) codewords.push_back(POCSAG_SYNCWORD);
for (; c < 16; c++) {
// Fill up 20 bits
if (type == MessageType::ALPHANUMERIC) {
if ((char_idx < message_size) || (ascii_char)) {
do {
bit_idx -= 7;
if (char_idx < message_size)
ascii_char = message[char_idx] & 0x7F;
else
ascii_char = 0; // Codeword padding
// Bottom's up
ascii_char = (ascii_char & 0xF0) >> 4 | (ascii_char & 0x0F) << 4; // *6543210 -> 3210*654
ascii_char = (ascii_char & 0xCC) >> 2 | (ascii_char & 0x33) << 2; // 3210*654 -> 103254*6
ascii_char = (ascii_char & 0xAA) >> 2 | (ascii_char & 0x55); // 103254*6 -> *0123456
codeword |= (ascii_char << bit_idx);
char_idx++;
} while (bit_idx > 11);
codeword &= 0x7FFFF800; // Trim data
codeword |= 0x80000000; // Message type
insert_BCH(BCH_code, &codeword);
codewords.push_back(codeword);
if (bit_idx != 11) {
bit_idx = 20 + bit_idx;
codeword = ascii_char << bit_idx;
} else {
bit_idx = 20 + 11;
codeword = 0;
}
} else {
codewords.push_back(POCSAG_IDLEWORD); // Batch padding
}
} else if (type == MessageType::NUMERIC_ONLY) {
if (char_idx < message_size) {
do {
bit_idx -= 4;
if (char_idx < message_size)
digit_code = get_digit_code(message[char_idx]);
else
digit_code = 3; // Space (codeword padding)
codeword |= (digit_code << bit_idx);
char_idx++;
} while (bit_idx > 11);
codeword |= 0x80000000; // Message type
insert_BCH(BCH_code, &codeword);
codewords.push_back(codeword);
bit_idx = 20 + 11;
codeword = 0;
} else {
codewords.push_back(POCSAG_IDLEWORD); // Batch padding
}
}
}
c = 0;
} while (char_idx < message_size);
}
// ----------------------------------------------------------------------------
// EccContainer
// ----------------------------------------------------------------------------
EccContainer::EccContainer() {
setup_ecc();
}
void EccContainer::setup_ecc() {
unsigned int srr = 0x3b4;
unsigned int i, n, j, k;
/* calculate all information needed to implement error correction */
// Note : this is only for 31,21 code used in pocsag & flex
// one should probably also make use of 32nd parity bit
for (i = 0; i <= 20; i++) {
ecs[i] = srr;
if ((srr & 0x01) != 0)
srr = (srr >> 1) ^ 0x3B4;
else
srr = srr >> 1;
}
/* bch holds a syndrome look-up table telling which bits to correct */
// first 5 bits hold location of first error; next 5 bits hold location
// of second error; bits 12 & 13 tell how many bits are bad
for (i = 0; i < 1024; i++) bch[i] = 0;
/* two errors in data */
for (n = 0; n <= 20; n++) {
for (i = 0; i <= 20; i++) {
j = (i << 5) + n;
k = ecs[n] ^ ecs[i];
bch[k] = j + 0x2000;
}
}
/* one error in data */
for (n = 0; n <= 20; n++) {
k = ecs[n];
j = n + (0x1f << 5);
bch[k] = j + 0x1000;
}
/* one error in data and one error in ecc portion */
for (n = 0; n <= 20; n++) {
for (i = 0; i < 10; i++) /* ecc screwed up bit */
{
k = ecs[n] ^ (1 << i);
j = n + (0x1f << 5);
bch[k] = j + 0x2000;
}
}
/* one error in ecc */
for (n = 0; n < 10; n++) {
k = 1 << n;
bch[k] = 0x3ff + 0x1000;
}
/* two errors in ecc */
for (n = 0; n < 10; n++) {
for (i = 0; i < 10; i++) {
if (i != n) {
k = (1 << n) ^ (1 << i);
bch[k] = 0x3ff + 0x2000;
}
}
}
}
int EccContainer::error_correct(uint32_t& val) {
int i, synd, errl, acc, pari, ecc, b1, b2;
errl = 0;
pari = 0;
ecc = 0;
for (i = 31; i >= 11; --i) {
if (val & (1 << i)) {
ecc = ecc ^ ecs[31 - i];
pari = pari ^ 0x01;
}
}
acc = 0;
for (i = 10; i >= 1; --i) {
acc = acc << 1;
if (val & (1 << i)) {
acc = acc ^ 0x01;
}
}
synd = ecc ^ acc;
errl = 0;
if (synd != 0) /* if nonzero syndrome we have error */
{
if (bch[synd] != 0) /* check for correctable error */
{
b1 = bch[synd] & 0x1f;
b2 = bch[synd] >> 5;
b2 = b2 & 0x1f;
if (b2 != 0x1f) {
val ^= 0x01 << (31 - b2);
ecc = ecc ^ ecs[b2];
}
if (b1 != 0x1f) {
val ^= 0x01 << (31 - b1);
ecc = ecc ^ ecs[b1];
}
errl = bch[synd] >> 12;
} else {
errl = 3;
}
if (errl == 1) pari = pari ^ 0x01;
}
if (errl == 4) errl = 3;
return errl;
}
// ----------------------------------------------------------------------------
// Numeric character table: 4-bit BCD -> ASCII
// ----------------------------------------------------------------------------
static const char numeric_chars[16] = {
'0', '1', '2', '3', '4', '5', '6', '7',
'8', '9', 'R', 'U', ' ', '-', ']', '['};
// Extract and bit-reverse a 4-bit nibble from a message codeword.
// POCSAG numeric digits are transmitted LSB first, so bit 30 (first transmitted)
// is the LSB of the digit value, not the MSB.
static uint8_t decode_nibble(uint32_t codeword, int nibble_idx) {
int bit_pos = 30 - nibble_idx * 4;
// bit_pos is the first transmitted bit (LSB of digit)
// bit_pos-3 is the last transmitted bit (MSB of digit)
uint8_t n = 0;
n |= ((codeword >> (bit_pos - 3)) & 1) << 3; // MSB of digit
n |= ((codeword >> (bit_pos - 2)) & 1) << 2;
n |= ((codeword >> (bit_pos - 1)) & 1) << 1;
n |= ((codeword >> (bit_pos - 0)) & 1) << 0; // LSB of digit
return n;
}
// ----------------------------------------------------------------------------
// Heuristic message type detection (first batch only)
// ----------------------------------------------------------------------------
// Count trailing fill characters in alpha buffer (NULL or space).
static int count_alpha_fill(const std::string& data) {
if (data.empty()) return 0;
int fill = 0;
for (int i = data.size() - 1; i >= 0; --i) {
char c = data[i];
if (c == '\0' || c == ' ')
fill++;
else
break;
}
return fill;
}
// Count trailing fill nibbles (0xC = space) in numeric buffer.
static int count_numeric_fill(const uint8_t* nibbles, int count) {
int fill = 0;
for (int i = count - 1; i >= 0; --i) {
if (nibbles[i] == 0x0C)
fill++;
else
break;
}
return fill;
}
// Score alpha interpretation: +3 alphanumeric/space, -2 other printable, -5 control.
static int score_alpha(const std::string& data, int fill) {
int score = 0;
int content = 0;
int len = data.size();
for (int i = 0; i < len; ++i) {
unsigned char c = data[i];
// Skip trailing fill
if (i >= len - fill && (c == 0 || c == ' '))
continue;
if (c == 0)
continue;
content++;
if ((c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') ||
(c >= '0' && c <= '9') || c == ' ')
score += 3;
else if (c >= 0x20 && c <= 0x7E)
score -= 2;
else if (c == '\n' || c == '\r' || c == '\t' || c == 0x04)
score += 0;
else
score -= 5;
}
if (content > 0)
score += fill * fill * 3 + fill * 5;
return score;
}
// Score numeric interpretation on raw nibbles.
static int score_numeric(const uint8_t* nibbles, int count, int fill) {
int raw_score = 0;
int scored = 0;
int digits = 0;
int u_count = 0;
// Pre-scan for U nibbles
for (int i = 0; i < count; ++i) {
if (nibbles[i] == 0x0C && i >= count - fill)
continue;
if (nibbles[i] == 0x0B)
u_count++;
}
bool urgent_prefix = (u_count == 1);
for (int i = 0; i < count; ++i) {
uint8_t n = nibbles[i];
if (n == 0x0C && i >= count - fill)
continue;
scored++;
if (n <= 0x09) {
raw_score += 3;
digits++;
} else if (n == 0x0B) {
raw_score += urgent_prefix ? -1 : -15;
} else if (n == 0x0A) {
raw_score -= 5;
} else {
raw_score -= 2;
}
}
int score = scored > 0 ? raw_score * 4 / 7 : 0;
// Phone numbers have at most ~15 digits
if (digits > 15)
score -= (digits - 15) * 5;
// Fill bonus (weaker than alpha: 1/16 vs 1/128 coincidence rate)
score += fill * fill;
return score;
}
DetectedType detect_message_type(const std::string& alpha,
const uint8_t* nibbles,
uint8_t nibble_count,
uint8_t msg_codewords) {
if (alpha.empty() && nibble_count == 0)
return DET_TONE;
// Long messages can't be numeric (phone numbers are short)
if (msg_codewords >= 8)
return DET_ALPHA;
int alpha_fill = count_alpha_fill(alpha);
int numeric_fill = count_numeric_fill(nibbles, nibble_count);
int sa = score_alpha(alpha, alpha_fill) + 2; // Alpha prior bias
int sn = score_numeric(nibbles, nibble_count, numeric_fill);
// Short message boost for alpha (1-2 data codewords)
if (msg_codewords <= 3)
sa += 3;
return (sn > sa) ? DET_NUMERIC : DET_ALPHA;
}
// ----------------------------------------------------------------------------
// Batch decoder
// ----------------------------------------------------------------------------
bool pocsag_decode_batch(const POCSAGPacket& batch, POCSAGState& state) {
constexpr uint8_t codeword_max = 16;
state.output.clear();
/* Only reset numeric accumulator when starting a new message,
* not on continuation batches — numeric_buf must be cumulative
* across multi-batch messages. */
const bool continuing_numeric =
(state.mode != STATE_HAVE_ADDRESS) &&
(state.out_type == MESSAGE) &&
state.type_decided &&
(state.detected == DET_NUMERIC);
if (!continuing_numeric)
state.numeric_len = 0;
/* Preserve new_message across batch boundary when STATE_HAVE_ADDRESS
* persists — the address was at the end of the previous batch and
* we haven't displayed it yet. Otherwise reset for this batch. */
if (state.mode != STATE_HAVE_ADDRESS)
state.new_message = false;
// Temporary nibble buffer for first-batch numeric decode.
uint8_t nibbles[max_batch_nibbles];
uint8_t nibble_count = 0;
uint8_t msg_codewords = 0;
// Also build raw alpha for heuristic (before non-printable replacement).
std::string raw_alpha{};
// Track whether any characters came from uncorrectable codewords.
// If so, heuristic scoring is unreliable — skip it and default to alpha.
bool has_bad_chars = false;
while (state.codeword_index < codeword_max) {
auto codeword = batch[state.codeword_index];
bool is_address = (codeword & 0x80000000U) == 0;
// Single ECC call: fix errors and get error count.
auto error_count = state.ecc->error_correct(codeword);
switch (state.mode) {
case STATE_CLEAR:
if (is_address && codeword != POCSAG_IDLEWORD) {
state.function = (codeword >> 11) & 3;
state.address = (codeword >> 10) & 0x1FFFF8U;
/* Frame number = lower 3 bits of RIC, derived from the
* address codeword's position in the batch (not the
* message codeword's position). codeword_index 0-15
* maps to frames 0-7 via index >> 1. */
state.address |= (state.codeword_index >> 1);
state.mode = STATE_HAVE_ADDRESS;
state.out_type = ADDRESS;
state.errors = error_count;
state.new_message = true;
state.type_decided = false;
state.detected = DET_UNKNOWN;
state.ascii_idx = 0;
state.ascii_data = 0;
state.prev_cw_err = 0;
state.cur_cw_err = 0;
nibble_count = 0;
msg_codewords = 0;
raw_alpha.clear();
} else if (codeword == POCSAG_IDLEWORD) {
state.out_type = IDLE;
}
break;
case STATE_HAVE_ADDRESS:
if (is_address) {
// Got another address. Run heuristic before returning if we have pending data.
if (!state.type_decided && msg_codewords > 0) {
state.detected = has_bad_chars
? DET_ALPHA
: detect_message_type(raw_alpha, nibbles, nibble_count, msg_codewords);
state.type_decided = true;
state.msg_codewords = msg_codewords;
if (state.detected == DET_NUMERIC) {
state.numeric_len = 0;
for (uint8_t ni = 0; ni < nibble_count && state.numeric_len < sizeof(state.numeric_buf); ++ni)
state.numeric_buf[state.numeric_len++] = numeric_chars[nibbles[ni] & 0x0F];
}
}
state.mode = STATE_CLEAR;
return true;
}
/* Frame number already applied in STATE_CLEAR.
* Transition to message decoding. */
state.mode = STATE_GETTING_MSG;
[[fallthrough]];
case STATE_GETTING_MSG:
if (is_address) {
// Message ended. Run heuristic before returning.
if (!state.type_decided && msg_codewords > 0) {
state.detected = has_bad_chars
? DET_ALPHA
: detect_message_type(raw_alpha, nibbles, nibble_count, msg_codewords);
state.type_decided = true;
state.msg_codewords = msg_codewords;
if (state.detected == DET_NUMERIC) {
state.numeric_len = 0;
for (uint8_t ni = 0; ni < nibble_count && state.numeric_len < sizeof(state.numeric_buf); ++ni)
state.numeric_buf[state.numeric_len++] = numeric_chars[nibbles[ni] & 0x0F];
}
}
state.mode = STATE_CLEAR;
return true;
}
state.out_type = MESSAGE;
state.errors += error_count;
msg_codewords++;
// Track per-codeword error level for character coloring.
// 0=clean, 1-2=corrected, 3=uncorrectable.
state.prev_cw_err = state.cur_cw_err;
state.cur_cw_err = (error_count >= 3) ? 3 : error_count;
// --- Alpha decode (always) ---
// Bits remaining from previous codeword inherit prev_cw_err.
// New 20 bits from this codeword use cur_cw_err.
// Characters spanning boundary get the worst of both.
uint32_t bits_from_prev = state.ascii_idx; // leftover bits before adding new ones
state.ascii_data |= ((uint64_t)((codeword >> 11) & 0xFFFFF)) << (44 - state.ascii_idx);
state.ascii_idx += 20;
while (state.ascii_idx >= 7) {
// Per-character error level from codeword error tracking.
// Characters spanning a codeword boundary get the worst level.
uint8_t char_err;
if (bits_from_prev >= 7) {
// Entire character from previous codeword's leftover bits.
char_err = state.prev_cw_err;
bits_from_prev -= 7;
} else if (bits_from_prev > 0) {
// Character spans boundary.
char_err = std::max(state.prev_cw_err, state.cur_cw_err);
bits_from_prev = 0;
} else {
// Entirely from current codeword.
char_err = state.cur_cw_err;
}
// Extract top 7 bits from accumulator
char ascii_char = (state.ascii_data >> 57) & 0x7F;
state.ascii_data <<= 7;
state.ascii_idx -= 7;
// Reverse bits (LSB-first encoding)
ascii_char = (ascii_char & 0xF0) >> 4 | (ascii_char & 0x0F) << 4;
ascii_char = (ascii_char & 0xCC) >> 2 | (ascii_char & 0x33) << 2;
ascii_char = (ascii_char & 0xAA) >> 2 | (ascii_char & 0x55);
// Store raw char for heuristic
if (!state.type_decided)
raw_alpha += ascii_char;
// Substitute '?' for characters from uncorrectable codewords
if (char_err >= 3) {
state.output += "?";
has_bad_chars = true;
} else if (ascii_char < 32 || ascii_char > 126)
state.output += ".";
else
state.output += ascii_char;
}
// --- Numeric decode ---
// First batch: accumulate nibbles locally for heuristic scoring.
// Continuation batches: if already decided numeric, decode directly
// into numeric_buf for the app layer.
if (!state.type_decided && nibble_count + 5 <= max_batch_nibbles) {
for (int n = 0; n < 5; ++n) {
nibbles[nibble_count++] = decode_nibble(codeword, n);
}
} else if (state.type_decided && state.detected == DET_NUMERIC &&
state.numeric_len + 5 <= (uint8_t)sizeof(state.numeric_buf)) {
for (int n = 0; n < 5; ++n) {
uint8_t nib = decode_nibble(codeword, n);
state.numeric_buf[state.numeric_len++] = numeric_chars[nib & 0x0F];
}
}
break;
}
state.codeword_index++;
}
// End of batch. If we have message data and type not yet decided, run heuristic.
if (state.out_type == MESSAGE && !state.type_decided && msg_codewords > 0) {
state.detected = has_bad_chars
? DET_ALPHA
: detect_message_type(raw_alpha, nibbles, nibble_count, msg_codewords);
state.type_decided = true;
state.msg_codewords = msg_codewords;
if (state.detected == DET_NUMERIC) {
state.numeric_len = 0;
for (uint8_t ni = 0; ni < nibble_count && state.numeric_len < sizeof(state.numeric_buf); ++ni)
state.numeric_buf[state.numeric_len++] = numeric_chars[nibbles[ni] & 0x0F];
}
}
return false;
}
} /* namespace pocsag */