mirror of
https://github.com/f4exb/sdrangel.git
synced 2024-11-09 01:56:05 -05:00
275 lines
9.9 KiB
C++
275 lines
9.9 KiB
C++
///////////////////////////////////////////////////////////////////////////////////
|
|
// Copyright (C) 2020 Edouard Griffiths, F4EXB //
|
|
// //
|
|
// Inspired by: https://github.com/myriadrf/LoRa-SDR //
|
|
// //
|
|
// 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 as 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 V3 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/>. //
|
|
///////////////////////////////////////////////////////////////////////////////////
|
|
|
|
#ifndef PLUGINS_CHANNELTX_MODCHIRPCHAT_CHIRPCHATMODENCODERLORA_H_
|
|
#define PLUGINS_CHANNELTX_MODCHIRPCHAT_CHIRPCHATMODENCODERLORA_H_
|
|
|
|
#include <vector>
|
|
#include <QByteArray>
|
|
|
|
class ChirpChatModEncoderLoRa
|
|
{
|
|
public:
|
|
static void addChecksum(QByteArray& bytes);
|
|
static void encodeBytes(
|
|
const QByteArray& bytes,
|
|
std::vector<unsigned short>& symbols,
|
|
unsigned int nbSymbolBits,
|
|
bool hasHeader,
|
|
bool hasCRC,
|
|
unsigned int nbParityBits
|
|
);
|
|
|
|
private:
|
|
static void encodeFec(
|
|
std::vector<uint8_t> &codewords,
|
|
unsigned int nbParityBits,
|
|
unsigned int& cOfs,
|
|
unsigned int& dOfs,
|
|
const uint8_t *bytes,
|
|
const unsigned int codewordCount
|
|
);
|
|
|
|
static const unsigned int headerParityBits = 4;
|
|
static const unsigned int headerSymbols = 8;
|
|
static const unsigned int headerCodewords = 5;
|
|
|
|
/***********************************************************************
|
|
* Round functions
|
|
**********************************************************************/
|
|
static inline unsigned roundUp(unsigned num, unsigned factor)
|
|
{
|
|
return ((num + factor - 1) / factor) * factor;
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Encode a 4 bit word into a 8 bits with parity
|
|
* Non standard version used in sx1272.
|
|
* https://en.wikipedia.org/wiki/Hamming_code
|
|
**********************************************************************/
|
|
static inline unsigned char encodeHamming84sx(const unsigned char x)
|
|
{
|
|
auto d0 = (x >> 0) & 0x1;
|
|
auto d1 = (x >> 1) & 0x1;
|
|
auto d2 = (x >> 2) & 0x1;
|
|
auto d3 = (x >> 3) & 0x1;
|
|
|
|
unsigned char b = x & 0xf;
|
|
b |= (d0 ^ d1 ^ d2) << 4;
|
|
b |= (d1 ^ d2 ^ d3) << 5;
|
|
b |= (d0 ^ d1 ^ d3) << 6;
|
|
b |= (d0 ^ d2 ^ d3) << 7;
|
|
return b;
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Encode a 4 bit word into a 7 bits with parity.
|
|
* Non standard version used in sx1272.
|
|
**********************************************************************/
|
|
static inline unsigned char encodeHamming74sx(const unsigned char x)
|
|
{
|
|
auto d0 = (x >> 0) & 0x1;
|
|
auto d1 = (x >> 1) & 0x1;
|
|
auto d2 = (x >> 2) & 0x1;
|
|
auto d3 = (x >> 3) & 0x1;
|
|
|
|
unsigned char b = x & 0xf;
|
|
b |= (d0 ^ d1 ^ d2) << 4;
|
|
b |= (d1 ^ d2 ^ d3) << 5;
|
|
b |= (d0 ^ d1 ^ d3) << 6;
|
|
return b;
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Encode a 4 bit word into a 6 bits with parity.
|
|
**********************************************************************/
|
|
static inline unsigned char encodeParity64(const unsigned char b)
|
|
{
|
|
auto x = b ^ (b >> 1) ^ (b >> 2);
|
|
auto y = x ^ b ^ (b >> 3);
|
|
return ((x & 1) << 4) | ((y & 1) << 5) | (b & 0xf);
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Encode a 4 bit word into a 5 bits with parity.
|
|
**********************************************************************/
|
|
static inline unsigned char encodeParity54(const unsigned char b)
|
|
{
|
|
auto x = b ^ (b >> 2);
|
|
x = x ^ (x >> 1);
|
|
return (b & 0xf) | ((x << 4) & 0x10);
|
|
}
|
|
|
|
/***********************************************************************
|
|
* CRC reverse engineered from Sx1272 data stream.
|
|
* Modified CCITT crc with masking of the output with an 8bit lfsr
|
|
**********************************************************************/
|
|
static inline uint16_t crc16sx(uint16_t crc, const uint16_t poly)
|
|
{
|
|
for (int i = 0; i < 8; i++)
|
|
{
|
|
if (crc & 0x8000) {
|
|
crc = (crc << 1) ^ poly;
|
|
} else {
|
|
crc <<= 1;
|
|
}
|
|
}
|
|
|
|
return crc;
|
|
}
|
|
|
|
static inline uint8_t xsum8(uint8_t t)
|
|
{
|
|
t ^= t >> 4;
|
|
t ^= t >> 2;
|
|
t ^= t >> 1;
|
|
|
|
return (t & 1);
|
|
}
|
|
|
|
static inline uint16_t sx1272DataChecksum(const uint8_t *data, int length)
|
|
{
|
|
uint16_t res = 0;
|
|
uint8_t v = 0xff;
|
|
uint16_t crc = 0;
|
|
|
|
for (int i = 0; i < length; i++)
|
|
{
|
|
crc = crc16sx(res, 0x1021);
|
|
v = xsum8(v & 0xB8) | (v << 1);
|
|
res = crc ^ data[i];
|
|
}
|
|
|
|
res ^= v;
|
|
v = xsum8(v & 0xB8) | (v << 1);
|
|
res ^= v << 8;
|
|
|
|
return res;
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Specific checksum for header
|
|
**********************************************************************/
|
|
static inline uint8_t headerChecksum(const uint8_t *h)
|
|
{
|
|
auto a0 = (h[0] >> 4) & 0x1;
|
|
auto a1 = (h[0] >> 5) & 0x1;
|
|
auto a2 = (h[0] >> 6) & 0x1;
|
|
auto a3 = (h[0] >> 7) & 0x1;
|
|
|
|
auto b0 = (h[0] >> 0) & 0x1;
|
|
auto b1 = (h[0] >> 1) & 0x1;
|
|
auto b2 = (h[0] >> 2) & 0x1;
|
|
auto b3 = (h[0] >> 3) & 0x1;
|
|
|
|
auto c0 = (h[1] >> 0) & 0x1;
|
|
auto c1 = (h[1] >> 1) & 0x1;
|
|
auto c2 = (h[1] >> 2) & 0x1;
|
|
auto c3 = (h[1] >> 3) & 0x1;
|
|
|
|
uint8_t res;
|
|
res = (a0 ^ a1 ^ a2 ^ a3) << 4;
|
|
res |= (a3 ^ b1 ^ b2 ^ b3 ^ c0) << 3;
|
|
res |= (a2 ^ b0 ^ b3 ^ c1 ^ c3) << 2;
|
|
res |= (a1 ^ b0 ^ b2 ^ c0 ^ c1 ^ c2) << 1;
|
|
res |= a0 ^ b1 ^ c0 ^ c1 ^ c2 ^ c3;
|
|
|
|
return res;
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Whitening generator reverse engineered from Sx1272 data stream.
|
|
* Each bit of a codeword is combined with the output from a different position in the whitening sequence.
|
|
**********************************************************************/
|
|
static inline void Sx1272ComputeWhitening(uint8_t *buffer, uint16_t bufferSize, const int bitOfs, const int nbParityBits)
|
|
{
|
|
static const int ofs0[8] = {6,4,2,0,-112,-114,-302,-34 }; // offset into sequence for each bit
|
|
static const int ofs1[5] = {6,4,2,0,-360 }; // different offsets used for single parity mode (1 == nbParityBits)
|
|
static const int whiten_len = 510; // length of whitening sequence
|
|
static const uint64_t whiten_seq[8] = { // whitening sequence
|
|
0x0102291EA751AAFFL,0xD24B050A8D643A17L,0x5B279B671120B8F4L,0x032B37B9F6FB55A2L,
|
|
0x994E0F87E95E2D16L,0x7CBCFC7631984C26L,0x281C8E4F0DAEF7F9L,0x1741886EB7733B15L
|
|
};
|
|
const int *ofs = (1 == nbParityBits) ? ofs1 : ofs0;
|
|
int i, j;
|
|
|
|
for (j = 0; j < bufferSize; j++)
|
|
{
|
|
uint8_t x = 0;
|
|
|
|
for (i = 0; i < 4 + nbParityBits; i++)
|
|
{
|
|
int t = (ofs[i] + j + bitOfs + whiten_len) % whiten_len;
|
|
|
|
if (whiten_seq[t >> 6] & ((uint64_t)1 << (t & 0x3F))) {
|
|
x |= 1 << i;
|
|
}
|
|
}
|
|
|
|
buffer[j] ^= x;
|
|
}
|
|
}
|
|
|
|
/***********************************************************************
|
|
* Diagonal interleaver + deinterleaver
|
|
**********************************************************************/
|
|
static inline void diagonalInterleaveSx(
|
|
const uint8_t *codewords,
|
|
const size_t numCodewords,
|
|
uint16_t *symbols,
|
|
const size_t nbSymbolBits,
|
|
const size_t nbParityBits
|
|
)
|
|
{
|
|
for (size_t x = 0; x < numCodewords / nbSymbolBits; x++)
|
|
{
|
|
const size_t cwOff = x*nbSymbolBits;
|
|
const size_t symOff = x*(4 + nbParityBits);
|
|
|
|
for (size_t k = 0; k < 4 + nbParityBits; k++)
|
|
{
|
|
for (size_t m = 0; m < nbSymbolBits; m++)
|
|
{
|
|
const size_t i = (m + k + nbSymbolBits) % nbSymbolBits;
|
|
const auto bit = (codewords[cwOff + i] >> k) & 0x1;
|
|
symbols[symOff + k] |= (bit << m);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/***********************************************************************
|
|
* https://en.wikipedia.org/wiki/Gray_code
|
|
**********************************************************************/
|
|
|
|
/*
|
|
* A more efficient version, for Gray codes of 16 or fewer bits.
|
|
*/
|
|
static inline unsigned short grayToBinary16(unsigned short num)
|
|
{
|
|
num = num ^ (num >> 8);
|
|
num = num ^ (num >> 4);
|
|
num = num ^ (num >> 2);
|
|
num = num ^ (num >> 1);
|
|
return num;
|
|
}
|
|
};
|
|
|
|
#endif // PLUGINS_CHANNELTX_MODCHIRPCHAT_CHIRPCHATMODENCODERLORA_H_
|