piotr | 437f546 | 2014-02-04 17:57:25 +0100 | [diff] [blame] | 1 | /* -*- c++ -*- */ |
| 2 | /* |
| 3 | * @file |
| 4 | * @author Piotr Krysik <pkrysik@stud.elka.pw.edu.pl> |
| 5 | * @section LICENSE |
| 6 | * |
| 7 | * This program is free software; you can redistribute it and/or modify |
| 8 | * it under the terms of the GNU General Public License as published by |
| 9 | * the Free Software Foundation; either version 3, or (at your option) |
| 10 | * any later version. |
| 11 | * |
| 12 | * This program is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 15 | * GNU General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU General Public License |
| 18 | * along with this program; see the file COPYING. If not, write to |
| 19 | * the Free Software Foundation, Inc., 51 Franklin Street, |
| 20 | * Boston, MA 02110-1301, USA. |
| 21 | */ |
| 22 | |
| 23 | /* |
| 24 | * viterbi_detector: |
| 25 | * This part does the detection of received sequnece. |
| 26 | * Employed algorithm is viterbi Maximum Likehood Sequence Estimation. |
| 27 | * At this moment it gives hard decisions on the output, but |
| 28 | * it was designed with soft decisions in mind. |
| 29 | * |
| 30 | * SYNTAX: void viterbi_detector( |
| 31 | * const gr_complex * input, |
| 32 | * unsigned int samples_num, |
| 33 | * gr_complex * rhh, |
| 34 | * unsigned int start_state, |
| 35 | * const unsigned int * stop_states, |
| 36 | * unsigned int stops_num, |
| 37 | * float * output) |
| 38 | * |
| 39 | * INPUT: input: Complex received signal afted matched filtering. |
| 40 | * samples_num: Number of samples in the input table. |
| 41 | * rhh: The autocorrelation of the estimated channel |
| 42 | * impulse response. |
| 43 | * start_state: Number of the start point. In GSM each burst |
| 44 | * starts with sequence of three bits (0,0,0) which |
| 45 | * indicates start point of the algorithm. |
| 46 | * stop_states: Table with numbers of possible stop states. |
| 47 | * stops_num: Number of possible stop states |
| 48 | * |
| 49 | * |
| 50 | * OUTPUT: output: Differentially decoded hard output of the algorithm: |
| 51 | * -1 for logical "0" and 1 for logical "1" |
| 52 | * |
| 53 | * SUB_FUNC: none |
| 54 | * |
| 55 | * TEST(S): Tested with real world normal burst. |
| 56 | */ |
| 57 | |
| 58 | #include <gnuradio/gr_complex.h> |
| 59 | #include <gsm_constants.h> |
| 60 | #define PATHS_NUM (1 << (CHAN_IMP_RESP_LENGTH-1)) |
| 61 | |
| 62 | void viterbi_detector(const gr_complex * input, unsigned int samples_num, gr_complex * rhh, unsigned int start_state, const unsigned int * stop_states, unsigned int stops_num, float * output) |
| 63 | { |
| 64 | float increment[8]; |
| 65 | float path_metrics1[16]; |
| 66 | float path_metrics2[16]; |
| 67 | float * new_path_metrics; |
| 68 | float * old_path_metrics; |
| 69 | float * tmp; |
| 70 | float trans_table[BURST_SIZE][16]; |
| 71 | float pm_candidate1, pm_candidate2; |
| 72 | bool real_imag; |
| 73 | float input_symbol_real, input_symbol_imag; |
| 74 | unsigned int i, sample_nr; |
| 75 | |
| 76 | /* |
| 77 | * Setup first path metrics, so only state pointed by start_state is possible. |
| 78 | * Start_state metric is equal to zero, the rest is written with some very low value, |
| 79 | * which makes them practically impossible to occur. |
| 80 | */ |
| 81 | for(i=0; i<PATHS_NUM; i++){ |
| 82 | path_metrics1[i]=(-10e30); |
| 83 | } |
| 84 | path_metrics1[start_state]=0; |
| 85 | |
| 86 | /* |
| 87 | * Compute Increment - a table of values which does not change for subsequent input samples. |
| 88 | * Increment is table of reference levels for computation of branch metrics: |
| 89 | * branch metric = (+/-)received_sample (+/-) reference_level |
| 90 | */ |
| 91 | increment[0] = -rhh[1].imag() -rhh[2].real() -rhh[3].imag() +rhh[4].real(); |
| 92 | increment[1] = rhh[1].imag() -rhh[2].real() -rhh[3].imag() +rhh[4].real(); |
| 93 | increment[2] = -rhh[1].imag() +rhh[2].real() -rhh[3].imag() +rhh[4].real(); |
| 94 | increment[3] = rhh[1].imag() +rhh[2].real() -rhh[3].imag() +rhh[4].real(); |
| 95 | increment[4] = -rhh[1].imag() -rhh[2].real() +rhh[3].imag() +rhh[4].real(); |
| 96 | increment[5] = rhh[1].imag() -rhh[2].real() +rhh[3].imag() +rhh[4].real(); |
| 97 | increment[6] = -rhh[1].imag() +rhh[2].real() +rhh[3].imag() +rhh[4].real(); |
| 98 | increment[7] = rhh[1].imag() +rhh[2].real() +rhh[3].imag() +rhh[4].real(); |
| 99 | |
| 100 | |
| 101 | /* |
| 102 | * Computation of path metrics and decisions (Add-Compare-Select). |
| 103 | * It's composed of two parts: one for odd input samples (imaginary numbers) |
| 104 | * and one for even samples (real numbers). |
| 105 | * Each part is composed of independent (parallelisable) statements like |
| 106 | * this one: |
| 107 | * pm_candidate1 = old_path_metrics[0] - input_symbol_real - increment[7]; |
| 108 | * pm_candidate2 = old_path_metrics[8] - input_symbol_real + increment[0]; |
| 109 | * if(pm_candidate1 > pm_candidate2){ |
| 110 | * new_path_metrics[0] = pm_candidate1; |
| 111 | * trans_table[sample_nr][0] = -1.0; |
| 112 | * } |
| 113 | * else{ |
| 114 | * new_path_metrics[0] = pm_candidate2; |
| 115 | * trans_table[sample_nr][0] = 1.0; |
| 116 | * } |
| 117 | * This is very good point for optimisations (SIMD or OpenMP) as it's most time |
| 118 | * consuming part of this function. |
| 119 | */ |
| 120 | sample_nr=0; |
| 121 | old_path_metrics=path_metrics1; |
| 122 | new_path_metrics=path_metrics2; |
| 123 | while(sample_nr<samples_num){ |
| 124 | //Processing imag states |
| 125 | real_imag=1; |
| 126 | input_symbol_imag = input[sample_nr].imag(); |
| 127 | |
| 128 | pm_candidate1 = old_path_metrics[0] + input_symbol_imag - increment[2]; |
| 129 | pm_candidate2 = old_path_metrics[8] + input_symbol_imag + increment[5]; |
| 130 | if(pm_candidate1 > pm_candidate2){ |
| 131 | new_path_metrics[0] = pm_candidate1; |
| 132 | trans_table[sample_nr][0] = -1.0; |
| 133 | } |
| 134 | else{ |
| 135 | new_path_metrics[0] = pm_candidate2; |
| 136 | trans_table[sample_nr][0] = 1.0; |
| 137 | } |
| 138 | |
| 139 | pm_candidate1 = old_path_metrics[0] - input_symbol_imag + increment[2]; |
| 140 | pm_candidate2 = old_path_metrics[8] - input_symbol_imag - increment[5]; |
| 141 | if(pm_candidate1 > pm_candidate2){ |
| 142 | new_path_metrics[1] = pm_candidate1; |
| 143 | trans_table[sample_nr][1] = -1.0; |
| 144 | } |
| 145 | else{ |
| 146 | new_path_metrics[1] = pm_candidate2; |
| 147 | trans_table[sample_nr][1] = 1.0; |
| 148 | } |
| 149 | |
| 150 | pm_candidate1 = old_path_metrics[1] + input_symbol_imag - increment[3]; |
| 151 | pm_candidate2 = old_path_metrics[9] + input_symbol_imag + increment[4]; |
| 152 | if(pm_candidate1 > pm_candidate2){ |
| 153 | new_path_metrics[2] = pm_candidate1; |
| 154 | trans_table[sample_nr][2] = -1.0; |
| 155 | } |
| 156 | else{ |
| 157 | new_path_metrics[2] = pm_candidate2; |
| 158 | trans_table[sample_nr][2] = 1.0; |
| 159 | } |
| 160 | |
| 161 | pm_candidate1 = old_path_metrics[1] - input_symbol_imag + increment[3]; |
| 162 | pm_candidate2 = old_path_metrics[9] - input_symbol_imag - increment[4]; |
| 163 | if(pm_candidate1 > pm_candidate2){ |
| 164 | new_path_metrics[3] = pm_candidate1; |
| 165 | trans_table[sample_nr][3] = -1.0; |
| 166 | } |
| 167 | else{ |
| 168 | new_path_metrics[3] = pm_candidate2; |
| 169 | trans_table[sample_nr][3] = 1.0; |
| 170 | } |
| 171 | |
| 172 | pm_candidate1 = old_path_metrics[2] + input_symbol_imag - increment[0]; |
| 173 | pm_candidate2 = old_path_metrics[10] + input_symbol_imag + increment[7]; |
| 174 | if(pm_candidate1 > pm_candidate2){ |
| 175 | new_path_metrics[4] = pm_candidate1; |
| 176 | trans_table[sample_nr][4] = -1.0; |
| 177 | } |
| 178 | else{ |
| 179 | new_path_metrics[4] = pm_candidate2; |
| 180 | trans_table[sample_nr][4] = 1.0; |
| 181 | } |
| 182 | |
| 183 | pm_candidate1 = old_path_metrics[2] - input_symbol_imag + increment[0]; |
| 184 | pm_candidate2 = old_path_metrics[10] - input_symbol_imag - increment[7]; |
| 185 | if(pm_candidate1 > pm_candidate2){ |
| 186 | new_path_metrics[5] = pm_candidate1; |
| 187 | trans_table[sample_nr][5] = -1.0; |
| 188 | } |
| 189 | else{ |
| 190 | new_path_metrics[5] = pm_candidate2; |
| 191 | trans_table[sample_nr][5] = 1.0; |
| 192 | } |
| 193 | |
| 194 | pm_candidate1 = old_path_metrics[3] + input_symbol_imag - increment[1]; |
| 195 | pm_candidate2 = old_path_metrics[11] + input_symbol_imag + increment[6]; |
| 196 | if(pm_candidate1 > pm_candidate2){ |
| 197 | new_path_metrics[6] = pm_candidate1; |
| 198 | trans_table[sample_nr][6] = -1.0; |
| 199 | } |
| 200 | else{ |
| 201 | new_path_metrics[6] = pm_candidate2; |
| 202 | trans_table[sample_nr][6] = 1.0; |
| 203 | } |
| 204 | |
| 205 | pm_candidate1 = old_path_metrics[3] - input_symbol_imag + increment[1]; |
| 206 | pm_candidate2 = old_path_metrics[11] - input_symbol_imag - increment[6]; |
| 207 | if(pm_candidate1 > pm_candidate2){ |
| 208 | new_path_metrics[7] = pm_candidate1; |
| 209 | trans_table[sample_nr][7] = -1.0; |
| 210 | } |
| 211 | else{ |
| 212 | new_path_metrics[7] = pm_candidate2; |
| 213 | trans_table[sample_nr][7] = 1.0; |
| 214 | } |
| 215 | |
| 216 | pm_candidate1 = old_path_metrics[4] + input_symbol_imag - increment[6]; |
| 217 | pm_candidate2 = old_path_metrics[12] + input_symbol_imag + increment[1]; |
| 218 | if(pm_candidate1 > pm_candidate2){ |
| 219 | new_path_metrics[8] = pm_candidate1; |
| 220 | trans_table[sample_nr][8] = -1.0; |
| 221 | } |
| 222 | else{ |
| 223 | new_path_metrics[8] = pm_candidate2; |
| 224 | trans_table[sample_nr][8] = 1.0; |
| 225 | } |
| 226 | |
| 227 | pm_candidate1 = old_path_metrics[4] - input_symbol_imag + increment[6]; |
| 228 | pm_candidate2 = old_path_metrics[12] - input_symbol_imag - increment[1]; |
| 229 | if(pm_candidate1 > pm_candidate2){ |
| 230 | new_path_metrics[9] = pm_candidate1; |
| 231 | trans_table[sample_nr][9] = -1.0; |
| 232 | } |
| 233 | else{ |
| 234 | new_path_metrics[9] = pm_candidate2; |
| 235 | trans_table[sample_nr][9] = 1.0; |
| 236 | } |
| 237 | |
| 238 | pm_candidate1 = old_path_metrics[5] + input_symbol_imag - increment[7]; |
| 239 | pm_candidate2 = old_path_metrics[13] + input_symbol_imag + increment[0]; |
| 240 | if(pm_candidate1 > pm_candidate2){ |
| 241 | new_path_metrics[10] = pm_candidate1; |
| 242 | trans_table[sample_nr][10] = -1.0; |
| 243 | } |
| 244 | else{ |
| 245 | new_path_metrics[10] = pm_candidate2; |
| 246 | trans_table[sample_nr][10] = 1.0; |
| 247 | } |
| 248 | |
| 249 | pm_candidate1 = old_path_metrics[5] - input_symbol_imag + increment[7]; |
| 250 | pm_candidate2 = old_path_metrics[13] - input_symbol_imag - increment[0]; |
| 251 | if(pm_candidate1 > pm_candidate2){ |
| 252 | new_path_metrics[11] = pm_candidate1; |
| 253 | trans_table[sample_nr][11] = -1.0; |
| 254 | } |
| 255 | else{ |
| 256 | new_path_metrics[11] = pm_candidate2; |
| 257 | trans_table[sample_nr][11] = 1.0; |
| 258 | } |
| 259 | |
| 260 | pm_candidate1 = old_path_metrics[6] + input_symbol_imag - increment[4]; |
| 261 | pm_candidate2 = old_path_metrics[14] + input_symbol_imag + increment[3]; |
| 262 | if(pm_candidate1 > pm_candidate2){ |
| 263 | new_path_metrics[12] = pm_candidate1; |
| 264 | trans_table[sample_nr][12] = -1.0; |
| 265 | } |
| 266 | else{ |
| 267 | new_path_metrics[12] = pm_candidate2; |
| 268 | trans_table[sample_nr][12] = 1.0; |
| 269 | } |
| 270 | |
| 271 | pm_candidate1 = old_path_metrics[6] - input_symbol_imag + increment[4]; |
| 272 | pm_candidate2 = old_path_metrics[14] - input_symbol_imag - increment[3]; |
| 273 | if(pm_candidate1 > pm_candidate2){ |
| 274 | new_path_metrics[13] = pm_candidate1; |
| 275 | trans_table[sample_nr][13] = -1.0; |
| 276 | } |
| 277 | else{ |
| 278 | new_path_metrics[13] = pm_candidate2; |
| 279 | trans_table[sample_nr][13] = 1.0; |
| 280 | } |
| 281 | |
| 282 | pm_candidate1 = old_path_metrics[7] + input_symbol_imag - increment[5]; |
| 283 | pm_candidate2 = old_path_metrics[15] + input_symbol_imag + increment[2]; |
| 284 | if(pm_candidate1 > pm_candidate2){ |
| 285 | new_path_metrics[14] = pm_candidate1; |
| 286 | trans_table[sample_nr][14] = -1.0; |
| 287 | } |
| 288 | else{ |
| 289 | new_path_metrics[14] = pm_candidate2; |
| 290 | trans_table[sample_nr][14] = 1.0; |
| 291 | } |
| 292 | |
| 293 | pm_candidate1 = old_path_metrics[7] - input_symbol_imag + increment[5]; |
| 294 | pm_candidate2 = old_path_metrics[15] - input_symbol_imag - increment[2]; |
| 295 | if(pm_candidate1 > pm_candidate2){ |
| 296 | new_path_metrics[15] = pm_candidate1; |
| 297 | trans_table[sample_nr][15] = -1.0; |
| 298 | } |
| 299 | else{ |
| 300 | new_path_metrics[15] = pm_candidate2; |
| 301 | trans_table[sample_nr][15] = 1.0; |
| 302 | } |
| 303 | tmp=old_path_metrics; |
| 304 | old_path_metrics=new_path_metrics; |
| 305 | new_path_metrics=tmp; |
| 306 | |
| 307 | sample_nr++; |
| 308 | if(sample_nr==samples_num) |
| 309 | break; |
| 310 | |
| 311 | //Processing real states |
| 312 | real_imag=0; |
| 313 | input_symbol_real = input[sample_nr].real(); |
| 314 | |
| 315 | pm_candidate1 = old_path_metrics[0] - input_symbol_real - increment[7]; |
| 316 | pm_candidate2 = old_path_metrics[8] - input_symbol_real + increment[0]; |
| 317 | if(pm_candidate1 > pm_candidate2){ |
| 318 | new_path_metrics[0] = pm_candidate1; |
| 319 | trans_table[sample_nr][0] = -1.0; |
| 320 | } |
| 321 | else{ |
| 322 | new_path_metrics[0] = pm_candidate2; |
| 323 | trans_table[sample_nr][0] = 1.0; |
| 324 | } |
| 325 | |
| 326 | pm_candidate1 = old_path_metrics[0] + input_symbol_real + increment[7]; |
| 327 | pm_candidate2 = old_path_metrics[8] + input_symbol_real - increment[0]; |
| 328 | if(pm_candidate1 > pm_candidate2){ |
| 329 | new_path_metrics[1] = pm_candidate1; |
| 330 | trans_table[sample_nr][1] = -1.0; |
| 331 | } |
| 332 | else{ |
| 333 | new_path_metrics[1] = pm_candidate2; |
| 334 | trans_table[sample_nr][1] = 1.0; |
| 335 | } |
| 336 | |
| 337 | pm_candidate1 = old_path_metrics[1] - input_symbol_real - increment[6]; |
| 338 | pm_candidate2 = old_path_metrics[9] - input_symbol_real + increment[1]; |
| 339 | if(pm_candidate1 > pm_candidate2){ |
| 340 | new_path_metrics[2] = pm_candidate1; |
| 341 | trans_table[sample_nr][2] = -1.0; |
| 342 | } |
| 343 | else{ |
| 344 | new_path_metrics[2] = pm_candidate2; |
| 345 | trans_table[sample_nr][2] = 1.0; |
| 346 | } |
| 347 | |
| 348 | pm_candidate1 = old_path_metrics[1] + input_symbol_real + increment[6]; |
| 349 | pm_candidate2 = old_path_metrics[9] + input_symbol_real - increment[1]; |
| 350 | if(pm_candidate1 > pm_candidate2){ |
| 351 | new_path_metrics[3] = pm_candidate1; |
| 352 | trans_table[sample_nr][3] = -1.0; |
| 353 | } |
| 354 | else{ |
| 355 | new_path_metrics[3] = pm_candidate2; |
| 356 | trans_table[sample_nr][3] = 1.0; |
| 357 | } |
| 358 | |
| 359 | pm_candidate1 = old_path_metrics[2] - input_symbol_real - increment[5]; |
| 360 | pm_candidate2 = old_path_metrics[10] - input_symbol_real + increment[2]; |
| 361 | if(pm_candidate1 > pm_candidate2){ |
| 362 | new_path_metrics[4] = pm_candidate1; |
| 363 | trans_table[sample_nr][4] = -1.0; |
| 364 | } |
| 365 | else{ |
| 366 | new_path_metrics[4] = pm_candidate2; |
| 367 | trans_table[sample_nr][4] = 1.0; |
| 368 | } |
| 369 | |
| 370 | pm_candidate1 = old_path_metrics[2] + input_symbol_real + increment[5]; |
| 371 | pm_candidate2 = old_path_metrics[10] + input_symbol_real - increment[2]; |
| 372 | if(pm_candidate1 > pm_candidate2){ |
| 373 | new_path_metrics[5] = pm_candidate1; |
| 374 | trans_table[sample_nr][5] = -1.0; |
| 375 | } |
| 376 | else{ |
| 377 | new_path_metrics[5] = pm_candidate2; |
| 378 | trans_table[sample_nr][5] = 1.0; |
| 379 | } |
| 380 | |
| 381 | pm_candidate1 = old_path_metrics[3] - input_symbol_real - increment[4]; |
| 382 | pm_candidate2 = old_path_metrics[11] - input_symbol_real + increment[3]; |
| 383 | if(pm_candidate1 > pm_candidate2){ |
| 384 | new_path_metrics[6] = pm_candidate1; |
| 385 | trans_table[sample_nr][6] = -1.0; |
| 386 | } |
| 387 | else{ |
| 388 | new_path_metrics[6] = pm_candidate2; |
| 389 | trans_table[sample_nr][6] = 1.0; |
| 390 | } |
| 391 | |
| 392 | pm_candidate1 = old_path_metrics[3] + input_symbol_real + increment[4]; |
| 393 | pm_candidate2 = old_path_metrics[11] + input_symbol_real - increment[3]; |
| 394 | if(pm_candidate1 > pm_candidate2){ |
| 395 | new_path_metrics[7] = pm_candidate1; |
| 396 | trans_table[sample_nr][7] = -1.0; |
| 397 | } |
| 398 | else{ |
| 399 | new_path_metrics[7] = pm_candidate2; |
| 400 | trans_table[sample_nr][7] = 1.0; |
| 401 | } |
| 402 | |
| 403 | pm_candidate1 = old_path_metrics[4] - input_symbol_real - increment[3]; |
| 404 | pm_candidate2 = old_path_metrics[12] - input_symbol_real + increment[4]; |
| 405 | if(pm_candidate1 > pm_candidate2){ |
| 406 | new_path_metrics[8] = pm_candidate1; |
| 407 | trans_table[sample_nr][8] = -1.0; |
| 408 | } |
| 409 | else{ |
| 410 | new_path_metrics[8] = pm_candidate2; |
| 411 | trans_table[sample_nr][8] = 1.0; |
| 412 | } |
| 413 | |
| 414 | pm_candidate1 = old_path_metrics[4] + input_symbol_real + increment[3]; |
| 415 | pm_candidate2 = old_path_metrics[12] + input_symbol_real - increment[4]; |
| 416 | if(pm_candidate1 > pm_candidate2){ |
| 417 | new_path_metrics[9] = pm_candidate1; |
| 418 | trans_table[sample_nr][9] = -1.0; |
| 419 | } |
| 420 | else{ |
| 421 | new_path_metrics[9] = pm_candidate2; |
| 422 | trans_table[sample_nr][9] = 1.0; |
| 423 | } |
| 424 | |
| 425 | pm_candidate1 = old_path_metrics[5] - input_symbol_real - increment[2]; |
| 426 | pm_candidate2 = old_path_metrics[13] - input_symbol_real + increment[5]; |
| 427 | if(pm_candidate1 > pm_candidate2){ |
| 428 | new_path_metrics[10] = pm_candidate1; |
| 429 | trans_table[sample_nr][10] = -1.0; |
| 430 | } |
| 431 | else{ |
| 432 | new_path_metrics[10] = pm_candidate2; |
| 433 | trans_table[sample_nr][10] = 1.0; |
| 434 | } |
| 435 | |
| 436 | pm_candidate1 = old_path_metrics[5] + input_symbol_real + increment[2]; |
| 437 | pm_candidate2 = old_path_metrics[13] + input_symbol_real - increment[5]; |
| 438 | if(pm_candidate1 > pm_candidate2){ |
| 439 | new_path_metrics[11] = pm_candidate1; |
| 440 | trans_table[sample_nr][11] = -1.0; |
| 441 | } |
| 442 | else{ |
| 443 | new_path_metrics[11] = pm_candidate2; |
| 444 | trans_table[sample_nr][11] = 1.0; |
| 445 | } |
| 446 | |
| 447 | pm_candidate1 = old_path_metrics[6] - input_symbol_real - increment[1]; |
| 448 | pm_candidate2 = old_path_metrics[14] - input_symbol_real + increment[6]; |
| 449 | if(pm_candidate1 > pm_candidate2){ |
| 450 | new_path_metrics[12] = pm_candidate1; |
| 451 | trans_table[sample_nr][12] = -1.0; |
| 452 | } |
| 453 | else{ |
| 454 | new_path_metrics[12] = pm_candidate2; |
| 455 | trans_table[sample_nr][12] = 1.0; |
| 456 | } |
| 457 | |
| 458 | pm_candidate1 = old_path_metrics[6] + input_symbol_real + increment[1]; |
| 459 | pm_candidate2 = old_path_metrics[14] + input_symbol_real - increment[6]; |
| 460 | if(pm_candidate1 > pm_candidate2){ |
| 461 | new_path_metrics[13] = pm_candidate1; |
| 462 | trans_table[sample_nr][13] = -1.0; |
| 463 | } |
| 464 | else{ |
| 465 | new_path_metrics[13] = pm_candidate2; |
| 466 | trans_table[sample_nr][13] = 1.0; |
| 467 | } |
| 468 | |
| 469 | pm_candidate1 = old_path_metrics[7] - input_symbol_real - increment[0]; |
| 470 | pm_candidate2 = old_path_metrics[15] - input_symbol_real + increment[7]; |
| 471 | if(pm_candidate1 > pm_candidate2){ |
| 472 | new_path_metrics[14] = pm_candidate1; |
| 473 | trans_table[sample_nr][14] = -1.0; |
| 474 | } |
| 475 | else{ |
| 476 | new_path_metrics[14] = pm_candidate2; |
| 477 | trans_table[sample_nr][14] = 1.0; |
| 478 | } |
| 479 | |
| 480 | pm_candidate1 = old_path_metrics[7] + input_symbol_real + increment[0]; |
| 481 | pm_candidate2 = old_path_metrics[15] + input_symbol_real - increment[7]; |
| 482 | if(pm_candidate1 > pm_candidate2){ |
| 483 | new_path_metrics[15] = pm_candidate1; |
| 484 | trans_table[sample_nr][15] = -1.0; |
| 485 | } |
| 486 | else{ |
| 487 | new_path_metrics[15] = pm_candidate2; |
| 488 | trans_table[sample_nr][15] = 1.0; |
| 489 | } |
| 490 | tmp=old_path_metrics; |
| 491 | old_path_metrics=new_path_metrics; |
| 492 | new_path_metrics=tmp; |
| 493 | |
| 494 | sample_nr++; |
| 495 | } |
| 496 | |
| 497 | /* |
| 498 | * Find the best from the stop states by comparing their path metrics. |
| 499 | * Not every stop state is always possible, so we are searching in |
| 500 | * a subset of them. |
| 501 | */ |
| 502 | unsigned int best_stop_state; |
| 503 | float stop_state_metric, max_stop_state_metric; |
| 504 | best_stop_state = stop_states[0]; |
| 505 | max_stop_state_metric = old_path_metrics[best_stop_state]; |
| 506 | for(i=1; i< stops_num; i++){ |
| 507 | stop_state_metric = old_path_metrics[stop_states[i]]; |
| 508 | if(stop_state_metric > max_stop_state_metric){ |
| 509 | max_stop_state_metric = stop_state_metric; |
| 510 | best_stop_state = stop_states[i]; |
| 511 | } |
| 512 | } |
| 513 | |
| 514 | /* |
| 515 | * This table was generated with hope that it gives a litle speedup during |
| 516 | * traceback stage. |
| 517 | * Received bit is related to the number of state in the trellis. |
| 518 | * I've numbered states so their parity (number of ones) is related |
| 519 | * to a received bit. |
| 520 | */ |
| 521 | static const unsigned int parity_table[PATHS_NUM] = { 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 0, }; |
| 522 | |
| 523 | /* |
| 524 | * Table of previous states in the trellis diagram. |
| 525 | * For GMSK modulation every state has two previous states. |
| 526 | * Example: |
| 527 | * previous_state_nr1 = prev_table[current_state_nr][0] |
| 528 | * previous_state_nr2 = prev_table[current_state_nr][1] |
| 529 | */ |
| 530 | static const unsigned int prev_table[PATHS_NUM][2] = { {0,8}, {0,8}, {1,9}, {1,9}, {2,10}, {2,10}, {3,11}, {3,11}, {4,12}, {4,12}, {5,13}, {5,13}, {6,14}, {6,14}, {7,15}, {7,15}, }; |
| 531 | |
| 532 | /* |
| 533 | * Traceback and differential decoding of received sequence. |
| 534 | * Decisions stored in trans_table are used to restore best path in the trellis. |
| 535 | */ |
| 536 | sample_nr=samples_num; |
| 537 | unsigned int state_nr=best_stop_state; |
| 538 | unsigned int decision; |
| 539 | bool out_bit=0; |
| 540 | |
| 541 | while(sample_nr>0){ |
| 542 | sample_nr--; |
| 543 | decision = (trans_table[sample_nr][state_nr]>0); |
| 544 | |
| 545 | if(decision != out_bit) |
| 546 | output[sample_nr]=-trans_table[sample_nr][state_nr]; |
| 547 | else |
| 548 | output[sample_nr]=trans_table[sample_nr][state_nr]; |
| 549 | |
| 550 | out_bit = out_bit ^ real_imag ^ parity_table[state_nr]; |
| 551 | state_nr = prev_table[state_nr][decision]; |
| 552 | real_imag = !real_imag; |
| 553 | } |
| 554 | } |