1 /* 2 ** Copyright 2011, The Android Open-Source Project 3 ** 4 ** Licensed under the Apache License, Version 2.0 (the "License"); 5 ** you may not use this file except in compliance with the License. 6 ** You may obtain a copy of the License at 7 ** 8 ** http://www.apache.org/licenses/LICENSE-2.0 9 ** 10 ** Unless required by applicable law or agreed to in writing, software 11 ** distributed under the License is distributed on an "AS IS" BASIS, 12 ** WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. 13 ** See the License for the specific language governing permissions and 14 ** limitations under the License. 15 */ 16 17 //#define LOG_NDEBUG 0 18 #define LOG_TAG "echo_reference" 19 20 #include <errno.h> 21 #include <stdlib.h> 22 #include <pthread.h> 23 #include <cutils/log.h> 24 #include <system/audio.h> 25 #include <audio_utils/resampler.h> 26 #include <audio_utils/echo_reference.h> 27 28 // echo reference state: bit field indicating if read, write or both are active. 29 enum state { 30 ECHOREF_IDLE = 0x00, // idle 31 ECHOREF_READING = 0x01, // reading is active 32 ECHOREF_WRITING = 0x02 // writing is active 33 }; 34 35 struct echo_reference { 36 struct echo_reference_itfe itfe; 37 int status; // init status 38 uint32_t state; // active state: reading, writing or both 39 audio_format_t rd_format; // read sample format 40 uint32_t rd_channel_count; // read number of channels 41 uint32_t rd_sampling_rate; // read sampling rate in Hz 42 size_t rd_frame_size; // read frame size (bytes per sample) 43 audio_format_t wr_format; // write sample format 44 uint32_t wr_channel_count; // write number of channels 45 uint32_t wr_sampling_rate; // write sampling rate in Hz 46 size_t wr_frame_size; // write frame size (bytes per sample) 47 void *buffer; // main buffer 48 size_t buf_size; // main buffer size in frames 49 size_t frames_in; // number of frames in main buffer 50 void *wr_buf; // buffer for input conversions 51 size_t wr_buf_size; // size of conversion buffer in frames 52 size_t wr_frames_in; // number of frames in conversion buffer 53 void *wr_src_buf; // resampler input buf (either wr_buf or buffer used by write()) 54 struct timespec wr_render_time; // latest render time indicated by write() 55 // default ALSA gettimeofday() format 56 int32_t playback_delay; // playback buffer delay indicated by last write() 57 pthread_mutex_t lock; // mutex protecting read/write concurrency 58 pthread_cond_t cond; // condition signaled when data is ready to read 59 struct resampler_itfe *down_sampler; // input resampler 60 struct resampler_buffer_provider provider; // resampler buffer provider 61 }; 62 63 64 int echo_reference_get_next_buffer(struct resampler_buffer_provider *buffer_provider, 65 struct resampler_buffer* buffer) 66 { 67 struct echo_reference *er; 68 69 if (buffer_provider == NULL) { 70 return -EINVAL; 71 } 72 73 er = (struct echo_reference *)((char *)buffer_provider - 74 offsetof(struct echo_reference, provider)); 75 76 if (er->wr_src_buf == NULL || er->wr_frames_in == 0) { 77 buffer->raw = NULL; 78 buffer->frame_count = 0; 79 return -ENODATA; 80 } 81 82 buffer->frame_count = (buffer->frame_count > er->wr_frames_in) ? er->wr_frames_in : buffer->frame_count; 83 // this is er->rd_channel_count here as we resample after stereo to mono conversion if any 84 buffer->i16 = (int16_t *)er->wr_src_buf + (er->wr_buf_size - er->wr_frames_in) * er->rd_channel_count; 85 86 return 0; 87 } 88 89 void echo_reference_release_buffer(struct resampler_buffer_provider *buffer_provider, 90 struct resampler_buffer* buffer) 91 { 92 struct echo_reference *er; 93 94 if (buffer_provider == NULL) { 95 return; 96 } 97 98 er = (struct echo_reference *)((char *)buffer_provider - 99 offsetof(struct echo_reference, provider)); 100 101 er->wr_frames_in -= buffer->frame_count; 102 } 103 104 static void echo_reference_reset_l(struct echo_reference *er) 105 { 106 LOGV("echo_reference_reset_l()"); 107 free(er->buffer); 108 er->buffer = NULL; 109 er->buf_size = 0; 110 er->frames_in = 0; 111 free(er->wr_buf); 112 er->wr_buf = NULL; 113 er->wr_buf_size = 0; 114 er->wr_render_time.tv_sec = 0; 115 er->wr_render_time.tv_nsec = 0; 116 } 117 118 static int echo_reference_write(struct echo_reference_itfe *echo_reference, 119 struct echo_reference_buffer *buffer) 120 { 121 struct echo_reference *er = (struct echo_reference *)echo_reference; 122 int status = 0; 123 124 if (er == NULL) { 125 return -EINVAL; 126 } 127 128 pthread_mutex_lock(&er->lock); 129 130 if (buffer == NULL) { 131 LOGV("echo_reference_write() stop write"); 132 er->state &= ~ECHOREF_WRITING; 133 echo_reference_reset_l(er); 134 goto exit; 135 } 136 137 LOGV("echo_reference_write() START trying to write %d frames", buffer->frame_count); 138 LOGV("echo_reference_write() playbackTimestamp:[%d].[%d], er->playback_delay:[%d]", 139 (int)buffer->time_stamp.tv_sec, 140 (int)buffer->time_stamp.tv_nsec, er->playback_delay); 141 142 //LOGV("echo_reference_write() %d frames", buffer->frame_count); 143 // discard writes until a valid time stamp is provided. 144 145 if ((buffer->time_stamp.tv_sec == 0) && (buffer->time_stamp.tv_nsec == 0) && 146 (er->wr_render_time.tv_sec == 0) && (er->wr_render_time.tv_nsec == 0)) { 147 goto exit; 148 } 149 150 if ((er->state & ECHOREF_WRITING) == 0) { 151 LOGV("echo_reference_write() start write"); 152 if (er->down_sampler != NULL) { 153 er->down_sampler->reset(er->down_sampler); 154 } 155 er->state |= ECHOREF_WRITING; 156 } 157 158 if ((er->state & ECHOREF_READING) == 0) { 159 goto exit; 160 } 161 162 er->wr_render_time.tv_sec = buffer->time_stamp.tv_sec; 163 er->wr_render_time.tv_nsec = buffer->time_stamp.tv_nsec; 164 165 er->playback_delay = buffer->delay_ns; 166 167 void *srcBuf; 168 size_t inFrames; 169 // do stereo to mono and down sampling if necessary 170 if (er->rd_channel_count != er->wr_channel_count || 171 er->rd_sampling_rate != er->wr_sampling_rate) { 172 if (er->wr_buf_size < buffer->frame_count) { 173 er->wr_buf_size = buffer->frame_count; 174 //max buffer size is normally function of read sampling rate but as write sampling rate 175 //is always more than read sampling rate this works 176 er->wr_buf = realloc(er->wr_buf, er->wr_buf_size * er->rd_frame_size); 177 } 178 179 inFrames = buffer->frame_count; 180 if (er->rd_channel_count != er->wr_channel_count) { 181 // must be stereo to mono 182 int16_t *src16 = (int16_t *)buffer->raw; 183 int16_t *dst16 = (int16_t *)er->wr_buf; 184 size_t frames = buffer->frame_count; 185 while (frames--) { 186 *dst16++ = (int16_t)(((int32_t)*src16 + (int32_t)*(src16 + 1)) >> 1); 187 src16 += 2; 188 } 189 } 190 if (er->wr_sampling_rate != er->rd_sampling_rate) { 191 if (er->down_sampler == NULL) { 192 int rc; 193 LOGV("echo_reference_write() new ReSampler(%d, %d)", 194 er->wr_sampling_rate, er->rd_sampling_rate); 195 er->provider.get_next_buffer = echo_reference_get_next_buffer; 196 er->provider.release_buffer = echo_reference_release_buffer; 197 rc = create_resampler(er->wr_sampling_rate, 198 er->rd_sampling_rate, 199 er->rd_channel_count, 200 RESAMPLER_QUALITY_VOIP, 201 &er->provider, 202 &er->down_sampler); 203 if (rc != 0) { 204 er->down_sampler = NULL; 205 LOGV("echo_reference_write() failure to create resampler %d", rc); 206 status = -ENODEV; 207 goto exit; 208 } 209 } 210 // er->wr_src_buf and er->wr_frames_in are used by getNexBuffer() called by the resampler 211 // to get new frames 212 if (er->rd_channel_count != er->wr_channel_count) { 213 er->wr_src_buf = er->wr_buf; 214 } else { 215 er->wr_src_buf = buffer->raw; 216 } 217 er->wr_frames_in = buffer->frame_count; 218 // inFrames is always more than we need here to get frames remaining from previous runs 219 // inFrames is updated by resample() with the number of frames produced 220 LOGV("echo_reference_write() ReSampling(%d, %d)", 221 er->wr_sampling_rate, er->rd_sampling_rate); 222 er->down_sampler->resample_from_provider(er->down_sampler, 223 (int16_t *)er->wr_buf, &inFrames); 224 LOGV_IF(er->wr_frames_in != 0, 225 "echo_reference_write() er->wr_frames_in not 0 (%d) after resampler", 226 er->wr_frames_in); 227 } 228 srcBuf = er->wr_buf; 229 } else { 230 inFrames = buffer->frame_count; 231 srcBuf = buffer->raw; 232 } 233 234 if (er->frames_in + inFrames > er->buf_size) { 235 LOGV("echo_reference_write() increasing buffer size from %d to %d", 236 er->buf_size, er->frames_in + inFrames); 237 er->buf_size = er->frames_in + inFrames; 238 er->buffer = realloc(er->buffer, er->buf_size * er->rd_frame_size); 239 } 240 memcpy((char *)er->buffer + er->frames_in * er->rd_frame_size, 241 srcBuf, 242 inFrames * er->rd_frame_size); 243 er->frames_in += inFrames; 244 245 LOGV("EchoReference::write_log() inFrames:[%d], mFramesInOld:[%d], "\ 246 "mFramesInNew:[%d], er->buf_size:[%d], er->wr_render_time:[%d].[%d]," 247 "er->playback_delay:[%d]", 248 inFrames, er->frames_in - inFrames, er->frames_in, er->buf_size, 249 (int)er->wr_render_time.tv_sec, 250 (int)er->wr_render_time.tv_nsec, er->playback_delay); 251 252 pthread_cond_signal(&er->cond); 253 exit: 254 pthread_mutex_unlock(&er->lock); 255 LOGV("echo_reference_write() END"); 256 return status; 257 } 258 259 #define MIN_DELAY_UPDATE_NS 62500 // delay jump threshold to update ref buffer 260 // 0.5 samples at 8kHz in nsecs 261 262 263 static int echo_reference_read(struct echo_reference_itfe *echo_reference, 264 struct echo_reference_buffer *buffer) 265 { 266 struct echo_reference *er = (struct echo_reference *)echo_reference; 267 268 if (er == NULL) { 269 return -EINVAL; 270 } 271 272 pthread_mutex_lock(&er->lock); 273 274 if (buffer == NULL) { 275 LOGV("EchoReference::read() stop read"); 276 er->state &= ~ECHOREF_READING; 277 goto exit; 278 } 279 280 LOGV("EchoReference::read() START, delayCapture:[%d],er->frames_in:[%d],buffer->frame_count:[%d]", 281 buffer->delay_ns, er->frames_in, buffer->frame_count); 282 283 if ((er->state & ECHOREF_READING) == 0) { 284 LOGV("EchoReference::read() start read"); 285 echo_reference_reset_l(er); 286 er->state |= ECHOREF_READING; 287 } 288 289 if ((er->state & ECHOREF_WRITING) == 0) { 290 memset(buffer->raw, 0, er->rd_frame_size * buffer->frame_count); 291 buffer->delay_ns = 0; 292 goto exit; 293 } 294 295 // LOGV("EchoReference::read() %d frames", buffer->frame_count); 296 297 // allow some time for new frames to arrive if not enough frames are ready for read 298 if (er->frames_in < buffer->frame_count) { 299 uint32_t timeoutMs = (uint32_t)((1000 * buffer->frame_count) / er->rd_sampling_rate / 2); 300 struct timespec ts; 301 302 ts.tv_sec = timeoutMs/1000; 303 ts.tv_nsec = timeoutMs%1000; 304 pthread_cond_timedwait_relative_np(&er->cond, &er->lock, &ts); 305 306 if (er->frames_in < buffer->frame_count) { 307 LOGV("EchoReference::read() waited %d ms but still not enough frames"\ 308 " er->frames_in: %d, buffer->frame_count = %d", 309 timeoutMs, er->frames_in, buffer->frame_count); 310 buffer->frame_count = er->frames_in; 311 } 312 } 313 314 int64_t timeDiff; 315 struct timespec tmp; 316 317 if ((er->wr_render_time.tv_sec == 0 && er->wr_render_time.tv_nsec == 0) || 318 (buffer->time_stamp.tv_sec == 0 && buffer->time_stamp.tv_nsec == 0)) { 319 LOGV("read: NEW:timestamp is zero---------setting timeDiff = 0, "\ 320 "not updating delay this time"); 321 timeDiff = 0; 322 } else { 323 if (buffer->time_stamp.tv_nsec < er->wr_render_time.tv_nsec) { 324 tmp.tv_sec = buffer->time_stamp.tv_sec - er->wr_render_time.tv_sec - 1; 325 tmp.tv_nsec = 1000000000 + buffer->time_stamp.tv_nsec - er->wr_render_time.tv_nsec; 326 } else { 327 tmp.tv_sec = buffer->time_stamp.tv_sec - er->wr_render_time.tv_sec; 328 tmp.tv_nsec = buffer->time_stamp.tv_nsec - er->wr_render_time.tv_nsec; 329 } 330 timeDiff = (((int64_t)tmp.tv_sec * 1000000000 + tmp.tv_nsec)); 331 332 int64_t expectedDelayNs = er->playback_delay + buffer->delay_ns - timeDiff; 333 334 LOGV("expectedDelayNs[%lld] = er->playback_delay[%d] + delayCapture[%d] - timeDiff[%lld]", 335 expectedDelayNs, er->playback_delay, buffer->delay_ns, timeDiff); 336 337 if (expectedDelayNs > 0) { 338 int64_t delayNs = ((int64_t)er->frames_in * 1000000000) / er->rd_sampling_rate; 339 340 delayNs -= expectedDelayNs; 341 342 if (abs(delayNs) >= MIN_DELAY_UPDATE_NS) { 343 if (delayNs < 0) { 344 size_t previousFrameIn = er->frames_in; 345 er->frames_in = (expectedDelayNs * er->rd_sampling_rate)/1000000000; 346 int offset = er->frames_in - previousFrameIn; 347 LOGV("EchoReference::readlog: delayNs = NEGATIVE and ENOUGH : "\ 348 "setting %d frames to zero er->frames_in: %d, previousFrameIn = %d", 349 offset, er->frames_in, previousFrameIn); 350 351 if (er->frames_in > er->buf_size) { 352 er->buf_size = er->frames_in; 353 er->buffer = realloc(er->buffer, er->frames_in * er->rd_frame_size); 354 LOGV("EchoReference::read: increasing buffer size to %d", er->buf_size); 355 } 356 357 if (offset > 0) 358 memset((char *)er->buffer + previousFrameIn * er->rd_frame_size, 359 0, offset * er->rd_frame_size); 360 } else { 361 size_t previousFrameIn = er->frames_in; 362 int framesInInt = (int)(((int64_t)expectedDelayNs * 363 (int64_t)er->rd_sampling_rate)/1000000000); 364 int offset = previousFrameIn - framesInInt; 365 366 LOGV("EchoReference::readlog: delayNs = POSITIVE/ENOUGH :previousFrameIn: %d,"\ 367 "framesInInt: [%d], offset:[%d], buffer->frame_count:[%d]", 368 previousFrameIn, framesInInt, offset, buffer->frame_count); 369 370 if (framesInInt < (int)buffer->frame_count) { 371 if (framesInInt > 0) { 372 memset((char *)er->buffer + framesInInt * er->rd_frame_size, 373 0, (buffer->frame_count-framesInInt) * er->rd_frame_size); 374 LOGV("EchoReference::read: pushing [%d] zeros into ref buffer", 375 (buffer->frame_count-framesInInt)); 376 } else { 377 LOGV("framesInInt = %d", framesInInt); 378 } 379 framesInInt = buffer->frame_count; 380 } else { 381 if (offset > 0) { 382 memcpy(er->buffer, (char *)er->buffer + (offset * er->rd_frame_size), 383 framesInInt * er->rd_frame_size); 384 LOGV("EchoReference::read: shifting ref buffer by [%d]",framesInInt); 385 } 386 } 387 er->frames_in = (size_t)framesInInt; 388 } 389 } else { 390 LOGV("EchoReference::read: NOT ENOUGH samples to update %lld", delayNs); 391 } 392 } else { 393 LOGV("NEGATIVE expectedDelayNs[%lld] = "\ 394 "er->playback_delay[%d] + delayCapture[%d] - timeDiff[%lld]", 395 expectedDelayNs, er->playback_delay, buffer->delay_ns, timeDiff); 396 } 397 } 398 399 memcpy(buffer->raw, 400 (char *)er->buffer, 401 buffer->frame_count * er->rd_frame_size); 402 403 er->frames_in -= buffer->frame_count; 404 memcpy(er->buffer, 405 (char *)er->buffer + buffer->frame_count * er->rd_frame_size, 406 er->frames_in * er->rd_frame_size); 407 408 // As the reference buffer is now time aligned to the microphone signal there is a zero delay 409 buffer->delay_ns = 0; 410 411 LOGV("EchoReference::read() END %d frames, total frames in %d", 412 buffer->frame_count, er->frames_in); 413 414 pthread_cond_signal(&er->cond); 415 416 exit: 417 pthread_mutex_unlock(&er->lock); 418 return 0; 419 } 420 421 422 int create_echo_reference(audio_format_t rdFormat, 423 uint32_t rdChannelCount, 424 uint32_t rdSamplingRate, 425 audio_format_t wrFormat, 426 uint32_t wrChannelCount, 427 uint32_t wrSamplingRate, 428 struct echo_reference_itfe **echo_reference) 429 { 430 struct echo_reference *er; 431 432 LOGV("create_echo_reference()"); 433 434 if (echo_reference == NULL) { 435 return -EINVAL; 436 } 437 438 *echo_reference = NULL; 439 440 if (rdFormat != AUDIO_FORMAT_PCM_16_BIT || 441 rdFormat != wrFormat) { 442 LOGW("create_echo_reference bad format rd %d, wr %d", rdFormat, wrFormat); 443 return -EINVAL; 444 } 445 if ((rdChannelCount != 1 && rdChannelCount != 2) || 446 wrChannelCount != 2) { 447 LOGW("create_echo_reference bad channel count rd %d, wr %d", rdChannelCount, wrChannelCount); 448 return -EINVAL; 449 } 450 451 if (wrSamplingRate < rdSamplingRate) { 452 LOGW("create_echo_reference bad smp rate rd %d, wr %d", rdSamplingRate, wrSamplingRate); 453 return -EINVAL; 454 } 455 456 er = (struct echo_reference *)calloc(1, sizeof(struct echo_reference)); 457 458 er->itfe.read = echo_reference_read; 459 er->itfe.write = echo_reference_write; 460 461 er->state = ECHOREF_IDLE; 462 er->rd_format = rdFormat; 463 er->rd_channel_count = rdChannelCount; 464 er->rd_sampling_rate = rdSamplingRate; 465 er->wr_format = wrFormat; 466 er->wr_channel_count = wrChannelCount; 467 er->wr_sampling_rate = wrSamplingRate; 468 er->rd_frame_size = audio_bytes_per_sample(rdFormat) * rdChannelCount; 469 er->wr_frame_size = audio_bytes_per_sample(wrFormat) * wrChannelCount; 470 *echo_reference = &er->itfe; 471 return 0; 472 } 473 474 void release_echo_reference(struct echo_reference_itfe *echo_reference) { 475 struct echo_reference *er = (struct echo_reference *)echo_reference; 476 477 if (er == NULL) { 478 return; 479 } 480 481 LOGV("EchoReference dstor"); 482 echo_reference_reset_l(er); 483 if (er->down_sampler != NULL) { 484 release_resampler(er->down_sampler); 485 } 486 free(er); 487 } 488 489
