1 /* GStreamer DCA parser
2 * Copyright (C) 2010 Tim-Philipp Müller <tim centricular net>
4 * This library is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Library General Public
6 * License as published by the Free Software Foundation; either
7 * version 2 of the License, or (at your option) any later version.
9 * This library is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Library General Public License for more details.
14 * You should have received a copy of the GNU Library General Public
15 * License along with this library; if not, write to the
16 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
17 * Boston, MA 02110-1301, USA.
21 * SECTION:element-dcaparse
22 * @short_description: DCA (DTS Coherent Acoustics) parser
23 * @see_also: #GstAmrParse, #GstAACParse, #GstAc3Parse
25 * This is a DCA (DTS Coherent Acoustics) parser.
28 * <title>Example launch line</title>
30 * gst-launch-1.0 filesrc location=abc.dts ! dcaparse ! dtsdec ! audioresample ! audioconvert ! autoaudiosink
36 * - should accept framed and unframed input (needs decodebin fixes first)
37 * - seeking in raw .dts files doesn't seem to work, but duration estimate ok
39 * - if frames have 'odd' durations, the frame durations (plus timestamps)
40 * aren't adjusted up occasionally to make up for rounding error gaps.
41 * (e.g. if 512 samples per frame @ 48kHz = 10.666666667 ms/frame)
50 #include "gstdcaparse.h"
51 #include <gst/base/base.h>
52 #include <gst/pbutils/pbutils.h>
54 GST_DEBUG_CATEGORY_STATIC (dca_parse_debug);
55 #define GST_CAT_DEFAULT dca_parse_debug
57 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
60 GST_STATIC_CAPS ("audio/x-dts,"
61 " framed = (boolean) true,"
62 " channels = (int) [ 1, 8 ],"
63 " rate = (int) [ 8000, 192000 ],"
64 " depth = (int) { 14, 16 },"
65 " endianness = (int) { LITTLE_ENDIAN, BIG_ENDIAN }, "
66 " block-size = (int) [ 1, MAX], " " frame-size = (int) [ 1, MAX]"));
68 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
71 GST_STATIC_CAPS ("audio/x-dts; " "audio/x-private1-dts"));
73 static void gst_dca_parse_finalize (GObject * object);
75 static gboolean gst_dca_parse_start (GstBaseParse * parse);
76 static gboolean gst_dca_parse_stop (GstBaseParse * parse);
77 static GstFlowReturn gst_dca_parse_handle_frame (GstBaseParse * parse,
78 GstBaseParseFrame * frame, gint * skipsize);
79 static GstFlowReturn gst_dca_parse_pre_push_frame (GstBaseParse * parse,
80 GstBaseParseFrame * frame);
81 static GstCaps *gst_dca_parse_get_sink_caps (GstBaseParse * parse,
83 static gboolean gst_dca_parse_set_sink_caps (GstBaseParse * parse,
86 #define gst_dca_parse_parent_class parent_class
87 G_DEFINE_TYPE (GstDcaParse, gst_dca_parse, GST_TYPE_BASE_PARSE);
90 gst_dca_parse_class_init (GstDcaParseClass * klass)
92 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
93 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
94 GObjectClass *object_class = G_OBJECT_CLASS (klass);
96 GST_DEBUG_CATEGORY_INIT (dca_parse_debug, "dcaparse", 0,
97 "DCA audio stream parser");
99 object_class->finalize = gst_dca_parse_finalize;
101 parse_class->start = GST_DEBUG_FUNCPTR (gst_dca_parse_start);
102 parse_class->stop = GST_DEBUG_FUNCPTR (gst_dca_parse_stop);
103 parse_class->handle_frame = GST_DEBUG_FUNCPTR (gst_dca_parse_handle_frame);
104 parse_class->pre_push_frame =
105 GST_DEBUG_FUNCPTR (gst_dca_parse_pre_push_frame);
106 parse_class->get_sink_caps = GST_DEBUG_FUNCPTR (gst_dca_parse_get_sink_caps);
107 parse_class->set_sink_caps = GST_DEBUG_FUNCPTR (gst_dca_parse_set_sink_caps);
109 gst_element_class_add_pad_template (element_class,
110 gst_static_pad_template_get (&sink_template));
111 gst_element_class_add_pad_template (element_class,
112 gst_static_pad_template_get (&src_template));
114 gst_element_class_set_static_metadata (element_class,
115 "DTS Coherent Acoustics audio stream parser", "Codec/Parser/Audio",
116 "DCA parser", "Tim-Philipp Müller <tim centricular net>");
120 gst_dca_parse_reset (GstDcaParse * dcaparse)
122 dcaparse->channels = -1;
124 dcaparse->depth = -1;
125 dcaparse->endianness = -1;
126 dcaparse->block_size = -1;
127 dcaparse->frame_size = -1;
128 dcaparse->last_sync = 0;
129 dcaparse->sent_codec_tag = FALSE;
133 gst_dca_parse_init (GstDcaParse * dcaparse)
135 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (dcaparse),
137 gst_dca_parse_reset (dcaparse);
138 dcaparse->baseparse_chainfunc =
139 GST_BASE_PARSE_SINK_PAD (GST_BASE_PARSE (dcaparse))->chainfunc;
141 GST_PAD_SET_ACCEPT_INTERSECT (GST_BASE_PARSE_SINK_PAD (dcaparse));
142 GST_PAD_SET_ACCEPT_TEMPLATE (GST_BASE_PARSE_SINK_PAD (dcaparse));
146 gst_dca_parse_finalize (GObject * object)
148 G_OBJECT_CLASS (parent_class)->finalize (object);
152 gst_dca_parse_start (GstBaseParse * parse)
154 GstDcaParse *dcaparse = GST_DCA_PARSE (parse);
156 GST_DEBUG_OBJECT (parse, "starting");
158 gst_dca_parse_reset (dcaparse);
164 gst_dca_parse_stop (GstBaseParse * parse)
166 GST_DEBUG_OBJECT (parse, "stopping");
172 gst_dca_parse_parse_header (GstDcaParse * dcaparse,
173 const GstByteReader * reader, guint * frame_size,
174 guint * sample_rate, guint * channels, guint * depth,
175 gint * endianness, guint * num_blocks, guint * samples_per_block,
176 gboolean * terminator)
178 static const int sample_rates[16] = { 0, 8000, 16000, 32000, 0, 0, 11025,
179 22050, 44100, 0, 0, 12000, 24000, 48000, 96000, 192000
181 static const guint8 channels_table[16] = { 1, 2, 2, 2, 2, 3, 3, 4, 4, 5,
184 GstByteReader r = *reader;
189 if (gst_byte_reader_get_remaining (&r) < (4 + sizeof (hdr)))
192 marker = gst_byte_reader_peek_uint32_be_unchecked (&r);
194 /* raw big endian or 14-bit big endian */
195 if (marker == 0x7FFE8001 || marker == 0x1FFFE800) {
196 for (i = 0; i < G_N_ELEMENTS (hdr); ++i)
197 hdr[i] = gst_byte_reader_get_uint16_be_unchecked (&r);
199 /* raw little endian or 14-bit little endian */
200 if (marker == 0xFE7F0180 || marker == 0xFF1F00E8) {
201 for (i = 0; i < G_N_ELEMENTS (hdr); ++i)
202 hdr[i] = gst_byte_reader_get_uint16_le_unchecked (&r);
207 GST_LOG_OBJECT (dcaparse, "dts sync marker 0x%08x at offset %u", marker,
208 gst_byte_reader_get_pos (reader));
211 if (marker == 0x1FFFE800 || marker == 0xFF1F00E8) {
212 if ((hdr[2] & 0xFFF0) != 0x07F0)
214 /* discard top 2 bits (2 void), shift in 2 */
215 hdr[0] = (hdr[0] << 2) | ((hdr[1] >> 12) & 0x0003);
216 /* discard top 4 bits (2 void, 2 shifted into hdr[0]), shift in 4 etc. */
217 hdr[1] = (hdr[1] << 4) | ((hdr[2] >> 10) & 0x000F);
218 hdr[2] = (hdr[2] << 6) | ((hdr[3] >> 8) & 0x003F);
219 hdr[3] = (hdr[3] << 8) | ((hdr[4] >> 6) & 0x00FF);
220 hdr[4] = (hdr[4] << 10) | ((hdr[5] >> 4) & 0x03FF);
221 hdr[5] = (hdr[5] << 12) | ((hdr[6] >> 2) & 0x0FFF);
222 hdr[6] = (hdr[6] << 14) | ((hdr[7] >> 0) & 0x3FFF);
223 g_assert (hdr[0] == 0x7FFE && hdr[1] == 0x8001);
226 GST_LOG_OBJECT (dcaparse, "frame header: %04x%04x%04x%04x",
227 hdr[2], hdr[3], hdr[4], hdr[5]);
229 *terminator = (hdr[2] & 0x80) ? FALSE : TRUE;
230 *samples_per_block = ((hdr[2] >> 10) & 0x1f) + 1;
231 *num_blocks = ((hdr[2] >> 2) & 0x7F) + 1;
232 *frame_size = (((hdr[2] & 0x03) << 12) | (hdr[3] >> 4)) + 1;
233 chans = ((hdr[3] & 0x0F) << 2) | (hdr[4] >> 14);
234 *sample_rate = sample_rates[(hdr[4] >> 10) & 0x0F];
235 lfe = (hdr[5] >> 9) & 0x03;
237 GST_TRACE_OBJECT (dcaparse, "frame size %u, num_blocks %u, rate %u, "
238 "samples per block %u", *frame_size, *num_blocks, *sample_rate,
241 if (*num_blocks < 6 || *frame_size < 96 || *sample_rate == 0)
244 if (marker == 0x1FFFE800 || marker == 0xFF1F00E8)
245 *frame_size = (*frame_size * 16) / 14; /* FIXME: round up? */
247 if (chans < G_N_ELEMENTS (channels_table))
248 *channels = channels_table[chans] + ((lfe) ? 1 : 0);
253 *depth = (marker == 0x1FFFE800 || marker == 0xFF1F00E8) ? 14 : 16;
255 *endianness = (marker == 0xFE7F0180 || marker == 0xFF1F00E8) ?
256 G_LITTLE_ENDIAN : G_BIG_ENDIAN;
258 GST_TRACE_OBJECT (dcaparse, "frame size %u, channels %u, rate %u, "
259 "num_blocks %u, samples_per_block %u", *frame_size, *channels,
260 *sample_rate, *num_blocks, *samples_per_block);
266 gst_dca_parse_find_sync (GstDcaParse * dcaparse, GstByteReader * reader,
267 gsize bufsize, guint32 * sync)
269 guint32 best_sync = 0;
270 guint best_offset = G_MAXUINT;
273 /* FIXME: verify syncs via _parse_header() here already */
275 /* Raw little endian */
276 off = gst_byte_reader_masked_scan_uint32 (reader, 0xffffffff, 0xfe7f0180,
278 if (off >= 0 && off < best_offset) {
280 best_sync = 0xfe7f0180;
284 off = gst_byte_reader_masked_scan_uint32 (reader, 0xffffffff, 0x7ffe8001,
286 if (off >= 0 && off < best_offset) {
288 best_sync = 0x7ffe8001;
291 /* FIXME: check next 2 bytes as well for 14-bit formats (but then don't
292 * forget to adjust the *skipsize= in _check_valid_frame() */
294 /* 14-bit little endian */
295 off = gst_byte_reader_masked_scan_uint32 (reader, 0xffffffff, 0xff1f00e8,
297 if (off >= 0 && off < best_offset) {
299 best_sync = 0xff1f00e8;
302 /* 14-bit big endian */
303 off = gst_byte_reader_masked_scan_uint32 (reader, 0xffffffff, 0x1fffe800,
305 if (off >= 0 && off < best_offset) {
307 best_sync = 0x1fffe800;
310 if (best_offset == G_MAXUINT)
318 gst_dca_parse_handle_frame (GstBaseParse * parse,
319 GstBaseParseFrame * frame, gint * skipsize)
321 GstDcaParse *dcaparse = GST_DCA_PARSE (parse);
322 GstBuffer *buf = frame->buffer;
324 gboolean parser_in_sync;
327 guint size = 0, rate, chans, num_blocks, samples_per_block, depth;
332 GstFlowReturn ret = GST_FLOW_EOS;
334 gst_buffer_map (buf, &map, GST_MAP_READ);
336 if (G_UNLIKELY (map.size < 16)) {
341 parser_in_sync = !GST_BASE_PARSE_LOST_SYNC (parse);
343 gst_byte_reader_init (&r, map.data, map.size);
345 if (G_LIKELY (parser_in_sync && dcaparse->last_sync != 0)) {
346 off = gst_byte_reader_masked_scan_uint32 (&r, 0xffffffff,
347 dcaparse->last_sync, 0, map.size);
350 if (G_UNLIKELY (off < 0)) {
351 off = gst_dca_parse_find_sync (dcaparse, &r, map.size, &sync);
354 /* didn't find anything that looks like a sync word, skip */
356 *skipsize = map.size - 3;
357 GST_DEBUG_OBJECT (dcaparse, "no sync, skipping %d bytes", *skipsize);
361 GST_LOG_OBJECT (parse, "possible sync %08x at buffer offset %d", sync, off);
363 /* possible frame header, but not at offset 0? skip bytes before sync */
369 /* make sure the values in the frame header look sane */
370 if (!gst_dca_parse_parse_header (dcaparse, &r, &size, &rate, &chans, &depth,
371 &endianness, &num_blocks, &samples_per_block, &terminator)) {
376 GST_LOG_OBJECT (parse, "got frame, sync %08x, size %u, rate %d, channels %d",
377 sync, size, rate, chans);
379 dcaparse->last_sync = sync;
381 /* FIXME: Don't look for a second syncword, there are streams out there
382 * that consistently contain garbage between every frame so we never ever
383 * find a second consecutive syncword.
384 * See https://bugzilla.gnome.org/show_bug.cgi?id=738237
387 parser_draining = GST_BASE_PARSE_DRAINING (parse);
389 if (!parser_in_sync && !parser_draining) {
390 /* check for second frame to be sure */
391 GST_DEBUG_OBJECT (dcaparse, "resyncing; checking next frame syncword");
392 if (map.size >= (size + 16)) {
393 guint s2, r2, c2, n2, s3;
396 GST_MEMDUMP ("buf", map.data, size + 16);
397 gst_byte_reader_init (&r, map.data, map.size);
398 gst_byte_reader_skip_unchecked (&r, size);
400 if (!gst_dca_parse_parse_header (dcaparse, &r, &s2, &r2, &c2, NULL, NULL,
402 GST_DEBUG_OBJECT (dcaparse, "didn't find second syncword");
407 /* ok, got sync now, let's assume constant frame size */
408 gst_base_parse_set_min_frame_size (parse, size);
410 /* wait for some more data */
411 GST_LOG_OBJECT (dcaparse,
412 "next sync out of reach (%" G_GSIZE_FORMAT " < %u)", map.size,
422 /* metadata handling */
423 block_size = num_blocks * samples_per_block;
425 if (G_UNLIKELY (dcaparse->rate != rate || dcaparse->channels != chans
426 || dcaparse->depth != depth || dcaparse->endianness != endianness
427 || (!terminator && dcaparse->block_size != block_size)
428 || (size != dcaparse->frame_size))) {
431 caps = gst_caps_new_simple ("audio/x-dts",
432 "framed", G_TYPE_BOOLEAN, TRUE,
433 "rate", G_TYPE_INT, rate, "channels", G_TYPE_INT, chans,
434 "endianness", G_TYPE_INT, endianness, "depth", G_TYPE_INT, depth,
435 "block-size", G_TYPE_INT, block_size, "frame-size", G_TYPE_INT, size,
437 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
438 gst_caps_unref (caps);
440 dcaparse->rate = rate;
441 dcaparse->channels = chans;
442 dcaparse->depth = depth;
443 dcaparse->endianness = endianness;
444 dcaparse->block_size = block_size;
445 dcaparse->frame_size = size;
447 gst_base_parse_set_frame_rate (parse, rate, block_size, 0, 0);
451 gst_buffer_unmap (buf, &map);
453 if (ret == GST_FLOW_OK && size <= map.size) {
454 ret = gst_base_parse_finish_frame (parse, frame, size);
463 * MPEG-PS private1 streams add a 2 bytes "Audio Substream Headers" for each
464 * buffer (not each frame) with the offset of the next frame's start.
465 * These 2 bytes can be dropped safely as they do not include any timing
466 * information, only the offset to the start of the next frame.
467 * See gstac3parse.c for a more detailed description.
471 gst_dca_parse_chain_priv (GstPad * pad, GstObject * parent, GstBuffer * buffer)
473 GstDcaParse *dcaparse = GST_DCA_PARSE (parent);
478 size = gst_buffer_get_size (buffer);
480 newbuf = gst_buffer_copy_region (buffer, GST_BUFFER_COPY_ALL, 2, size - 2);
481 gst_buffer_unref (buffer);
482 ret = dcaparse->baseparse_chainfunc (pad, parent, newbuf);
484 gst_buffer_unref (buffer);
492 remove_fields (GstCaps * caps)
496 n = gst_caps_get_size (caps);
497 for (i = 0; i < n; i++) {
498 GstStructure *s = gst_caps_get_structure (caps, i);
500 gst_structure_remove_field (s, "framed");
505 gst_dca_parse_get_sink_caps (GstBaseParse * parse, GstCaps * filter)
507 GstCaps *peercaps, *templ;
510 templ = gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD (parse));
512 GstCaps *fcopy = gst_caps_copy (filter);
513 /* Remove the fields we convert */
514 remove_fields (fcopy);
515 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), fcopy);
516 gst_caps_unref (fcopy);
518 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), NULL);
521 /* Remove the framed field */
522 peercaps = gst_caps_make_writable (peercaps);
523 remove_fields (peercaps);
525 res = gst_caps_intersect_full (peercaps, templ, GST_CAPS_INTERSECT_FIRST);
526 gst_caps_unref (peercaps);
527 gst_caps_unref (templ);
533 GstCaps *intersection;
536 gst_caps_intersect_full (filter, res, GST_CAPS_INTERSECT_FIRST);
537 gst_caps_unref (res);
545 gst_dca_parse_set_sink_caps (GstBaseParse * parse, GstCaps * caps)
548 GstDcaParse *dcaparse = GST_DCA_PARSE (parse);
550 s = gst_caps_get_structure (caps, 0);
551 if (gst_structure_has_name (s, "audio/x-private1-dts")) {
552 gst_pad_set_chain_function (parse->sinkpad, gst_dca_parse_chain_priv);
554 gst_pad_set_chain_function (parse->sinkpad, dcaparse->baseparse_chainfunc);
560 gst_dca_parse_pre_push_frame (GstBaseParse * parse, GstBaseParseFrame * frame)
562 GstDcaParse *dcaparse = GST_DCA_PARSE (parse);
564 if (!dcaparse->sent_codec_tag) {
568 taglist = gst_tag_list_new_empty ();
571 caps = gst_pad_get_current_caps (GST_BASE_PARSE_SRC_PAD (parse));
572 gst_pb_utils_add_codec_description_to_tag_list (taglist,
573 GST_TAG_AUDIO_CODEC, caps);
574 gst_caps_unref (caps);
576 gst_base_parse_merge_tags (parse, taglist, GST_TAG_MERGE_REPLACE);
577 gst_tag_list_unref (taglist);
579 /* also signals the end of first-frame processing */
580 dcaparse->sent_codec_tag = TRUE;