Recognize AAC-HE SBR with upsampling and correct duration, bitrate, seek and resume behaviour for such files. When SBR upsampling is used the decoder outputs the double amount of samples per frame. As the seek and resume functions do not know about this fact a special handling is introduced. Fixes issues reported in FS#11916.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@29186 a1c6a512-1295-4272-9138-f99709370657

2 files changed, 51 insertions, 5 deletions

diff --git a/apps/codecs/aac.c b/apps/codecs/aac.c
index af40385..37e525e 100644
--- a/apps/codecs/aac.c
+++ b/apps/codecs/aac.c
@@ -59,6 +59,7 @@ enum codec_status codec_main(void)
     NeAACDecHandle decoder;
     int err;
     uint32_t s = 0;
+    uint32_t sbr_fac = 1;
     unsigned char c = 0;
     void *ret;
 
@@ -143,13 +144,26 @@ next_track:
         decoder->fb_intermed[1] = &gb_fb_intermed[1][0];
     }
 
+#ifdef SBR_DEC
+    /* The file uses SBR. */
+    if (decoder->forceUpSampling) {
+        sbr_fac = 2;
+    } else {
+        sbr_fac = 1;
+    }
+#endif
+
     ci->id3->frequency = s;
 
     i = 0;
     
     if (file_offset > 0) {
+        /* Resume the desired (byte) position. Important: When resuming SBR
+         * upsampling files the resulting sound_samples_done must be expanded 
+         * by a factor of 2. This is done via using sbr_fac. */
         if (alac_seek_raw(&demux_res, &input_stream, file_offset,
                           &sound_samples_done, (int*) &i)) {
+            sound_samples_done *= sbr_fac;
             elapsed_time = (sound_samples_done * 10) / (ci->id3->frequency / 100);
             ci->set_elapsed(elapsed_time);
         } else {
@@ -174,9 +188,14 @@ next_track:
 
         /* Deal with any pending seek requests */
         if (ci->seek_time) {
+            /* Seek to the desired time position. Important: When seeking in SBR
+             * upsampling files the seek_time must be divided by 2 when calling 
+             * alac_seek and the resulting sound_samples_done must be expanded 
+             * by a factor 2. This is done via using sbr_fac. */
             if (alac_seek(&demux_res, &input_stream,
-                          ((ci->seek_time-1)/10)*(ci->id3->frequency/100),
+                          ((ci->seek_time-1)/10/sbr_fac)*(ci->id3->frequency/100),
                           &sound_samples_done, (int*) &i)) {
+                sound_samples_done *= sbr_fac;
                 elapsed_time = (sound_samples_done * 10) / (ci->id3->frequency / 100);
                 ci->set_elapsed(elapsed_time);
             
@@ -234,6 +253,10 @@ next_track:
         /* Output the audio */
         ci->yield();
         
+        /* Ensure correct sample_duration is used. For SBR upsampling files
+         * sample_duration is only half the size of real output frame size. */
+        sample_duration *= sbr_fac;
+        
         framelength = (frame_info.samples >> 1) - lead_trim;
         
         if (i == demux_res.num_sample_byte_sizes - 1 && framelength > 0)
@@ -266,7 +289,7 @@ next_track:
         {
             /* frame_info.samples can be 0 for the first frame */
             lead_trim -= (i > 0 || frame_info.samples)
-                ? (frame_info.samples >> 1) : sample_duration;
+                ? (frame_info.samples >> 1) : (uint32_t)framelength;
 
             if (lead_trim < 0 || ci->id3->lead_trim == 0)
             {
@@ -275,7 +298,7 @@ next_track:
         }
 
         /* Update the elapsed-time indicator */
-        sound_samples_done += sample_duration;
+        sound_samples_done += framelength;
         elapsed_time = (sound_samples_done * 10) / (ci->id3->frequency / 100);
         ci->set_elapsed(elapsed_time);
         i++;
diff --git a/apps/metadata/mp4.c b/apps/metadata/mp4.c
index a59b3f9..c9c691f 100644
--- a/apps/metadata/mp4.c
+++ b/apps/metadata/mp4.c
@@ -73,6 +73,9 @@
 #define MP4_udta FOURCC('u', 'd', 't', 'a')
 #define MP4_extra FOURCC('-', '-', '-', '-')
 
+/* Used to correct id3->samples, if SBR upsampling was detected in esds atom. */
+static bool SBR_upsampling_used = false;
+
 /* Read the tag data from an MP4 file, storing up to buffer_size bytes in
  * buffer.
  */
@@ -272,7 +275,6 @@ static bool read_mp4_esds(int fd, struct mp3entry* id3, uint32_t* size)
     
         if (type == 5)
         {
-            DEBUGF("MP4: SBR\n");
             unsigned int old_index = index;
 
             sbr = true;
@@ -342,6 +344,12 @@ static bool read_mp4_esds(int fd, struct mp3entry* id3, uint32_t* size)
              * decoding (parts of) the file.
              */
             id3->frequency *= 2;
+            
+            /* Set this to true to be able to calculate the correct runtime 
+             * and bitrate. */
+            SBR_upsampling_used = true;
+
+            sbr = true;
         }
     }
     
@@ -665,6 +673,7 @@ static bool read_mp4_container(int fd, struct mp3entry* id3,
                 {
                     uint32_t subsize;
                     uint32_t subtype;
+                    bool sbr_used;
 
                     /* Get frequency from the decoder info tag, if possible. */
                     lseek(fd, 2, SEEK_CUR);
@@ -676,7 +685,14 @@ static bool read_mp4_container(int fd, struct mp3entry* id3,
                     
                     if (subtype == MP4_esds)
                     {
-                        read_mp4_esds(fd, id3, &size);
+                        sbr_used = read_mp4_esds(fd, id3, &size);
+                        if (sbr_used)
+                        {
+                            if (SBR_upsampling_used)
+                                DEBUGF("MP4: AAC-HE, SBR upsampling\n");
+                            else
+                                DEBUGF("MP4: AAC-HE, SBR\n");
+                        }
                     }
                 }
             }
@@ -730,6 +746,7 @@ static bool read_mp4_container(int fd, struct mp3entry* id3,
 
 bool get_mp4_metadata(int fd, struct mp3entry* id3)
 {
+    SBR_upsampling_used = false;
     id3->codectype = AFMT_UNKNOWN;
     id3->filesize = 0;
     errno = 0;
@@ -743,6 +760,12 @@ bool get_mp4_metadata(int fd, struct mp3entry* id3)
             logf("Not an ALAC or AAC file");
             return false;
         }
+        
+        /* SBR upsampling will output double amount of samples per frame. */
+        if (SBR_upsampling_used)
+        {
+            id3->samples *= 2;
+        }
 
         id3->length = ((int64_t) id3->samples * 1000) / id3->frequency;