dv.c

FFmpeg is an audio/video conversion tool. It includes libavcodec, the leading open source codec libr

    buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
	     (0 << 3) |     /* FSC: for 50Mb/s 0 - first channel; 1 - second */
	     7;             /* reserved -- always 1 */
    buf[2] = dif_num;       /* DIF block number Video: 0-134, Audio: 0-8 */
    return 3;
}

static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
{
    if (syb_num == 0 || syb_num == 6) {
	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
		 (0<<4)  | /* AP3 (Subcode application ID) */
		 0x0f;     /* reserved -- always 1 */
    } 
    else if (syb_num == 11) {
	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
                 0x7f;     /* reserved -- always 1 */
    }
    else {
	buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
                 (0<<4)  | /* APT (Track application ID) */
		 0x0f;     /* reserved -- always 1 */
    }
    buf[1] = 0xf0 |            /* reserved -- always 1 */
             (syb_num & 0x0f); /* SSYB number 0 - 11   */
    buf[2] = 0xff;             /* reserved -- always 1 */
    return 3;
}

static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
{
    int i, j, k;
    
    for (i = 0; i < c->sys->difseg_size; i++) {
       memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
       
       /* DV header: 1DIF */
       buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
       buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
       buf += 72; /* unused bytes */
				   
       /* DV subcode: 2DIFs */
       for (j = 0; j < 2; j++) {
          buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
	  for (k = 0; k < 6; k++) {
	     buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
	     buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
	  }
	  buf += 29; /* unused bytes */
       }
       
       /* DV VAUX: 3DIFs */
       for (j = 0; j < 3; j++) {
	  buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
	  for (k = 0; k < 15 ; k++)
	     buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
	  buf += 2; /* unused bytes */
       } 
       
       /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
       for (j = 0; j < 135; j++) {
            if (j%15 == 0) {
	        buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
	        buf += dv_write_pack(dv_aaux_packs_dist[i][j/15], c, buf);
		buf += 72; /* shuffled PCM audio */
	    }
	    buf += dv_write_dif_id(dv_sect_video, i, j, buf);
	    buf += 77; /* 1 video macro block: 1 bytes control
			                       4 * 14 bytes Y 8x8 data
			                           10 bytes Cr 8x8 data
						   10 bytes Cb 8x8 data */
       }
    }
}

static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, uint8_t* frame_ptr)
{
    int i, j, d, of;
    for (i = 0; i < c->sys->difseg_size; i++) {
       frame_ptr += 6 * 80; /* skip DIF segment header */
       for (j = 0; j < 9; j++) {
          for (d = 8; d < 80; d+=2) {
	     of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
	     frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
	     frame_ptr[d+1] = pcm[of*2]; //        that DV is a big endian PCM       
          }
          frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
       }
    }
}

static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
{
    int i, j;
    int ptr = 0;

    for (i = 0; i < c->sys->difseg_size; i++) {
       ptr += 6 * 80; /* skip DIF segment header */
       for (j = 0; j < 135; j++) {
            if (j%15 == 0)
	        ptr += 80; /* skip Audio DIF */
	    ptr += 3;
	    memcpy(frame_ptr + ptr, video_data + ptr, 77);
	    ptr += 77;
       }
    }
}

/* 
 * This is the dumbest implementation of all -- it simply looks at
 * a fixed offset and if pack isn't there -- fails. We might want
 * to have a fallback mechanism for complete search of missing packs.
 */
static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
{
    int offs;
    
    switch (t) {
    case dv_audio_source:
          offs = (80*6 + 80*16*3 + 3);
	  break;
    case dv_audio_control:
          offs = (80*6 + 80*16*4 + 3);
	  break;
    case dv_video_control:
          offs = (80*5 + 48 + 5);
          break;
    default:
          return NULL;
    }   

    return (frame[offs] == t ? &frame[offs] : NULL);
}

/* 
 * There's a couple of assumptions being made here:
 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
 *    We can pass them upwards when ffmpeg will be ready to deal with them.
 * 2. We don't do software emphasis.
 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
 *    are converted into 16bit linear ones.
 */
static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2)
{
    int size, i, j, d, of, smpls, freq, quant, half_ch;
    uint16_t lc, rc;
    const DVprofile* sys;
    const uint8_t* as_pack;
    
    as_pack = dv_extract_pack(frame, dv_audio_source);
    if (!as_pack)    /* No audio ? */
	return 0;
   
    sys = dv_frame_profile(frame);
    smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
    freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
    quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
    
    if (quant > 1)
	return -1; /* Unsupported quantization */

    size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
    half_ch = sys->difseg_size/2;

    /* for each DIF segment */
    for (i = 0; i < sys->difseg_size; i++) {
       frame += 6 * 80; /* skip DIF segment header */
       if (quant == 1 && i == half_ch) {
           if (!pcm2)
               break;
           else
               pcm = pcm2;
       }

       for (j = 0; j < 9; j++) {
          for (d = 8; d < 80; d += 2) {
	     if (quant == 0) {  /* 16bit quantization */
		 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
                 if (of*2 >= size)
		     continue;
		 
		 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
		 pcm[of*2+1] = frame[d]; //        that DV is a big endian PCM
		 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
		     pcm[of*2+1] = 0;
	      } else {           /* 12bit quantization */
		 lc = ((uint16_t)frame[d] << 4) | 
		      ((uint16_t)frame[d+2] >> 4);
		 rc = ((uint16_t)frame[d+1] << 4) |
	              ((uint16_t)frame[d+2] & 0x0f);
		 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
		 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));

		 of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
                 if (of*2 >= size)
		     continue;

		 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
		 pcm[of*2+1] = lc >> 8; //        that DV is a big endian PCM
		 of = sys->audio_shuffle[i%half_ch+half_ch][j] + 
		      (d - 8)/3 * sys->audio_stride;
		 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
		 pcm[of*2+1] = rc >> 8; //        that DV is a big endian PCM
		 ++d;
	      }
	  }
		
	  frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
        }
    }

    return size;
}

static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
{
    const uint8_t* as_pack;
    const DVprofile* sys;
    int freq, smpls, quant, i;

    sys = dv_frame_profile(frame);
    as_pack = dv_extract_pack(frame, dv_audio_source);
    if (!as_pack || !sys) {    /* No audio ? */
        c->ach = 0;
	return -1;
    }
    
    smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
    freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
    quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
    c->ach = (quant && freq == 2) ? 2 : 1;

    /* The second stereo channel could appear in IEC 61834 stream only */
    if (c->ach == 2 && !c->ast[1]) {
        c->ast[1] = av_new_stream(c->fctx, 0);
	if (c->ast[1]) {
            av_set_pts_info(c->ast[1], 64, 1, 30000);
	    c->ast[1]->codec.codec_type = CODEC_TYPE_AUDIO;
	    c->ast[1]->codec.codec_id = CODEC_ID_PCM_S16LE;
	} else
	    c->ach = 1;
    }
    for (i=0; i<c->ach; i++) {
       c->ast[i]->codec.sample_rate = dv_audio_frequency[freq];
       c->ast[i]->codec.channels = 2;
       c->ast[i]->codec.bit_rate = 2 * dv_audio_frequency[freq] * 16;
    }
    
    return (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
}

static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
{
    const DVprofile* sys; 
    const uint8_t* vsc_pack;
    AVCodecContext* avctx;
    int apt, is16_9;
    int size = 0;
    
    sys = dv_frame_profile(frame);
    if (sys) {
        avctx = &c->vst->codec;
	
	avctx->frame_rate = sys->frame_rate;
        avctx->frame_rate_base = sys->frame_rate_base;
        avctx->width = sys->width;
        avctx->height = sys->height;
        avctx->pix_fmt = sys->pix_fmt;
        
	/* finding out SAR is a little bit messy */
	vsc_pack = dv_extract_pack(frame, dv_video_control);
        apt = frame[4] & 0x07;
	is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
	                       (!apt && (vsc_pack[2] & 0x07) == 0x07)));
	avctx->sample_aspect_ratio = sys->sar[is16_9];
	
	size = sys->frame_size;
    }
    return size;
}

/* 
 * The following 6 functions constitute our interface to the world
 */

int dv_assemble_frame(DVMuxContext *c, AVStream* st,
                      const uint8_t* data, int data_size, uint8_t** frame)
{
    uint8_t pcm[8192];
    int fsize, reqasize;
   
    *frame = &c->frame_buf[0];
    if (c->has_audio && c->has_video) {  /* must be a stale frame */
        dv_format_frame(c, *frame);
	c->frames++;
	c->has_audio = c->has_video = 0;
    }
    
    if (st->codec.codec_type == CODEC_TYPE_VIDEO) {
        /* FIXME: we have to have more sensible approach than this one */
	if (c->has_video)
	    av_log(&st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
	    
        dv_inject_video(c, data, *frame);
	c->has_video = 1;
	data_size = 0;
    } 
    
    reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
    fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
    if (st->codec.codec_type == CODEC_TYPE_AUDIO || (c->has_video && fsize >= reqasize)) { 
	if (fsize + data_size >= reqasize && !c->has_audio) {
            if (fsize >= reqasize) {
	        fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
            } else {
	        fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);