Re: audio device on ubuntu

On Sun, Aug 30, 2009 at 9:42 AM, Carl Karsten<[email protected]> wrote:
> On Sat, Aug 29, 2009 at 11:29 PM, Dan Dennedy<[email protected]> wrote:
>> On Sat, Aug 29, 2009 at 5:45 PM, Carl Karsten<[email protected]> wrote:
>>> It seams to be playing at 1x.
>>
>> That message is not based on what speed you are playing at; it is
>> based on what speed it was at when you recorded. It could be that the
>> message only appears for a couple of frames and you have some other
>> audio output problem. Do you think you are getting the message for
>> each frame while playing.
>
> yes.  played 30 frames, got 30 lines of # no audio

OK

>>
>>> It is dv from dvgrab/dvswitch, should be
>>> fine.
>
> Actually, some new dvswitch  code was written back in March that
> creates a dv stream from alsa (black video.)  There is a chance that
> it isn't quite right.
>
> you mind reviewing some code:
>
> dvsource-alsa calls dv_buffer_set_audio in dif_audio.c
>
> http://svn.debian.org/wsvn/dvswitch/trunk/src/dvsource-alsa.c
> http://svn.debian.org/wsvn/dvswitch/trunk/src/dif_audio.c
>
> void dv_buffer_set_audio(uint8_t * buffer,
>                         enum dv_sample_rate sample_rate_code,
>                         unsigned frame_count, const int16_t * samples)
> {
>    const struct dv_system * system = dv_buffer_system(buffer);
>
>    assert(sample_rate_code >= 0 && sample_rate_code < dv_sample_rate_count);
>    assert(frame_count >= system->audio_frame_counts[sample_rate_code].min &&
>           frame_count <= system->audio_frame_counts[sample_rate_code].max);
>
>    bool use_12bit = sample_rate_code == dv_sample_rate_32k;
>    unsigned sample_count = frame_count * 2; // stereo
>
>    // Each audio block has a 3-byte block id, a 5-byte AAUX
>    // pack, and 72 bytes of samples.  Audio block 3 in each
>    // sequence has an AS (audio source) pack, audio block 4
>    // has an ASC (audio source control) pack, and the other
>    // packs seem to be optional.
>    static const uint8_t aaux_blank_pack[DIF_PACK_SIZE] = {
>        0xFF, 0xFF, 0xFF, 0xFF, 0xFF
>    };
>    uint8_t aaux_as_pack[DIF_PACK_SIZE] = {
>        // pack id; 0x50 for AAUX source
>        0x50,
>        // bits 0-5: number of audio frames in video frame minus minimum value
>        // bit 6: flag "should be 1"
>        // bit 7: flag for unlocked audio sampling
>        (frame_count - system->audio_frame_counts[sample_rate_code].min)
>        | (1 << 6) | (1 << 7),
>        // bits 0-3: audio mode
>        // bit 4: flag for independent channels
>        // bit 5: flag for "lumped" stereo (?)
>        // bits 6-7: number of audio channels per block minus 1
>        use_12bit << 6,
>        // bits 0-4: system type; 0x0 for DV
>        // bit 5: frame rate; 0 for 29.97 fps, 1 for 25 fps
>        // bit 6: flag for multi-language audio
>        // bit 7: ?
>        dv_buffer_system_code(buffer) << 5,
>        // bits 0-2: quantisation; 0 for 16-bit LPCM, 1 for 12-bit
>        // bits 3-5: sample rate code
>        // bit 6: time constant of emphasis; must be 1
>        // bit 7: flag for no emphasis
>        use_12bit | (sample_rate_code << 3) | (1 << 6) | (1 << 7)
>    };
>    static const uint8_t aaux_asc_pack[DIF_PACK_SIZE] = {
>        // pack id; 0x51 for AAUX source control
>        0x51,
>        // bits 0-1: emphasis flag and ?
>        // bits 2-3: compression level; 0 for once
>        // bits 4-5: input type; 1 for digital
>        // bits 6-7: copy generation management system; 0 for unrestricted
>        (1 << 4),
>        // bits 0-2: ?
>        // bits 3-5: recording mode; 1 for original (XXX should indicate dub)
>        // bit 6: recording end flag, inverted
>        // bit 7: recording start flag, inverted
>        (1 << 3) | (1 << 6) | (1 << 7),
>        // bits 0-6: speed; 0x20 seems to be normal
>        // bit 7: direction: 1 for forward
>        0x20 | (1 << 7),

Here is how libdv interprets this and causes 'no audio' messages:
int dv_is_normal_speed (dv_decoder_t *dv)
{
  int normal_speed = TRUE;
	
  if (dv->std == e_dv_std_iec_61834) {
    normal_speed = (dv->audio->aaux_asc.pc3.speed == 0x20);
  } else if (dv->std == e_dv_std_smpte_314m) {
    if(dv->audio->aaux_as.pc3.system) {
      /* PAL */
      normal_speed = (dv->audio->aaux_asc.pc3.speed == 0x64);
    } else {
      /* NTSC */
      normal_speed = (dv->audio->aaux_asc.pc3.speed == 0x78);
    } /* else */
  }
  return normal_speed;

I bet dvsource-alsa.c is signaling this as a smpte stream and not iec.
As can be seen below from libdv, that is determined by the APT field
in some DV header.

    dv->std = ((header->apt) ? e_dv_std_smpte_314m : e_dv_std_iec_61834);

>From dvsource-alsa.c:

                int apt = (system == &dv_system_625_50) ? 0 : 1;

Here is how libdv determines system:

    dv->system = (header->dsf || (dv_system_50_fields (dv) == 1)) ?
      e_dv_system_625_50 : e_dv_system_525_60;

dsf field comes from a VAUX header. Here is the core part of
dv_system_50_fields:

  if ((id = dv -> vaux_pack [0x60]) != 0xff) {
    if (dv -> vaux_data [id] [2] & 0x20) {
      return 1;
    }
    return 0;
  }

So, you see it is not real simple. Now, we have to sort through
dvsource-alsa.c:dv_buffer_fill_dummy() to figure out what is wrong.

Ya know what? I am kind of resenting the fact that myself and others
have put years of work into libdv, and Roman put years into FFmpeg's
DV. There's two solid implementations that could have been reused. But
no, Ben had to go and reinvent the wheel. I don't want to look into
the finer details of DV again at this time. I already spent enough
time doing that. I no longer maintain Kino and libdv. Ask Ben to fix
it.

-- 
+-DRD-+

------------------------------------------------------------------------------
Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
trial. Simplify your report design, integration and deployment - and focus on 
what you do best, core application coding. Discover what's new with 
Crystal Reports now.  http://p.sf.net/sfu/bobj-july