Message ID | 1408085019-5277-1-git-send-email-gtmkramer@xs4all.nl (mailing list archive) |
---|---|
State | Accepted |
Delegated to: | Takashi Iwai |
Headers | show |
15.08.2014 12:43, Jurgen Kramer wrote: > Functions 'snd_pcm_format_silence_64' and 'snd_pcm_format_size' also need to be > able to handle the DSD formats. 0 is not a DSD silence value. It is the most negative value, similar to -32768 for S16 samples. We need something that will average to zero. ffmpeg uses a repeating pattern of 0x69 here, with this comment: /* 0x69 = 01101001 * This pattern "on repeat" makes a low energy 352.8 kHz tone * and a high energy 1.0584 MHz tone which should be filtered * out completely by any playback system --> silence */ https://ffmpeg.org/pipermail/ffmpeg-cvslog/2014-April/076427.html > > Signed-off-by: Jurgen Kramer <gtmkramer@xs4all.nl> > --- > src/pcm/pcm_misc.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/src/pcm/pcm_misc.c b/src/pcm/pcm_misc.c > index 24d52f9..c52e7bf 100644 > --- a/src/pcm/pcm_misc.c > +++ b/src/pcm/pcm_misc.c > @@ -317,11 +317,13 @@ ssize_t snd_pcm_format_size(snd_pcm_format_t format, size_t samples) > switch (format) { > case SNDRV_PCM_FORMAT_S8: > case SNDRV_PCM_FORMAT_U8: > + case SNDRV_PCM_FORMAT_DSD_U8: > return samples; > case SNDRV_PCM_FORMAT_S16_LE: > case SNDRV_PCM_FORMAT_S16_BE: > case SNDRV_PCM_FORMAT_U16_LE: > case SNDRV_PCM_FORMAT_U16_BE: > + case SNDRV_PCM_FORMAT_DSD_U16_LE: > return samples * 2; > case SNDRV_PCM_FORMAT_S18_3LE: > case SNDRV_PCM_FORMAT_S18_3BE: > @@ -387,6 +389,8 @@ u_int64_t snd_pcm_format_silence_64(snd_pcm_format_t format) > case SNDRV_PCM_FORMAT_S20_3BE: > case SNDRV_PCM_FORMAT_S18_3LE: > case SNDRV_PCM_FORMAT_S18_3BE: > + case SNDRV_PCM_FORMAT_DSD_U8: > + case SNDRV_PCM_FORMAT_DSD_U16_LE: > return 0; > case SNDRV_PCM_FORMAT_U8: > return 0x8080808080808080ULL; >
At Fri, 15 Aug 2014 15:02:35 +0600, Alexander E. Patrakov wrote: > > 15.08.2014 12:43, Jurgen Kramer wrote: > > Functions 'snd_pcm_format_silence_64' and 'snd_pcm_format_size' also need to be > > able to handle the DSD formats. > > 0 is not a DSD silence value. It is the most negative value, similar to > -32768 for S16 samples. We need something that will average to zero. > > ffmpeg uses a repeating pattern of 0x69 here, with this comment: > > /* 0x69 = 01101001 > * This pattern "on repeat" makes a low energy 352.8 kHz tone > * and a high energy 1.0584 MHz tone which should be filtered > * out completely by any playback system --> silence > */ > > https://ffmpeg.org/pipermail/ffmpeg-cvslog/2014-April/076427.html Good point. We need a fix in the kernel side, too. Takashi
diff --git a/src/pcm/pcm_misc.c b/src/pcm/pcm_misc.c index 24d52f9..c52e7bf 100644 --- a/src/pcm/pcm_misc.c +++ b/src/pcm/pcm_misc.c @@ -317,11 +317,13 @@ ssize_t snd_pcm_format_size(snd_pcm_format_t format, size_t samples) switch (format) { case SNDRV_PCM_FORMAT_S8: case SNDRV_PCM_FORMAT_U8: + case SNDRV_PCM_FORMAT_DSD_U8: return samples; case SNDRV_PCM_FORMAT_S16_LE: case SNDRV_PCM_FORMAT_S16_BE: case SNDRV_PCM_FORMAT_U16_LE: case SNDRV_PCM_FORMAT_U16_BE: + case SNDRV_PCM_FORMAT_DSD_U16_LE: return samples * 2; case SNDRV_PCM_FORMAT_S18_3LE: case SNDRV_PCM_FORMAT_S18_3BE: @@ -387,6 +389,8 @@ u_int64_t snd_pcm_format_silence_64(snd_pcm_format_t format) case SNDRV_PCM_FORMAT_S20_3BE: case SNDRV_PCM_FORMAT_S18_3LE: case SNDRV_PCM_FORMAT_S18_3BE: + case SNDRV_PCM_FORMAT_DSD_U8: + case SNDRV_PCM_FORMAT_DSD_U16_LE: return 0; case SNDRV_PCM_FORMAT_U8: return 0x8080808080808080ULL;
Functions 'snd_pcm_format_silence_64' and 'snd_pcm_format_size' also need to be able to handle the DSD formats. Signed-off-by: Jurgen Kramer <gtmkramer@xs4all.nl> --- src/pcm/pcm_misc.c | 4 ++++ 1 file changed, 4 insertions(+)