[mythtv-users] ac3 passthrough results in accelerated video

[John P Poet]

On Thu, May 29, 2008 at 8:24 PM, Craig Treleaven <ctreleaven at cogeco.ca> wrote:
> At 9:31 AM -0700 5/29/08, scram69 wrote:
>>On Thu, May 29, 2008 at 9:15 AM, Michael T. Dean
>><mtdean at thirdcontact.com> wrote:
>>> On 05/29/2008 11:30 AM, scram69 wrote:
>>>> Recently upgraded to 0.21 so I could finally get ac3 passthrough
>>>> working on my osx frontend.  When it works, it's great - however...
>>>>
>>>> Just about every other time I start to watch a recording, the
>>>> playback, both video and sound, is accelerated, as if I were watching
>>>> using 2x-3x timestretch or fast forward.  (I bring up the time stretch
>>>> menu, and it indicates 1.0x)  If I uncheck ac3 passthrough in the
>>>> Setup>General menu, the problem goes away.
>>>>
>>>> Unfortunately, the frontend log output is _exactly_ the same when I
>>>> see this accelerated playback as when I do not.
>>>>
>>>> There have been a number of posts about ac3 passthrough in 0.21, but
>>>> in searching, I haven't come across this issue.  Has anyone else
>>>> experienced this?
>>>
>>> Does it always happen on any given recording or are you saying that on a
>>> particular recording, you'll sometimes see it play incorrectly and
>>> sometimes see it play correctly?
>>>
>>> If it always happens on a given recording, it's just a matter of your
>>> broadcaster choosing to use 32kHz AC-3, which the vast majority of
>>> computer-based sound cards cannot handle (they play it as 48kHz, thus
>>> the "timestrech"--which is actually more of a "chipmunking" as there
>>> shouldn't be pitch correction).  Myth just keeps the video in sync with
>>> the audio as passthrough means that adjusting the audio stream is
>>> impossible.
>>
>>Sorry, I should have been more clear in my first post.  On any
>>particular digital recording (does not happen to recordings from the
>>analog source), sometimes I will get accelerated playback, and other
>>times it will play just fine.  The broadcaster for the digital source
>>in this case is Comcast cable QAM-256.
>>
>>But yes, "chipmunking" is a more accurate description of the audio as
>>there is no pitch correction.  And it does seem to explain why the
>>video would be affected by an audio setting.  Perhaps sometimes myth
>>mis-identifies the AC-3 frequency, and other times gets it right?
>>>
>>> Oh, and if you enable timestretch--even to 1.05x or something--it will
>>> "make it work" by disabling passthrough.  TTBOMK, even if you then go
>>> back to 1.0x, it will still "work" (as Myth won't re-enable passthrough
>>> unless you exit playback and restart playback).
>>
>>Actually, if I enable timestretch during these accelerated playback
>>episodes I get a return to the pre-0.21 AC-3 passthrough behavior -
>>loud static popping.
>
> Me too.  I'm running .21-fixes (from a couple of months ago) on a Core2Duo mini.  The problem recordings will play in accelerated mode one, two or three times in a row and then play fine.  Lots of recordings play fine.  I think I only see this problem with ABC, NBC and CBS; not with the Canadian networks.  BTW, all are recorded OTA via an HDHomerun.
>
> Also, the frontend hangs on exiting playback, sometimes.  I keep meaning to set up logging but haven't done it yet.
>
> Neither of these problems happened to any extent under .20-fixes.  As you might imagine, it is having a detrimental effect on the SAF (spousal approval factor).  Thanks to scram69 for pointing out the issue and the correlation with digital audio--I had no idea that was related!
>
> On the Mac, what is a good tool identify the types of audio and video data in a file?  VLC doesn't give any details.


One *big* change between 0.20 and 0.21 is that once you turn on
timestetch while watching a show, it never turns back off -- even if
you go back to 1.0x speed.  As part of this, it *decodes* DD5.1,
adjusts it's speed (if necessary), then re-encodes the audio back into
DD5.1 (preserving the original channel allocation) and sends it out
your S/PDIF.  This is actually all pretty cool, but it has introduced
some new problems.

If you are getting loud static popping, I wonder if it is not
correctly re-encoding the DD5.1?  I have not tried this in a *long*
time, but you used to be able to change to a different ac3 library by
compiling with:

configure --enable-liba52bin --disable-decoder=ac3 --enable-decoder=liba52

Mark Spieth would have to comment on whether that still works.  Of
course your problem might also be that no ac3/a52 library is being
linked in at all...


If timestretch is "on", then it can occasionally get confused by the
audiostream and miss-detect "6 channels" as "2 channel", which results
in a chipmunk sound.  For example see:

http://svn.mythtv.org/trac/ticket/5126


You can post a (possibly) useful log if you run mythfrontend with audio logging:

mythfrontend -v audio,libav,playback -l somelogfile


Hope some of this info is useful.

John
-- 
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?