[MPlayer-users] Bug: Poor quality decoding of ac3 audio

[Rashkae]

On 11-03-02 08:07 PM, Vladimir Mosgalin wrote:
> Hi Rashkae!
>
>   On 2011.03.02 at 19:36:32 -0500, Rashkae wrote next:
>
>    
>> On 11-03-02 02:07 PM, Vladimir Mosgalin wrote:
>>      
>>> I didn't check this sample, but some ac3 streams which have too loud
>>> sounds can get cracking and other artifacts when downmixing due to some
>>> overflows, the solution is to use
>>> -channels 6 -af volume=-10,pan=.. (downmixing string)
>>> or -channels 6 -af volume=-10,hrtf if you are downmixing for headphones.
>>>
>>>        
>> Downmixing with pan does indeed allow mplayer to play/decode the
>> audio without glitches.  I don't even have to reduce the volume,
>> using the 'basic' formula of pan=2:1:0:0:1:1:0:0:1:0.5:0.5:1:1 in
>> the documentation.
>>
>> However, I still think there is a flaw in mplayer somewhere,
>> somehow.  It is my understanding that mplayer with default options
>> should be using liba52 to downmix the  ac3 stream, exactly the same
>> as a52dec does.. but I'm getting very different results between the
>> two programs.  As for whether the liba52 on my system (0.7.4) is
>> newer or older code than what Mplayer svn is using..I don't
>> understand the Mplayer internals nearly well enough to tell.
>>      
> Did you check result of
> a52dec -o wav -r
> and
> a52dec -o wavdolby -r
> ?
>    

I did check with -o wav -r.  I haven't tried anything with wavdolby

I'll have to produce some samples where the artifact I'm complaining 
about is more pronounced.  I think from episodes of Supernatural would 
be good, since the ''motor" sound effect used for the Impala automobile 
produced it quite often...

In any case, I thank you very much for your informative replies.  Are 
there any suggested overall best practice suggestions for pan 
downmixing?  Google can find some examples for me, but it seems like a 
very subjective field.  Anything resembling a consensus of good starting 
point?