[MPlayer-dev-eng] [PATCH] fix for -srate bug

[Ed Wildgoose]

Ed Wildgoose wrote:

> Hi Michael,
>
>
>>>> my resampler is already supported, just try -af
>>>> lavcresample=<samplerate>:<filter_length>:<linear_flag>:<polyphase_size>:< 
>>>>
>>>> cutoff>
>>>>     
>>>
>>> Aha.  Interesting.  Can you provide the numbers for this filter to 
>>> match
>>> those that were in your test?
>>>   
>>
>>
>> the code changed somewhat in the meantime so its not trivial ...
>> but <samplerate>:32:1:10 or <samplerate>:32:0:14 might be similar
>>  
>>
>
> I am just fiddling with this new resampler right now.  So far it looks 
> pretty good, but I need to set things up a little better to do some 
> back to back testing against the base setup.  CPU usage though is 
> about the same with this resampler compared to the built in mplayer 
> resampler
>
> I have found one small bug though.  If I set the "linear_flag" to 1, 
> and then try to resample from 48K to 44K, then the output is played 
> back too slowly (as though it doesn't reduce the number of samples 
> correctly)


I have done a little more listening, and have no complaints about the 
lavc resampler so far. 

I do notice though that there is a 6dB difference between the internal 
mplayer high quality resampler and the lavcresampler (lavc is louder).  
I'm assuming that the internal resampler is deliberately dropping the 
volume to avoid clipping due to the resampling, but I would have thought 
that 6dB was a bit extreme (a little over 3 dB should be enough?).  I 
haven't checked the code though... There seem to be many other places 
where mplayer drops the volume a bit.

Have you given any thought as to whether there is any issue here in 
practice?

Thanks

Ed W

P.S. Would love to switch to this version of mplayer by default, but the 
current code seems to have an issue with picking the english language 
audio track on my multi-language DVD's.  I seem to be getting random 
tracks now... Config/operator error, or known issue...?