[MPlayer-dev-eng] pre5?

[Anders Johansson]

Hi,

> On Thu, Jun 06, 2002 at 10:08:45PM +0200, Colin Leroy wrote:
> > On Thu, 6 Jun 2002 15:59:02 -0400 D Richard Felker III <dalias at aerifal.cx>
> > wrote:
> > 
> > >> ao sdl volume control (in case it's important ;-)
> > >
> > >Which is brain damaged -- it just duplicates the software volume
> > >control that already exists.
> > 
> > Software vol control is lower quality and requires kilometrical command
> > line ; if it is a feature duplication, it's an enhancement of the existing
> > feature imho.
> 
> OK. I guess I'm not the only one who thinks the AO plugins suck. 

When I designed the plugin interface I didn't know where it should go
cause I wasn't familiar with the design of Mplayer so it ended up at
the wrong place. If you are interested in why it looks the way it does
check the mailing list logs back in October 2001, the whole design
discussion was made back then.

> You don't need a long commandline though -- just put the stuff to
> load the volume plugin in your config file and then use * and / keys
> to control volume as usual.
> 
> BTW, once the new AO plugin architecture is in place, is anyone
> interested in fixing resample so the pitch isn't off by so much? It's
> frequently off by as much as 5% with the current implementation...

Can you hear it? I have never tried listening t music or anything like
that. The reason for the offset is that I am using a fix length poly
phase filterbank to do the resampling. The prototype filter is
designed using a Blackman window (necessary to minimize aliasing
components). I had a look at including the filter design into Mplayer
but the algorithm is huge, so I didn't think it was worth the
effort. Also I wanted to prioritize reducing the quantization effect
rather than the frequency error something I could not do if I
implemented the Blackman algorithm (cause requires post optimization
using a genetic algorithm). I have previously asked about design
strategies for the resampling on this list and back then no one
disagreed with me. The reason for the size of the frequency error is
that due to a bug (which I have never found - not in my code) I have
to select 2^n size filter (not optimal), otherwise the output data
gets over written. Perhaps this has been fixed I haven't checked
lately (I will do that once I have committed the EQ (soon ready :)).

> Rich
> 

By the way is anyone else except for me working on the new plugin
architecture?

// Anders