[FFmpeg-devel] [PATCH 2/2] lavfi: EBU R.128 scanner.

[Clément Bœsch]

On Mon, Oct 01, 2012 at 10:14:56PM +0200, Clément Bœsch wrote:
[...]
> > Looks great otherwise, I didn't read it very carefully, but I think we
> > can commit and let it mature then.
> > 
> 
> Right, then I'll push in a moment.
> 

Applied.

> Note: the version I'm going to push is slightly improved. I can't remember
> all the various changes but the main ones are:
>  - Integrated loudness results fixed
>  - I & LRA should be faster, with an histogram storing more info
>  - added a summary at the end
> 
> > I also wonder if we could make it a bit more flexible, for example to
> > make the short-term loudness and the momentary loudness windows
> > configurable (possibly useful for other purposes), but I don't
> > consider this blocking.
> 
> The recommendation is quite explicit:
> 
>     The momentary loudness uses a sliding rectangular time window of
>     length 0.4 s. The measurement is not gated.
> 
>     The short-term loudness uses a sliding rectangular time window of
>     length 3 s. The measurement is not gated. The update rate for ‘live
>     meters’ shall be at least 10 Hz.
> 
> I'm sure we can add various customisation options, but I wouldn't play too
> much with these settings. OTOH Maybe the accuracy could be configurable
> for instance (see HIST_GRAIN), at the cost of a likely slower code. It
> would also needs some extra checks to make sure the range is sane (you
> need to take into account various stuff like the range of energies in
> get_histogram()).
> 
> Anyway, the next step is to add the True Peak feature, and maybe some more
> fancy information on the graph. I'll work on this later.
> 
> But now, if anyone wants to play with the math behind the two filters (PRE
> and RLB) to merge them, as well as the support of other frequencies than
> 48kHz to avoid resampling, that would very welcome. It involves some kind
> of complex math according to what I read on HydrogenAudio. Here are some
> references I found over the threads:
> 
>   http://www.musicdsp.org/files/Audio-EQ-Cookbook.txt
>   http://www.scribd.com/doc/6531763/DirectForm-Filter-Parameter-Quantization
> 
> Oh and BTW, thanks a lot to cbsrobot for providing me some references
> results which greatly helped me to fix the integrated loudness.
> 

And forgot to add: there is some start/stop/reset commands to implement
(it's requested in the recommendation).

-- 
Clément B.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://ffmpeg.org/pipermail/ffmpeg-devel/attachments/20121001/9f9e7604/attachment.asc>