On Tue, Aug 05, 2014 at 04:23:23PM +0200, Raffaele Morelli wrote: > Ok, works but I have a couple of question. > > I see only 751 records in the outfile. How ebur128 works? Is the file > scanned entirely and how? Yes, the entire audio file is processed. Internally ebumeter and ebur128 compute two histograms, each consisting of 751 bins from -70 to +5 dB in steps of 0.1 dB. The first histogram is the 'momentary' loudness, the average over a sliding window of 400 ms. The second is the 'short-term' loudness, computed using a sliding window of 3 seconds. The histograms are then used to compute average loudness and loudness range according to the EBU-R128 standard. Ebumeter will show them in real-time. For the details, see my LAC 2011 paper which you can find here: <http://kokkinizita.linuxaudio.org/papers/index.html>. The output file of ebur128 consists of five columns: 1. X-axis, level in dB 2. Cumulative probability of the momentary loudness, this is the histogram integrated and normalised to 1. In other words the probability that the level is less than the X-axis value. 3. Same for the short-term loudness. 4. The momentary loudness histogram, counts for each bin. 5. Same for the short-term loudnes. The gnuplot commands I posted will plot 1 vs 4 or 1 vs 5. Ciao, -- FA A world of exhaustive, reliable metadata would be an utopia. It's also a pipe-dream, founded on self-delusion, nerd hubris and hysterically inflated market opportunities. (Cory Doctorow) _______________________________________________ Linux-audio-user mailing list Linux-audio-user@xxxxxxxxxxxxxxxxxxxx http://lists.linuxaudio.org/listinfo/linux-audio-user
