[SCA-Dance] MP3 encoding

tmcd at panix.com tmcd at panix.com
Sun Oct 9 01:54:52 EDT 2011


Some time back, I asked about what settings to use for MP3 recordings
for SCA dance music, to be played over a boom box for a dance
practice.  I think that everyone agreed with Dani Zweig,

> Use 192 or better, because you never know when you'll want to move
> them to a better system or use them in a place where the difference
> is noticeable or make copies for someone else - at which point
> saving a bit of time and a bit of storage will look like a false
> economy.
>
> We experimented at home, and determined that the difference between
> 128 and 192 was fairly noticeable, but the difference between 192
> and 256 only mattered in specialized circumstances.

In case anyone is curious or wants to do the same: I am using a
program named "lame" to create MP3s.  I read the man page and
http://wiki.hydrogenaudio.org/index.php?title=LAME#Recommended_encoder_settings,
and settled on these settings:

--vbr-new -V 0 -q 0 --replaygain-accurate --clipdetect

(In the current version of lame, some of those options implies some of
the others listed.  I prefer to be explicit.)

--vbr-new -V 0: The default encoding is CBR, Constant Bitrate.  It
turns out not to be very efficient in storage.  VBR, Variable Bitrate,
allows a larger bitrate when the music is complicated and needs it for
accuracy, but a smaller bitrate when it's not needed (like
near-silence at the start or end).  MP3 players made in the last few
years should support VBR.  -V 0 is the highest VBR quality, apparently
up near 245 Kbps.  The wiki page says "will normally produce
transparent encoding (transparent = most people can't distinguish the
MP3 from the original in an ABX blind test)."

I played a bit with -V 0 versus -V 4.  The wiki page says "background
noise and low bitrate requirement, small sizes ...  -V4 (~165 kbps),
... -V6 produces an 'acceptable' quality, while -V4 should be close to
perceptual transparency."  -V 4 on the test song saved only 25%,
so I decided to stick to the higher accuracy.

-q 0: "Bitrate is of course the main influence on quality.  The higher
the bitrate, the higher the quality.  But for a given bitrate, we have
a choice of algorithms to determine the best scalefactors and Huffman
encoding (noise shaping)."  0 is the slowest and best.  I am more than
willing to trade one-time encoding time to get better accuracy.

--replaygain-accurate --clipdetect
http://www.replaygain.org/: "a random play through a music collection
can have one leaping for the volume control every other track.
... within each audio file, information can be stored about what
volume change would be required to play each track or album at a
standard [subjective] loudness, and players can use this 'replay gain'
information to automatically nudge the volume up or down as required."
This is the slower but most accurate algorithm.  --clipdetect warns
about clipping, which sounds like a good idea to me.

Daniel de Lindo
-- 
Tim McDaniel; Reply-To: tmcd at panix.com


More information about the Sca-dance mailing list