I'm used to see confortable bitrate value with lame --standard. But here is a sample with an unexpected high bitrate: 250 kbps for the 30 samples sample, 240 kbps for the full track. It's an organ piece, without any short block, but a lot of 320 kbps frames (I can't understand why: other encoders don't show the same behaviour with this piece).


3.90.3 is for comparison more efficient: I couldn't ABX any difference, and bitrate is 30 kbps lower.
For me it's not a problem. But this sample might help developers to tune lame in order to increase efficiency (i.e. same quality level at lower bitrate) in some case (it's surely possible, in regard to 3.90.3 performances).

ftp://ftp2.foobar2000.net/foobar/Bach-Latry%20(2).fla


Other thing to note: with lame 3.97a3, fast standard is ~210 kbps, whereas standard is 250~kbps. Huge difference (can't ABX it yet).

3.90.3 -> 802 KB
3.93.1 -> 716 KB
3.96 -> 914 KB
3.97a3 -> 913 KB...

The stereo type repartition greatly differs between .90/93 & .96/97 btw...

(:hello: guru...)

I think that the higher bitrates for Lame 3.96/97 are due to the higher percentage of Stereo frames used - Lame 3.90.3 uses more MS frames and can therefore get away with using less bits for encoding. I also encountered this anomaly (usually my Lame 3.96 files are significantly smaller than 3.90.3 files) with music with lots of stereo separation like e.g. some Beatles records. Maybe Lame 3.96 is too careful with regard to using MS frames and therefore wastes bits - OR, maybe it does the right thing and improves on lame 3.90.3 by being more careful. Maybe the lame devs could look into this whether this behaviour of Lame 3.96 is a bug or an asset...