-V 0 vs. -b 320 vs. wav

2007-09-12 10:46:40

Hey folks,

Here are the files from the A/B testing referenced in the following thread:

http://www.hydrogenaudio.org/forums/index....showtopic=57462

There are 3 files for each A/B comparison (6 total)

Tina Turner - What's love got to do with it:
-30 second sample of mp3 (-V 0)
-5 second sub-sample of that mp3 decoded into wav format
-identical 5 second clip of original wav

Michael Jackson - Smooth Criminal
-20 second sample of mp3 (-b 320)
-8 second sub-sample of that mp3 decoded into wav format
-identical 8 second clip of original wav

-V 0 vs. -b 320 vs. wav

Reply #1 – 2007-09-12 12:09:02

If you're interested, try the attached lossyFLAC files. They're 342kbps and 377kbps. (Genuine lossless FLAC files would be over 800kbps).

Any FLAC decoder/player will work with lossyFLAC.

I suggest foobar2k as a good player / decoder / converter / ABX comparator / many other things.

Cheers,
David.

P.S. The .wav files you supplied aren't of identical sections of the audio. Same length, but not exactly the same sections.

-V 0 vs. -b 320 vs. wav

Reply #2 – 2007-09-13 00:36:30

Thanks David.  I'll check them out when I get home. I've already got fb2k (lean and mean, great player), as wmp lays an egg every time I try to run hi res samples (from Linn, Mitek etc) through it. I'll download the latest version or plug-ins for the abx capability (didn't know about that one).  As for the wavs, I haven't the first clue how they wound up being different (aside from the encoding).  I loaded both files (orig wav first, decoded mp3 second) Audacity, zoomed in, highlighted the respective chunks and clicked export.  Since I'm relying on drag and drop, it's certainly possible that the (start/end) could have been off by a millisecond or two, but the ringers are all mid-sample.  Still, sloppy evidence won't do (why give the naysayers the pleasure, eh?), and since I promised to test the 3.98 beta anyway, I think I'll take the following approach:

-Sample the problem clip from the wave
-Encode that clip to mp3 using razorlame (save new file)
-decode mp3 back to wave in Audacity.

This should preserve the length perfectly, right? As I think about it, my previous approach was more cumbersome (encode full length song to mp3, then sample clips from both files),  and probably error prone.  What do you think?

T

-V 0 vs. -b 320 vs. wav

Reply #3 – 2007-09-13 09:49:28

If you want to be sure you maintain the length and start/end points, you need to use an mp3 encoder and decoder that's aware of delay, padding, and accurate length metadata. Otherwise you'll see the extra samples which are inevitably added to the start and end of the track by the mp3 encoding/decoding process.

You won't be surprised to hear that lame is OK to use for encoding, and foobar2k is OK to use for decoding. Lame is also OK to use for decoding.

Cheers,
David.