OK some weren\'t satisfied with the piece used for this test so I did something more SPARSE. Again, 4 files, all of them now in 24-bit wav, but derived from wav\'s and MP3\'s of different bits/resolution.
One is the original un-processed 24-bit render with QLSO. One is the same wav, but dithered to 16-bit. One is an MP3 at 192kbps, the last is an MP3 at 320kbps. Post your results here or to email@example.com
Simon, this is a good torture test for the algorithm, and I like the little piece. The tutti \"ba ba ba, ba ba ba, crescendo\" figure is smooth in A4, and somewhat less so in the rest, especially when the brass buzz comes up. There is just more transparency in A4 all around...you hear the separation all the way through. The lower res files don\'t show this off like the original.
At the very top in A4, the upper pizz voices\' exciting of the room is smoother, too, and one tail doesn\'t go away as the subsequent plucks sound.
The emptier the soundstage becomes, the more differece you can discern. The best example for people to hear the processes lies at the 21 second mark, just past the last piccolo figure. Put the cursor right at 21 seconds and play each one out. You\'ll hear the tail die very nicely in A4 (right up to a little strange artifact). In the other three, you hear the quantizing noise whooshing over the final portion of the tail.
The mp3 encoder you\'re using does sound really good, and the exercise makes a great case for deriving your mp3s from 24 rather than 16-bit masters. It did a surprisingly good job of matching CD-quality on difficult material, sufficient to make me throw up my hands and not bother picking the pepper out of the gnat poop to find the lower res PCM.
I\'m curious what you used to quantize the 16-bit PCM, because I\'m surprised that the quantization noise on all three of the lower res versions is pretty much equal. That either speaks very well of the mp3 encoder or less well of the quantization process. I can\'t really distinguish much audible difference between them. That surprised me, not so much that I can\'t get the difference between the two mp3s, but that the 16-bit PCM doesn\'t distinguish itself more. It also may have been looking ahead and getting excited by that little artifact on the end, but that doesn\'t seem to be the case. Weird. Or good, take your pick. Also there\'s a teeny tiny pop in the release of one of those samples, which you can hear in all four. I thought I\'d be able to listen to that repeatedly and figure it out, but no. I guess I\'ve stood in front of too many drummers.
I may be totally wrong, but, in the spirit of reconciliation, I\'ll be glad to hang myself out to dry.
The only point I ever intended to make is that while mp3 is a really good thing, it does have to make decisions about what it thinks you hear vs. what it thinks you don\'t. And that many mix decisions are so context-driven that it\'s sometimes haphazard to second guess them out of that context (or at least to imply that the author called it wrong, when his decisions may well be picture perfect for the context). That somehow got a little lost in the piss and vinegar.
I love that piece you did there, really expressive. I love that multi-million dollar reverb too! If I keep listening I might change my mind [img]images/icons/smile.gif[/img] but here\'s what I think:
1. 24 bit
2. Lowest mp3 (192kb?)
3. highest mp3(360kb?)
4. 16 bit
I think that number 2 sounds a wee bit compressed.
Hey, you keep writing like that with QLSO and we\'ll have to keep challenging you to do more tests!