Quote:
Originally Posted by majkel /img/forum/go_quote.gif
This is the source of Feb's 99%, not that people don't hear and lie. This is a rude accusation, kind of "I don't believe you own the Orpheus - post a picture with yourself standing aside".
|
Again, you misunderstand or deliberately misconstrue my point. I am not accusing people of lying. (Though there are a fair number of people who simply parrot things that they have read here without ever conducting listening tests for themselves). For the most part, I believe that people are honestly reporting what they have heard. However, there are many explanations for what they hear that, in my view, make their generalized assertions invalid. For example:
--People base their conclusions and their advice on files of questionable origin. "I downloaded a 320kbps file and compared it to .wav, and the difference was night and day. The MP3 sounded like crap." The file is of unknown origin. For all we know, it may have been transcoded from a 64kbps file. These results are not generalizable.
--People base their conclusions on tests that they did with poorly-encoded files. "Back in 1999, I listened to an 128kbps MP3 that I made with MusicMatch, and it was awful. 128kbps MP3s suck." Obviously, the files they are using for their comparison are not representative of all MP3 files.
--People use unconventional encoder settings. "The LAME presets suck. I used these settings that I downloaded from some random website." One very common mistake that I've seen in this regard is people forcing the encoder to use a very high lowpass setting, which forces the encoder to waste bits for inaudible high-frequency sounds rather than using those bits for the midrange frequencies where they are actually needed.
--People make comparisons without volume matching the files. Regardless of our respective positions on ABX testing, etc., we should all be able to agree that differences in volume can be perceived as differences in quality.
--People can be influenced by their expectations when they do sighted-tests. I hate using the word "placebo" because it is so often misused here and carries such a stigma, but expectation biases are real, no matter how much people want to deny them.
Using a good encoder and an ABX utility can eliminate all of these issues. Given that every good ABX utility lets you listen to as much of a song as you want to, as many times as you want to, I believe that your criticisms of the ABX test in this context are unfounded. I believe that if you can't hear a difference between two tracks after listening to the entire track as many times as you'd like to, then no audible difference exists.
I recognize that you don't agree with me, and we can agree to disagree, but please recognize that am I not accusing you or anyone else of lying.
Quote:
| All non-believers - stop accusing, start listening. Full midrange reproduction begins in the 320kb/s bitrate of the mp3 format. Try it and believe it. |
I have engaged in extensive listening tests, and my results are different from yours. I have found through personal experience that when (a) I use a good encoder, (b) I volume-match files, and (c) I listen under non-sighted conditions, I cannot distinguish differences that I previously believed to exist.
