high BIT RATES - Don't believe it ! - Page 3

RedSmurf · 05:49 PM

Quote:

Originally Posted by squash54

GM raised an interesting point about the "upsampling" technology being used in some CD players and noted that there is not much out there about whether this can be applied to mp3's. I am hoping to get some feed-back about this issue from the Swiss company that provides this technology to CD player manufacturers. I will post anything I learn here.

Let me assure you: once a signal has been compressed into MP3 or another lossy format, it is permanently distorted. You can't remove this distortion and get the original signal back. Otherwise MP3 would be a lossless format. Basic information theory says that when you take a string of 88200 numbers and convert them in some way to 16000 numbers, you cannot restore the original 88200 numbers unless they contain enough redundancy so that their pattern can be described in 16000 numbers or less.
A friend once told me that he had the greatest compression technology ever. "If you send the same file through 'zip' over and over, it will become smaller each time and eventually you'd end up with only 2 bytes!" It's easy to see the flaw in this reasoning: if this were true, there would only be 65536 possible files in the entire universe. If you want to know more about information theory, look for "Shannon" on Wikipedia.

The only thing you could do is try to detect typical MP3 artefacts ('ringing' and 'glassy' sounds) and filter them out (which would require complex signal processing). However, there is no doubt that this will also remove sounds that were originally in the music. It's the same as with noise removal on images: you can try to filter out noise, but it will also remove features of the image itself. Often it's better to just live with the noise. Moreover, MP3 doesn't just introduce noise and artefacts, at low bitrates it will blatantly throw away frequencies.

The upsampling technology in CD players will not improve sound compared to a non-upsampling CD player that conforms to the same specifications. The human ear can only hear up to ±20kHz and a sound sampled at 44.1kHz can perfectly represent any signal that has frequency content up to 22.05kHz (Nyquist's theorem). The only advantage of upsampling is that it becomes easier to design good D/A convertors, because the margin for the filters required inside the convertors becomes much larger. In that aspect, an upsampling D/A convertor may produce a more faithful reproduction of a non-upsampling D/A convertor that is unable to reach the required specs due to technological limitations, but once you go above a certain threshold there is no audible difference anymore.

The only way to be sure about the quality is to start from actual CD audio and encode it at a sufficiently high bitrate. That's why I still prefer to buy CDs if they are reasonably priced