You know it's funny, I just re-read the original article there....when I first "skimmed it" I was only annoyed by his first claim, but then thought the rest of his comments were basically just regarding jitter issues. After Ethan's post, though, I read the article in its entirety, and realized how dangerous some of the statements are.
1) "Here's proof. Computers crash. Hard drives lose data. Files get fragmented. Norton and other utility programs are popular ways of "recovering" corrupted files. If all the one's and zeros worked perfectly every time, none of these problems would occur, since each file would do exactly as it was supposed to do each and every time."
This type of corruption is extremely rare, *and* you're going to know about it. Analog recording mediums like tape are as or more susceptible to degredation and failure, and in ways that you might not notice until it's too late.
2) "Even transferring a sound file from one hard drive to another changes the sound slightly. Never mind "saving" files to cdr or dat backup and then reloading back into a hard disk system. With all this technology available, musicians still have to balance (1) the amount of time it takes to learn these details (2) their time and budget and (3) the craft and passion of their music."
This statement is completely absurd and very dangerous. Data is not written to a hard drive or buffer with any regard to timing or magnitude. *Data* written to a CD is the same as data written to a hard drive or memory. If an error occurs, you're going to know about it. (this is not necessarily true when we talk about audio cd's, that's a different story)
3) "Whenever you change eq, compress or change levels in a hard disk system, you are making calculations in the computer. The actual word length is changing and the original "analysis" made by the A to D converters is being changed. The sound is being altered, and generally this means that some resolution of the tone is lost. This translates into a harder, shallower, less detailed sound. Smoothness, width, depth, and openness are some of the first qualities to go whenever these calculations are made."
Notice the complete lack of any quantifiable terminology? "Smoothness, width, depth, openness" blah. Yes, DSP implies performing calculations on digital audio. Yes, the sample values will be changed. If you want to think of this as "the original analysis made by the A to D converters is being changed", then so be it, even though technically the original analysis isn't being changed (duh).
He throws word length in there as though that in of itself is the problem. Yes, when you do DSP the word length is usually changing, but in the POSITIVE direction, and then the results of the DSP are dithered back down to the original word length. This process allows the DSP calculations greater accuracy, and doesn't somehow harm the sound by itself. The simple fact of the matter is that it IS POSSIBLE to have a digital processor that is the audible equivalent of an analog processor in the digital realm. In other words, if a digital processor can produce the same exact data file as an analog processor through a converter, then it's going to sound exactly the same. This doesn't mean that digital processing is just as good as analog though, it just means it has potential. Some things will sound better than others. Wow, there's a totally new concept.
4) "Dat machines and hard disk systems have moving components, moving tape etc. that are subject to errors, especially when handling huge sound files. Cable problems and word clock issues can cause jitter which affects solidity and resolution too."
A purely digital copy of a DAT or HD will not result in a degraded sound. Jitter is not copied. File size does not increase the chance of an error occuring, just the chance that an error will occur IN that particular large file. Not the same thing by a longshot.
5) "If you normalize your song/file, your computer is going through calculations that change the one's and zeros of your original signal. Huge amounts of level change will add some hardness and shrink the "sound stage"."
The "calculations" that the computer is going through when normalizing is a crazy little process called "addition". If a constant value is added to each sample in a file, then mathematically the waveform of the normalized file will be identical to the original, just with greater amplitude. In fact, the original samples can be "recovered" by simply subracting the constant value.
Anyhow, as I said earlier...digital audio can sound bad. Digital processing can sound bad. Analog audio can sound bad. Analog processing can sound bad. When you move a fader in your DAW software, you are multiplying a constant value against the sample stream. Any time you multiply, resolution does become an issue (e.g. what's 1/3? is it .3, .33, .333, .3333)....but so what. Anytime you move an analog fader you're relying on imprecise electronics to change the signal....so what.
I'm just an amature recording wannabe. Some of the things I say might not be entirely correct. However, I have learned one very important thing....if you can't hear it, it doesn't exist. It's best to gain as much listening experience as possible before you go head first into a bunch of technical mumbo jumbo. Eventually the techno mumbo jumbo will make sense, because you'll require it to quantify what you're hearing.
Slackmaster 2000
