Cool Edit Pro setting

So I finally figured out what was causing all manner of vagueries in my home copy of Cool Edit Pro. If I took a mono file, converted it to stereo, and then null-tested the channels, they would cancel, but I would get an odd, constant noise at -90dB or so. This would happen WHENEVER I would use any transform. If I normalized silence, I'd get that -90db noise. I couldn't figure out what was causing it.

Until today, that is. A little setting called "Dither transform results" was the culprit.

Now, I assume Dither Transform Results is a GOOD thing, as it apparently increases dynamic range when doing transforms in 16 bit (which is occasionally necessary). Why would this cause the side effect of -90DB noise, though, even when the transformed part of the wave was completely silent? I hesitate to keep this enabled now, even though it's probably a good thing in general.

Is my copy of CEP crapping out, or is there something bigger at work here?

 

Dither IS noise, and is necessary for rounding out the digital word length to reduce truncation distortion.

Anytime you do something to a file, you lengthen the word. But, if you work in 16-bit, there isn't enough room for that longer word, so the software must round off the numbers to keep it all in that 16-bit bucket. Dither rounds off those bits smoothly. 24-bit provides a larger bucket that gives much more room for those longer words. You can indeed get away with processing a 16-bit file IF you keep your transformations short, and few. You may not notice much, if any real difference in sound.

However, the best course of action, if you are going to do more than just simple editing, if possible, is to convert that 16-bit file to 32-bit first, then process to your heart's content. This is especially important if you plan on making gain changes. Gain changes requires the software to recalculate the whole file, which produces an enormous word length. It is better done in 32-bit. Then, when done processing, convert back down to 16-bit WITHOUT dither. Why? because once you dither something, you do not want to do it a second time. Double dithering produces worse results than not doing it at all. You figure most CDs were dithered already, so if you rip tracks from CDs and process them, don't dither it again! I learned this from Bob Katz writings, and he's right. I've accidentally dithered things twice, and the results were that of a veiled, slightly smeared sound. But, when I used his suggestion, things were acceptable. I lost a bit of depth, but still retained the crispness and stereo seperation of the unprocessed file.

But, David, keep that Dither Transforms enabled! What I have found, is that a dither amount of .07 is a good balance between distortion and smoothness when dithering from 32/24-bit to 16-bit. The software comes with the dither set at a default of 1 bit. That is too much for me. The older versions used .05. That is too little for me.

BTW, you want to take care of that noise you hear? That's partly why you can use noise shaping, to distribute that noise to parts of the frequency spectrum where it will be masked by the music. You have to experiment with what works best for you and the type of music you are working with, but I find that the C2 curve with a Triangular p.d.f. (probability of distribution factor) works best for me. Once you introduce dither, you can't get rid of it. Like tape hiss, dither noise is our friend.

I wouldn't worry about the silence. It's not going to be noticeable on most playback in most circumstances.