Yup, digital transmissions do allow more stuff to be packed into the same piece of spectrum. If an analog signal is to be broadcast at a certain frequency, you needed to have those dead air guard bands above and below because it's hard to build a radio circuit that can recover full signal at one frequency but totally reject the ones around it. Hard to build an analog filter like that that doesn't compromise the (analog) signal.
By contrast, it's easier to build a digital filter that has very sharp rejection. Since the signal is digital (and can have error-correction), you can work around interference pretty effectively, within the constraints of the data format you've imposed on the signal.
If you're old enough, you may remember back to the original car phones, which were very similar to a police radio system. There were a few frequency slots which the car phones would pick from automatically. I had a friend who had one of these in the early 80's (in his DeLorean, no less). At that time, in the Silicon Valley, there could be a grand total of 20 mobile phone calls happening simultaneously before you got some sort of special busy signal. A little different than today, where you can support thousands of digital mobile signals at any street corner. You can imagine how full the spectrum is now, and how hard it is for old-school electronics to operate properly.
There's something to that analog warmth in CDs, certainly the case when they were new (didn't the 25th anniversary of the CD just come?). In my earlier post, I blabbed a bit about the CD audio data format (16-bits/signal, 2 channels @ 44KHz sample rate, uncompressed). The sample rate was set at 44KHz because this should allow proper reconstruction of frequencies up to 22KHz from the digital signal. That's higher than most people can hear, so this should be great.
But in practice, as always, things aren't always that simple. You can recover a perfect digital stream, but when you convert that to analog audio, you run into those darn imperfections in the translation process again, and those have audible side effects.
If you were analyzing the volume level of a pure sine wave in analog, you'd see a smooth, continuous change in voltage, up and down. If you digitize that same signal and analyze that, you'll see the voltage change in discrete steps at the sampling rate, hardly identical. Now, if you have the sampling rate high enough and the steps small enough, you shouldn't notice the difference. But in the early CD players, there were a lot of problems in conversion to analog which passed through a lot of errors induce in the sampling process.
This is a quantization or aliasing error. You can see them on the computer screen that you're looking at now, in the form of jaggies on diagonal lines and in the fine features of your fonts. The sample frequency of your computer's screen (something like 80-100MHz delivered at 96 dots/inch) isn't sufficient to deliver full fidelity of things like diagonal lines or curves to your eyes. This same problem is happening to your ears when you recover digital audio - there's audio jaggies occuring, which your brain will interpret as a cloud of high-frequency fuzz.
The best way to fix this problem is to radically increase the sample rate and resolution. Paper is imaged digitally and curves, diagonals, and text don't seem to have a problem. This is because the quantization has been selected to be way out of most people's range of perception (paper is 2500dpi instead of 96dpi). Although that works for imaging paper, it works poorly if you're going to transmit digital data, because now you have much, much more data to move - it will take longer to send and cost more to store.
The other solution that's used to address the jaggies is called anti-aliasing, which is a computational blurring of the signal which should lessen the effects of insufficient sample rate. On your computer screen, instead of calculating which black and white dots to turn on to draw a diagonal line, you draw levels of gray to try to soften the edges and make it look like there was more data or resolution. In fact, there's actually less data accuracy than before. This is largely used on screen fonts to make them less bumpy looking. This is also done on all digital audio converters now - instead of reproducing steps in output levels, the steps are smoothed over with guesses at what they might have been in the original analog signal.
Hardly a good solution, but for 99.99% of the population, a solution which is good enough. It's hard to build a good LP record player, and easier to build a pretty good CD player. The resolution of CDs might be lower than optimal, but it made it possible to have convenient and durable CDs, plus you could have the ability to easily hop tracks.
I like the utility of digital audio as well as the next guy, and I guess I'm just used to the antialiasing now too. The compressed digital audio forms like mp3 are losing even more data, but perhaps the additional utility is worth it as well (easier to carry an iPod than a CD Walkman with 300 CDs!).
David Fung
