OK, let me remind those that find this post that I’m not interested in debate. Wait, there are exceptions: If you have an MS in Electrical Engineering and are (or have been) a working engineer (like me), then I’m kind of interested, particularly if you understand digital signal processing in any kind of detail.
Ah but then again it might not matter, because I trust my ears. I know when I hear something and I also know when the difference is so small I can not qualitative say that the difference I am hearing isn’t actually do to a change in attentiveness in myself.
But 44kHz and 16 bits is not one of those things. I know this because I have several recordings in both CD as well as SACD or DVD-A format, and the difference can be large.
But let’s go back to some of the technical reasons, and my ATTACK! on the anti-audiophile. The first thing to keep in mind is what the Nyquist theorem does and does not say. Nyquist (which is a mathamatical theorem, by the way), basically says that if I want to capture an audio-like signal (consisting of a combination of sinusoids according to Fourier theory), then I need to “sample” at twice the highest frequency of the signal.
The first caveat is that basic Nyquist does not refer to digital sampling per se. Nyquist assumes that your samples can be any height. When we now quantize our samples as in a CD signal, this now means that our samples can’t be any height at all, they can only be a height allowed by our digitization scheme. What this now means is that we can no longer perfectly reconstruct the original audio signal.
Now the argument from the anti-audiophiles is that 16 bits is enough: 16 bits gives you 2-to-the-16th-power number of sample heights, and the ear isn’t sensitivie to any more than that.
This is true and untrue. Actually, it’s completely untrue. All other things being equal, most studies claim that the ear can hear perhaps 17 or 18 bits. The anti-audiophile will argue that this is more than enough for recorded music, and one might concede that point.
But the important aspect of this is that, as your CD player attempts to recreate the audio signal, it stupidly just builds waveforms out of the quantized samples. Since these samples no longer correspond to the original waveform, what your audio system in effect does is introduce “quantization noise” as this new non-perfect waveform is no longer sinusoidal! And a waveform that is no longer sinusoidal not only has leeched energy from the original perfect tone, it has generated a host of crappy little parastici tones that have no musical relationship to the original. And now, multiply this noise by every single tone in your music and what you have is life-eating crud.
this is most dramatically seen in the residual “ringing” in a square wave, otherwise known as Gibbs phenomenon: Put in a perfect square wave into your CD sampler and out comes a square wave that is no longer square but where the horizontal parts are shaky. These shaky parts are in themselves audible and have no musical relationship to the original signal. This is one reason why CDs tend to fatigue your ears and sound.
All of this, of course, assumes that you have perfected everything else in the recording and mastering process. The moment we add imperfections such as jitter, things get bad quickly. In fact, go listen to early digital recordings: They’re glassy, harsh, and unconvincing-sounding. In other words, they sound like crap. The nature of the CD format is such that there is little or no headroom for such imperfections; they are immediately audible.
Another problem with CD is that the format has removed some of the more arcane but important supersonic queues that our ears are apparently aware of. Ah, you say, “supersonic” is by definition not audible. It therefore doesn’t matter. But to that this audiophile would respond, “don’t argue with me but get your own papers published that contradict current scientific knowledge”.
But the basic idea here is that our ears and brain can infer some supersonic aspects of an audio event based on certain very subtle but theoretically audible queues. To give an example, with many media two high frequency non-audible pitches can create sum and difference frequencies which are themselves audible. And then, somehow, our brain is apparently able to reconstruct some of the supersonic events that allow a recording to sound far more realistic. CD, however, eliminates this possibility due to the necessary anti-aliasing filters.
But none of this matters. The proof is in the pudding and the difference between CD and high-rez formats is quite audible (to the point to where my non-audiophile wife, who doesn’t even like music asked if the CD and SACD version of the live Kraftwerk album were the same recording).
Now don’t get me wrong, CD can get VERY good when everything is the recording chain is handled perfectly (having 24 bit recording technology helps, of course). Take Raising Sand on CD and put it on your high-end rig. WOW! It really sounds very good. But then again, the new chipsets are actually upsampling to 24 bits and then using some DSP magic to guess what those waveforms are supposed to be.
More later…perhaps I’ll take on why cables matter, and they do.