Over on the "Your Old Vinyl Collection" thread I posted a link to an article, entitled "Vinyl vs. CD Myths refuse to die." It's worth reading. In it the author links to ANOTHER article, "Is the sound on vinyl records better than on CDs or DVDs?", which is ALSO worth reading. The author of the first article says, "The content on that [second] page is, frankly, tripe. On a quick scan I counted six outright errors - mostly uninformed opinion being expressed as fact." Then he goes on to say, "Perhaps at some point I'll create a post here specifically refuting the technical errors on that page". A quick search of Rich Pell's articles on EE Times since January 2010 however does not show that he did.
No matter. The refutation is really easy. The article nicely summarizes the surprisingly prevalent misconceptions about digital audio, so I think it might be helpful to an informed discussion to point out exactly what the unnamed author of that undated and unreferenced article got exactly wrong.
It can be expressed in three letters: DAC.
To digitize a signal, you run it through an ADC, an analog-to-digital converter. This slices up the incoming analog waves into 44,000 slices per second based on 65,536 possible values. This part the article got exactly right.
The results are shown on the graph (pink line) overlaying the 10,000 Hz source sine wave (black line) and the graph correctly shows 4.4 samples.
Pretty rough approximation, you have to agree! It looks like a big ugly square wave -- not the nice smooth 10,000 Hz sine wave it is supposed to be reproducing.
If you've ever accidentally hooked up a speaker to a raw digital output (as I have), that's EXACTLY what you hear: ugly screeching square waves!
What the graph fails to point out, and what the author failed to include, and what a lot of people misunderstand, is that we don't listen to this square wave output. The raw digital bitstream must be first run through a DAC, digital-to-analog converter, to turn it back into sine waves so we can listen to it. Once this is done, something very similar to the original waveform comes out.
How similar? Well you might say "There's a lot of missing information in that square wave approximation" (if fact the article's author does say this) but you have to remember two things: 1) the DAC's job is to smooth out the square wave jumps in level (all 65,536 of them) and it uses the same process, in reverse, that the ADC used to create that square wave. Therefore with a matched ADC/DAC pair the end result is very, very close. 2) The example shown in the graph is a 10,000 Hz sine wave. We don't listen to 10,000 Hz sine waves. In fact the frequency range of even the highest percussion instruments, like a triangle, is only 3-5000 Hz. There are overtones and resonances in the 10,000 Hz range, but no fundamentals.
And remember, below 10,000 Hz you're getting a lot more than 4.4 samples per wavecycle. 8.8 for 5,000 Hz, 17.6 for 2,500 Hz, 35.2 for 1,250 HZ and so forth.
So do you lose "overtones and resonances" with a 44 Khz sampling frequency? Not with a good DAC, and not according to Nyquist theorum. If you aren't familiar with Nyquist it's worth Googling.
Going back to the graph above, would sampling at 192,000 times per second (red line) produce a more accurate result? Well yes -- if you were listening to the square wave output. But if you're using a DAC, the answer is, "Maybe yes, maybe no. It all depends on the DAC." Certainly (and proven in numerous double-blind tests) the human ear CANNOT HEAR THE DIFFERENCE under normal circumstances. That's the cold hard truth.
And that is why products that tout "super fidelity" are mostly blowing smoke up your ass (...to get closer to your wallet without being seen).
And why people who say "analog LPs produce a more accurate representation of the original analog signal" are full of shit.
Bookmarks