Ask an Audio geek

[IATIA]

Here we go.  My argument is that one cannot exactly digitize analog signal.  It is not that one cannot digitize an acceptable representation of an analog signal.

I also stated that the best one could do is to digitize 50% of the actual signal. Actually, I said that 50% was the best that could be done, but unachievable.

Here we sample a 1KHz signal at 1KHz - Flat line

---

---

Here we sample a 1KHz signal at 2KHz - Still a flat line

---

---

Here we sample a 1KHz signal at 4KHz - now we have a sawtooth wave

---

---

Here we sample a 1KHz signal at 8KHz - Now it is starting to resemble a sine wave

---

---

Here we sample a 1KHz signal at 16KHz - We now have a ‘reasonable’ facsimile of a sine wave

---

---

Now the sample point.  Do we sample on the rising edge, the falling edge or take an average reading?
None will get the actual peak

---

---

Here is a 4KHz and 20KHz sine wave sampled at 64KHz

---

---

We can keep zooming in and, just like a Mandelbrot Bug, keep finding more artifacts. The upper limit for audio frequency at STP is around 3GHz.  At that point, the molecules are too far apart to transfer the sound.

This ‘magical’ math that compensates for the information between the samples is pure guess work.  Maybe there is a pulse or a dip or it just follows the apparent path.

I have shown that a 1KHz signal needs a 16KHz sampling rate which means a 20KHz signal will need a 320KHz sampling rate and still does not cover anything in the 40KHz harmonics.

If we have an infinitesimally narrow sampling point at an infinitesimally high rate, we can still only check every other point.

So, I state it again.  One can get a reasonably accurate and acceptable digital representation of an analog signal that one would have difficulty at best to prove it was not an analog signal, but it IS NOT AN EXACT COPY of the analog signal and can never be.

(And there are significantly better ways to sample an analog signal then what is being used in CD/DVD technology today.)