|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
76.1.138.16
In Reply to: Comparison of a Raw Analog Audio signal to the CD audio and DVD audio output posted by Vinyleer on September 13, 2006 at 18:59:58:
This graph accurately reports what is MEASURED with a CD or DVD. This is simply the nature of sampled systems. Keeping in mind that at 10kHz, real audio signals are much smaller than they are at 20 Hz, which is why the CD has only a few bits.Now, the OUTPUT of a CD or DVD is very different than this, but it is also different than the original analog signal, UNLESS it is a very long, continuouse sine wave (not music). In that case the nyquest rules do apply and the analog signal is accurately reproduced.
I must agree that this doesn't really completely explain why vinyl is better than CDs/DVDs (the answer is also in the D/A conversion problem), but it is a useful, but over simplified, way to show the layman that the CD and DVD really don't truely capture the original signal.
Follow Ups:
"UNLESS it is a very long, continuouse sine wave (not music)"That is also blatantly wrong.
There is no requirement for a steady-state signal, much less a sine, in the whole Nyquist-Shannon-Kotelnikow shebang. You may want to dig out
Shannon's formulation of the sampling theorem and his proof. The proof is built entirely on the Fourier transform, which does not preclude transient/finite-time signals as it transforms signals to continuous spectra. (Perhaps you confuse with the Fourier series, which transforms periodic signals to a discrete spectrum.)
Hmm... so you are saying that phase and all is accurate with a single sine fragment? I don't think so. There simply is not enough information, especially at the limit of only two samples per cycle (at 20kHz). A key assumption in the formulation is that it is a repeating signal for some period of time. Such is the nature of frequency domain thinking.Sorry to disagree, but that's what this forum is for!
Phase and all is accurate within the constraints imposed by the Theorem, i.e. that the signal to be sampled has its Fourier transform confined to a band with the width of fs/2.The latter implies that after band-limiting to fs/2 the signal-to-be-sampled cannot have any fast changes anymore close to fs/2. This is the nature of band-limited signals, and has nothing to do with the sampling theorem per se. If you feel this is a problem then all you have to do is to increase fs.
---
I have said it many times before: if there is something wrong with digital audio then it is not because of the sampling theorem being wrong or partly-true (it is correct), but because there is no clue given, let alone a magic bullet(*), to make the band-limiting process painless, given the lowish sample rate of CD. And this gave the recording (gear) industry an opportunity to screw things up bigtime.
(* In fact there are copper bullets, if not silver ones, but too few people are aware of them. Keith Johnson was one of the first who understood this, I think.)
This post is made possible by the generous support of people like you and our sponsors: