@sleepwalker65 said:
People who argue that a discrete sampling protocol can never record and reconstruct continuous audio do not fully understand the nature of sound. Sounds are made up of waves. Sound waves are just sine wave compressions and rarefactions of various frequencies at various amplitudes. Granted, in a musical performance, there are a lot of frequencies and amplitudes, but they are just a lot of sine waves. If you apply a Fourier transform to an audio recording, you can decompose all of that noise into a collection of sine waves that have equations to describe their behavior. Because sine waves have mathematically regular behavior, once you know the frequency and the amplitude, you know everything you need. There isn’t any "information left out during the digitizing."
Since humans aren’t bats or dogs, we are only interested in waves we can hear, and for the absolute best of us, that spans the range from 20Hz to 20,000Hz. For most of us middle-agers and above, a more likely range is 20Hz to 12,000Hz. Nyquist-Shannon says that a filter can be constructed to accurately reconstruct a waveform from a stream of discrete samples if the sampling rate is 2 times the desired highest frequency. This isn’t magic, it is math. And it is why the CD Redbook standard of 16/44.1 is all that is really needed. Every rigorous listening test I have ever seen has shown that for the vast majority of listeners, even trained and professional listeners, it is very very difficult to tell the difference between a 16/44.1 recording and a 24/96 or 24/192 one.
Fourier analysis has been around since 1822, the Nyquist theorem was formulated in 1933, and Shannon’s extension was published in 1948. So these aren’t new ideas. Until the advent of the digital computer, applying these ideas to audio wasn’t practical. But since 1982, it has been practical, and over the last 36 years the technology has been refined.
If people prefer the sound of vinyl LPs to CDs, that’s fine. But they should stop justifying their preference by perpetrating myths.
the fact is that while the data is still represented as a staircase, (lollipop diagram if you prefer), the nature of sampling means that some information is left out during digitizing and then on playback, it is artificially synthesized. That is the indisputable flaw in the process. No matter what quantization resolution and what sampling rate, you cannot escape this fundamental.This is a oft repeated and persistent argument against digital audio, but it is a myth. It stems from a lack of understanding of sound, Fourier transforms, and the Nyquist-Shannon theorem.
People who argue that a discrete sampling protocol can never record and reconstruct continuous audio do not fully understand the nature of sound. Sounds are made up of waves. Sound waves are just sine wave compressions and rarefactions of various frequencies at various amplitudes. Granted, in a musical performance, there are a lot of frequencies and amplitudes, but they are just a lot of sine waves. If you apply a Fourier transform to an audio recording, you can decompose all of that noise into a collection of sine waves that have equations to describe their behavior. Because sine waves have mathematically regular behavior, once you know the frequency and the amplitude, you know everything you need. There isn’t any "information left out during the digitizing."
Since humans aren’t bats or dogs, we are only interested in waves we can hear, and for the absolute best of us, that spans the range from 20Hz to 20,000Hz. For most of us middle-agers and above, a more likely range is 20Hz to 12,000Hz. Nyquist-Shannon says that a filter can be constructed to accurately reconstruct a waveform from a stream of discrete samples if the sampling rate is 2 times the desired highest frequency. This isn’t magic, it is math. And it is why the CD Redbook standard of 16/44.1 is all that is really needed. Every rigorous listening test I have ever seen has shown that for the vast majority of listeners, even trained and professional listeners, it is very very difficult to tell the difference between a 16/44.1 recording and a 24/96 or 24/192 one.
Fourier analysis has been around since 1822, the Nyquist theorem was formulated in 1933, and Shannon’s extension was published in 1948. So these aren’t new ideas. Until the advent of the digital computer, applying these ideas to audio wasn’t practical. But since 1982, it has been practical, and over the last 36 years the technology has been refined.
If people prefer the sound of vinyl LPs to CDs, that’s fine. But they should stop justifying their preference by perpetrating myths.