It’s a nostalgic bias. Uncompressed digitally recorded sound is superior in every measurable way to vinyl records.
A common misconception is that the process of sampling the signal at finite time intervals means that all the signal is not captured, and therefore the reproduced sound is an approximation. This simply is not true, and only spruiked by those who don’t understand the maths. As long as a signal is recorded digitally with adequate resolution, then it can be reproduced precisely. No distortion. No approximation. Perfectly.
The original CD format of 16 bit, 44.1KHz sampling rate is more than adequate for human hearing. But, and this is a big but, the incoming signal MUST be filtered so that nothing above half the sampling frequency appears at the input of the analog to digital process. This is where problems start to occur, and a big part of CDs initial failure to impress audiophiles.
If any frequencies above half the sampling rate (also referred to as the Nyquist frequency) are digitised, they will be reproduced erroneously as frequency components lower than the Nyquist frequency. This effect is known as aliasing. In the early days of digital audio, equipment used sharp rolloff analog filters to remove everything above the Nyquist frequency. CD was only ever intended to reproduce up to 20KHz, so the filters started rolling off steeply at that point with the hope that attenuation at 22.05KHz was adequate to make any surviving aliased sounds inaudible. Human hearing “masks” sounds of a certain level threshold below ambient (both frequency and time dependent), effectively allowing unwanted noise to be hidden if it cannot effectively be removed.
These sharp rolloff anti-aliasing filters however caused other problems. Analog filters, by their nature, introduce phase distortion, and the steeper their rolloff the more phase distortion is added. This distortion may or may not be audible, but nonetheless it’s a valid concern, and it’s measurable. Digitised music created with analog anti-aliasing filters will exhibit phase distortion, particularly at the upper end of the frequency range. Further, depending on the source material there may still be aliasing artifacts in the reproduced sound.
This is where the concept of oversampling steps in. It’s a game changer. By sampling the audio at a higher rate, like two, four or more times, we can greatly reduce the rolloff required of the anti-aliasing filter. Instead of a rolloff covering 20KHz to 22.05KHz, it can now gently cover say 20KHz to 88.2KHz. Over this range, the filter’s attenuation at the Nyquist frequency can bring aliasing below the noise floor, while introducing barely measurable phase distortion that only occurs above the target frequency range anyway. We now have digitised a perfectly faithful version of the original signal, albeit at way too high a resolution for the final medium.
But it’s digital, and can now be processed by maths. It’s now a trivial task to digitally remove everything above the target frequency range and resample at the rate of the final medium. Digital filters don’t introduce distortion, they simply remove what it not needed. The final product is a series of numbers that when processed mathematically will plot to a curve that precisely matches the original input signal up to the intended frequency limit.
Now we come to the equipment whose purpose is to turn these numbers back into sound, and exactly the same issues arise, in reverse. The output of a digital to analog convertor, unprocessed, steps sharply between each voltage level. We need to apply an analog filter to remove the high (inaudible but still problematic) frequency components of this step shaped waveform. Guess what - phase distortion! Oversampling again comes to the rescue. Using maths on the numbers, we can increase the sampling rate as much as needed to bring this quantization noise to a sufficiently high frequency that the filter slope no longer causes significant phase distortion.
We now have an end to end system that can perfectly reproduce musical signals, but the early days of digital music were far from this and the lingering arguments about last millennium’s digital audio technology still haunt us.
We’re also now dealing with audio engineers that want music to be continuously loud, no room for dynamics in their mixes, and the horror of low rate digitally compressed audio (128kbps MP3, I’m looking at you). Digital music has never had the chance, at least in mainstream music, to shine nearly as brightly as it can, but it’s not the fault of the medium.
Why do people love vinyl so much? Because the audio engineers involved in the mastering process cared about the medium, understood its weaknesses, and strived to get the best they possibly could from it. Digital hasn’t had the chance to be nurtured like this but don’t be fooled, it’s actually capable of faithfully recreating the vinyl experience if we want it to—if only from the aural perspective. But who really wants the imperfections of vinyl in their music? Nostalgia buffs do.