24/192 digital audio format and why it doesn’t make sense. Part 1
Unfortunately, there is no point in recording music 24/192. Its fidelity does not dramatically exceed 16/44 or 16/48 formats, but it takes up 6 times more space.
Save and read later –
Earlier headlines reported that musician Neil Young and Apple founder Steve Jobs were discussing a possible launch of a service to download “uncompromising studio quality” music formats. Most of the newspapers, magazines and users were quite optimistic about the prospects of a digital music format with signal quantization in 24 bits, at a sampling frequency of 192 kHz.
Unfortunately, there is no point in recording music 24/192. Its fidelity does not dramatically exceed 16/44 or 16/48 formats, but it takes up 6 times more space.
Today, there are several problems associated with audio quality and the “application” of digital music distribution. The 24/192 format does not resolve any of them. As long as everyone regards this format as a panacea, we will not see any improvement in the field of music.
Let’s start with the bad news
Over the past few weeks, I have talked to smart, scientific people who believe in the 24/192 music format and don’t understand how anyone can disagree with it. They asked good questions that are worth answering in detail.
I also wondered what could be causing such active support for high sample rate digital audio. The responses showed that few people understand the basics of signal theory or the sampling theorem (the Kotelnikov or Nyquist-Shannon theorem), which is not surprising. Misunderstandings about mathematics, technology and physiology were evident in the speeches of many professionals with extensive experience in audio technology. Some have even argued that Kotelnikov’s theorem does not explain how digital audio works [1].
Disinformation and prejudice only play in the hands of charlatans. Let’s go through the basics of why the 24/192 format doesn’t make sense before presenting other more valid ideas.
Gentlemen, welcome! Your ears!
The ear listens with the help of hair cells, which are located on the resonant basilar membrane in the cochlea of the inner ear. Each hair cell is precisely tuned to a specific narrow frequency range, which is determined by the position of the cell on the membrane. The peak of the sensitivity is in the middle of the frequency range, which gradually decreases in both directions and takes an asymmetrical cone-shaped shape, overlapping the frequency ranges of neighboring cells. We do not hear sound if there are no hair cells tuned to that frequency.
The left side of the figure shows a cross section of a human snail with a basilar membrane (beige in color). The membrane is designed to resonate in different places along its length, depending on the incoming frequency: high frequencies resonate closer to the base and low frequencies at the opposite end. The figure shows the approximate locations of various frequencies.
The right side is a schematic diagram of the response of hair cells along the basilar membrane, as a group of overlapping signals.
The process is similar to an analog radio receiver, which receives the frequency signal to which it is tuned from a nearby radio station. The more the receiver and station frequencies do not match, the more unstable and distorted the signal will be, regardless of its strength. There are upper (and lower) levels of the frequency range beyond which hair cells cannot receive signals and we cannot hear anything.
Sample rate and audible frequency spectrum
I’m sure you’ve heard many times that frequencies 20 Hz to 20 kHz are the audible range of the human ear. It is very important to understand how scientists obtained such numbers.
First, we measure the “hearing threshold” across the entire audio range for a group of listeners. This allows us to construct a curve that represents the quietest sound the human ear can hear at any given frequency, measured under ideal conditions in healthy ears. An anechoic environment, accurate calibration of breeding equipment, and rigorous statistical analysis are an easy part of the experiment. Auditory concentration is lost very quickly, so the test must be performed while the subject is not tired. As a result, there are many breaks and pauses, and testing can take from several hours to several days, depending on the methodology.
