24/192 Music Downloads are Very Silly Indeed
24/192 Music Downloads
...and why they make no sense
Xiph homepage
Planet.Xiph
Monty's demos
Intro
Physiology
Audible spectrum
Golden ears
Spectrophiles
Intermodulation
Fallacies
Oversampling
16 vs. 24 bit
Noise
Dynamic range
Signal-to-noise
Why 24 bit?
Listening tests
Caveat Lector
Confirmation bias
Loudness tricks
Clipping
Different masters
Inadvertant cues
Better headphones
Lossless formats
Better masters
Surround
Outro
Further reading
Footnotes
Also see Xiph.Org's new<br>video, Digital Show<br>& Tell, for detailed demonstrations of digital sampling<br>in action on real equipment!
Articles last month revealed that musician Neil<br>Young and Apple's Steve Jobs discussed offering<br>digital music downloads of 'uncompromised studio quality'.<br>Much of the press and user commentary was particularly<br>enthusiastic about the prospect of uncompressed 24 bit 192kHz<br>downloads. 24/192 featured prominently in my own<br>conversations with Mr. Young's group several months ago.
Unfortunately, there is no point to distributing music in<br>24-bit/192kHz format. Its playback fidelity is slightly<br>inferior to 16/44.1 or 16/48, and it takes up 6 times the<br>space.
There are a few real problems with the audio quality and<br>'experience' of digitally distributed music today. 24/192<br>solves none of them. While everyone fixates on 24/192 as a<br>magic bullet, we're not going to see any actual<br>improvement.
First, the bad news
In the past few weeks, I've had conversations with<br>intelligent, scientifically minded individuals who believe<br>in 24/192 downloads and want to know how anyone could<br>possibly disagree. They asked good questions that deserve<br>detailed answers.
I was also interested in what motivated high-rate digital<br>audio advocacy. Responses indicate that few people<br>understand basic signal theory or the<br>sampling<br>theorem, which is hardly surprising. Misunderstandings<br>of the mathematics, technology, and physiology arose in most<br>of the conversations, often asserted by professionals who<br>otherwise possessed significant audio expertise. Some even<br>argued that the sampling theorem doesn't really explain how<br>digital audio actually works [1].
Misinformation and superstition only serve charlatans. So,<br>let's cover some of the basics of why 24/192 distribution<br>makes no sense before suggesting some improvements that<br>actually do.
Gentlemen, meet your ears
The ear hears via hair cells that sit on the resonant<br>basilar membrane in the cochlea. Each hair cell is<br>effectively tuned to a narrow frequency band determined by<br>its position on the membrane. Sensitivity peaks in the<br>middle of the band and falls off to either side in a<br>lopsided cone shape overlapping the bands of other nearby<br>hair cells. A sound is inaudible if there are no hair cells<br>tuned to hear it.
Above left: anatomical cutaway drawing of a human cochlea with the<br>basilar membrane colored in beige. The membrane is<br>tuned to resonate at different frequencies along its length,<br>with higher frequencies near the base and lower frequencies<br>at the apex. Approximate locations of several frequencies<br>are marked.
Above right: schematic diagram representing hair cell response<br>along the basilar membrane as a bank of overlapping filters.
This is similar to an analog radio that picks up the<br>frequency of a strong station near where the tuner is<br>actually set. The farther off the station's frequency is,<br>the weaker and more distorted it gets until it disappears<br>completely, no matter how strong. There is an upper (and<br>lower) audible frequency limit, past which the sensitivity<br>of the last hair cells drops to zero, and hearing ends.
Sampling rate and the audible spectrum
I'm sure you've heard this many, many times: The human<br>hearing range spans 20Hz to 20kHz. It's important to know<br>how researchers arrive at those specific numbers.
First, we measure the 'absolute threshold of hearing'<br>across the entire audio range for a group of listeners.<br>This gives us a curve representing the very quietest sound<br>the human ear can perceive for any given frequency as<br>measured in ideal circumstances on healthy ears. Anechoic<br>surroundings, precision calibrated playback equipment, and<br>rigorous statistical analysis are the easy part. Ears and<br>auditory concentration both fatigue quickly, so testing must<br>be done when a listener is fresh. That means lots of breaks<br>and pauses. Testing takes anywhere from many hours to many<br>days depending on the methodology.
Then we collect data for the opposite extreme, the<br>'threshold of pain'. This is the point where the audio<br>amplitude is so high that the ear's physical and neural<br>hardware is not only completely overwhelmed by the input,<br>but experiences physical pain. Collecting this data is<br>trickier. You don't want to permanently damage anyone's<br>hearing in the process.
Above: Approximate equal loudness curves derived from<br>Fletcher and Munson (1933) plus modern sources for<br>frequencies > 16kHz. The absolute threshold of hearing...