Now that almost all music software popularizes the FLAC lossless format, the MP3 format is a thing of the past. But the most common audio format in this memory, although the size is only a fraction of others, the sound quality is not bad, it is inevitable to wonder what black technology it uses.
Since Edison invented the gramophone, humans have been preserving and disseminating sound for more than 140 years.
But in the last 20 years, the birth of an audio format has changed the music experience. That is MP3.
As the world's most widely used audio format, MP3 was born for the first time to make music a mass consumer product that everyone can reach.
There should be many small partners have the first music player, that is, MP3 Player, even in that year, everyone defaulted to MP3 is synonymous with music.
But the same digital audio, MP3 is only one-twelfth the size of a standard CD, and for the human ear, the difference between the two, there is, but it is much more difficult to detect than picture compression and the like, so what does it do to music? What did you lose?
Today, let's talk about your most common audio format - MP3.
What exactly is missing from MP3?
What is the difference before and after compression?
If we want to reduce the size of a file, the most straightforward way is to compress.
Usually we understand compression, which is repetitive compression. For example, if you go to the supermarket to buy 5 bottles of Coca-Cola, you will not write 5 colas on the receipt, just write "Coca-Cola *5". This process is equivalent to encoding duplicate parts of the file in shorter bytes. The file itself does not lose data, and it does not lose any information after decoding, but only makes the file smaller.
This is a lossless pressure loss, which is actually the last step of MP3, which is done using an algorithm called Huffman Coding. But with this algorithm alone, mp3s won't shrink significantly.
Because the sound itself is an extremely chaotic, information entropy data is very high. So it's impossible to shrink it down to 10% of the size of the CD in this way.
Since the road of lossless compression does not work, it is better to simply throw away some information.
So what sound did MP3 lose?
The easiest way to know the answer is to compare.
We juxtapose mp3s of the same sound with its lossless version on two tracks, invert one of the tracks, and if the two sounds are the same, they cancel each other out, and we should get a mute effect. This is also how noise-cancelling headphones work.
But because MP3 is lossy compression, it actually sounds like this:
But only then we can't tell the difference between MP3 and lossless. If a piece of music is constantly switching between MP3 and lossless, can you really tell the difference? I'm sure you can't hear it.
This is the magic of the MP3 algorithm, its compression is not simply to throw away the sound data, but to lose the data at the same time, you can not perceive.
The birth history of MP3
The story of Brandenburg and Dieter Seitzer
In the late '70s, a German professor named Dieter Seitzer popped up with a forward-thinking idea. He wanted people to sit at home and use ISDN's phone line to call for music on demand, like a jukebox.
ISDN was a popular digital line at the time, also called "one-line pass", you can use it to make calls, send and receive faxes and the like, but its speed is only 128kbps.
So when Dieter patented the idea, the staff at the patent office told him it was no good. Unless you can increase ISDN's network rate by a factor of twelve, that's enough to transfer the amount of data for a CD.
Dieter listened, and estimated that he could not do anything about speeding up the Internet, but if he invented an audio format that was only 1/12 size of a CD, wouldn't he be able to engage in this "digital jukebox". So he turned around and gave the invention to a student of his name, Karlheinz Brandenburg.
Brandenburg's master's thesis focused on a speech compression algorithm called ATC, which Dieter found out about. But Brandenburg was also depressed when he received the task.
I thought tnnd you can't do a professor, how can my little doctoral student do it, but the task given by the supervisor can't be completed, so he plans to use a few years to prove that this is impossible, and then write a thesis, mix the doctorate into the hand to count the ball! And in the process of proving that it was impossible to invent an audio file that was 12 times smaller than a CD, he found that, heck, it was possible.
Psychoacoustics with MP3
The key to making Brandenburg think there's a drama is an extremely unpopular discipline called Psychoacoustics.
This is a branch of psychophysics, which originated from the exploration of music and musical instruments, and later became the study of the relationship between human physiological senses and the objective sound world. It sounds very complicated, but it is actually very easy to understand, such as the sound positioning of the human ear, the study of the hearing range, all belong to the category of psychoacoustics.
There is a very classic study in psychoacoustics, which is the isophonic curve. This curve tells us that the human ear perceives sounds of different frequencies very differently. Human hearing ranges between 20-20,000Hz, and at different frequencies in this range, we hear different loudness.
Low-frequency sounds require greater sound pressure to be equal to smaller sound pressure at mid-range frequencies, resulting in sound sizes that sound equal.
For example, the bass needs to be louder to sound as loud as a guitar. This is also why the bass speakers of bass are much larger than other speakers.
The lowest point in the figure appears around 3000Hz, so it shows that people are most sensitive to the sound of this frequency, and only need relatively small sound pressure to hear. For example, most of the alarm sounds we hear have a fundamental frequency of 1000-3000Hz, so that the human ear can more easily capture these sounds and avoid danger.
Interestingly, as we age, our range of sound is getting smaller, and most adults can't hear sounds at frequencies above 16,000 Hz.
In this direction, the discoverers of the waiting curve have developed an even more magical thing, making us realize how huge the gap between our senses and objectivity really is.
This thing is "Masking".
Sound Masking
One afternoon in 1958, a psychologist named Licklider went to the dentist and told the doctor, "I don't need anesthesia." Then I pulled out a pair of headphones and started listening to deafening music. To the sound of great music, the dentist helped him get three caries, and he felt no pain as if he had been hit with an anesthetic.
Licklider named the technology Audiac, and he took it with him to go on a tour with the dentist to extract teeth, and also helped many women relieve the pain during childbirth.
Audiac uses a strong auditory stimulus to suppress pain, which is a cross-sensory masking effect.
In the auditory system, one sound is also masked by another sound that is emitted at the same time. For example, in a band, it is often the guitar sound in the C bit, but if suddenly there is an instrument of similar frequency added, such as trumpet. The sound of the guitar will be briefly drowned out. This process is called co-frequency masking.
Let's use an animation as an example, when a noise with low to high frequencies passes through a sine tone, the sound is masked by the noise.
That sounds like this:
For example, when you are in class, you want to use a cough to cover up the sound of your fart, then you had better meet three conditions, one is that the cough sound is longer than equal to the fart sound, and the other is that the loudness of the cough sound is greater than or equal to the fart sound. The third is to ensure that the frequency of the two is close. These three conditions are met, which is an excellent cover-up.
So what does this have to do with MP3? MP3's algorithm takes advantage of this feature of the human ear to throw away the drowned sounds in different frequencies of the song. This allows for minimal loss of sound quality while reducing file size.
Temporal masking
But that's not enough.
When we hear a noise stop abruptly, there is actually a gradually decreasing masking effect of 100-200ms. During the period after the noise has completely stopped, smaller sounds than his are obscured and we are completely inaudible, just as our ears take 200ms to regain consciousness.
Not only that, but the noise masks the sound that precedes it, and although it's only 50ms, it's been a long time for the senses, which means that our brains need a 50ms buffer to report to consciousness.
And this before-and-after process is called time masking.
The core of the MP3 compression algorithm is to use a carefully iterated human auditory psychology model for many years, to correspond every moment in the music to each frame in the MP3 file format (FRAME), check the frequency and time period of the above two masking effects in this frame, and throw away the audio information that is masked and we can't hear.
This process is not simply a precise and mechanical judgment, its background color is actually sensory control.
In the early days of the MP3 algorithm test, testers need to find problems with the MP3 compression algorithm in a large number of songs. They are compared between mp3 and lossless versions of various songs, and they are rated for each song they listen to, with a total of four grades: inaudible, a little different, a little ugly, and very ugly. In particular, the last two options can be said to be a very subjective judgment.
This means that the invention and improvement of the MP3 algorithm actually takes the subjective judgment of people as one of the goals of consideration. We can't say that this algorithm is completely subjective, and it's not absolutely objective, so its effect can't be even across all songs.
Vega and the battle for audio coding standards
Speaking of which, I have to mention a small story in the process of the invention of MP3. In the final stages of testing the MP3 compression algorithm, brandenburg and his colleagues felt that their algorithm was so good that it was difficult to hear the difference in almost all double-blind tests.
One day he stumbled upon a magazine that said people liked to test their speakers with Suzanne Vega's song Tom's Diner, and he happened to see the CD at Fraunhofer's lab, so he uploaded the song to his computer.
The song is very simple, it is a pure vocal, no accompaniment song. But when he processed the song through the MP3 compression algorithm, he got such an effect.
At the lower bit rate of MP3, Vega's voice becomes hoarse and unnatural. So in the year that followed, the R&D team made thousands of minor tweaks to the MP3 algorithm, and Brandenburg said he had listened to the song at least three thousand times, probably more than anyone else on the planet.
In the end, they successfully compressed this Tom's Diner, and through this song, they really improved the compression algorithm of MP3.
Many years later, Brandenburg actually met Vega and listened to her sing that Tom's Diner live. Although he has listened to it countless times, he says the song is still good.
Brandenburg finally published his paper in 1989. The next step is to bring this technology to the world. That is, in the early 90s, several emerging technologies suddenly appeared in the industry, and they were looking for a new audio coding standard to use, including the familiar "CD-ROM" and "DVD". So he and his team submitted entries to the Moving Image Expert Group (MPEG), where they competed with 13 other teams for new audio coding standards. The biggest competitor came from an organization called MUSICAM, and behind this organization was Philips, which at that time held the patent for CD discs, which can be said to be in the sky.
So although their team had less technical data and stronger sound quality, they eventually lost to MUSICAM.
Because MUSICAM's algorithm requires less processing power. In an era when processors were generally not very good at that time, it was indeed more advantageous.
So in those days, MP3 was a proper failure. Even its inventors have begun to work on new audio encodings. The MP3 was thus thrown into the dustbin of history.
It wasn't until the mid-1990s that two revolutionary technologies were born that brought the outcast of MP3 back to life: the World Wide Web and Windows 95.
A research and development team, also from Germany, developed a software player for MP3 and released it on Windows systems.
At that time, 1 GB-sized hard disks were just beginning to gain popularity, and the storage space was very precious, but the processor had a huge boost. Therefore, the smaller MP3 was slowly accepted by everyone, and unexpectedly became the new audio coding standard. On July 14, 1995, MP3's birthday, Karlheinz Brandenburg and his colleagues at the Fraunhofer Institute decided to name the extension of this compression algorithm, MPEG-2 Audio Layer III, according to the industry standard's full name, .MP3.
By the late 1990s, "MP3" replaced "SEX" as the most searched word on search engines. Once, on a business trip to Hong Kong, Brandenburg saw 30 different brands of MP3 players in the window and thought, "Okay, we finally won." ”
MP3 is disappearing
The birth of MP3 is much more complicated than I thought, it is a scientific research that took many years and countless iterations to obtain. This achievement can be said to have reshaped the human music industry. It was also from MP3 that music became a mass consumer product that everyone could touch.
From vinyl and tape to CDs and MP3s, every technological innovation is changing people's music experience and the way people consume music. And MP3 is unique in this history. Those who admire it believe that MP3 is so great that it makes it easy for everyone to enjoy music, while those who oppose it see it as a beast because it swallows up the copyright on which record companies lived and the golden age of the record industry.
Today, digital music is still alive, but MP3 is on the verge of being obsolete. We no longer need to download music to the player, and then use the player to listen to music, everyone is using mobile phones to listen to songs online, 5G communication and hundreds of gigabytes of memory, making audio compression more unnecessary, music platforms have gradually shifted to lossless formats such as FLAC.
But we all remember the days when we listened to MP3s, and the music that accompanied us.
Author: Yangzi Video Production: Zhang Yong & Little Claw
Animation: Tianyu Mei Editor: Huan Yan (Video) & Xuan Xuan (Tweet)
Images, sources:
Not present S2E1 MP3 - Heavy light
Equal-loudness contour
Karlheinz Brandenburg，Dieter Seitzer，Huffmancoding -Wikipedia
How MP3 Got Its Groove -Internet hall of fame
The History of MP3 Technology -ThoughtCo.
The Human Auditory System -ScienceDirect
The Music Industry Strikes Back | System Shock Ep 2 -Bloomberg Quicktake: Originals
How a Suzanne Vega Song Helped Create the MP3 - Great Big Story
Perceptual Coding: How Mp3 Compression Works -WaybackMachine
SOUND THAT KILLS PAIN - Will Stephenson
Voice Signal Processing - Han Jiqing
History of Acoustics - Psychoacoustics - Zhihu Sub-Fish
MP3 came to an end, and the technology born for the telephone and radio changed the entire music industry - Curiosity Daily
This article is reprinted from the WeChat public account with permission: Bad Review Author: Bad Review Jun
The reproduced content represents the views of the author only
It does not represent the position of the Institute of High Energy of the Chinese Academy of Sciences
Editor: Liu Yulong
Great videos Don't miss out
[Oriental time and space] My family and my country Chen Hesheng
Xingbiao pays attention to the WeChat public account of "Institute of High Energy of the Chinese Academy of Sciences"
Don't miss the highlights