Since traffic is not valuable, the sound quality provided by music software is getting higher and higher. In addition to traditional lossy... oh no, standard quality, there are also lossless formats.
Today, there is even a quality that is “more lossless than lossless”. Of course, this requires you to spend money to become a more luxurious member than a deluxe member in order to hear it.
So, is lossless music an IQ tax? Why is there any sound quality that is better than lossless?
In order to clarify these issues, we not only checked the information, but also found the host of the podcast "Nine Duan Qi Tan", Mr. Yu Mengqi, and the head of the audio of QQ Music Tianqin Lab, Mr. Zhao, and "questioned" them face to face. what's going on.
In fact, from a strict physical sense, no matter how much money you spend, the "lossless" music you hear cannot be truly lossless music.
As we said in the previous video, the sounds in nature are all continuous analog signals. Digital music stored in the form of 0s and 1s are discrete digital signals. So when recording the original sound of an instrument or singer into digital music, there is a conversion process.
The accuracy of this conversion is the key to making lossless music.
To convert analog signals in nature into digital music of 0s and 1s, you need to select points from the original sound and record their information. The denser the selected points, the more accurate the conversion. Three basic concepts are involved here: sampling, quantization and encoding. Don't be afraid, it's very simple. You can still install it after you learn it. The number of samples in one second is called the "sampling frequency". The sampling frequency of CD is 44.1kHz, which means that each second of music contains 44,100 sampling points.
And "quantization" is responsible for recording the vertical axis amplitude information. Each piece of music has a strong and weak sound. The difference between the strongest and the weakest is called "dynamic range", and the unit is what we often call "decibel" (dB). The so-called "quantization" is to divide the dynamic range into equal levels, and then classify the volume of the sampling point into the nearest value. Of course, the more detailed the
quantification, the better. The precision of quantization is called "bit depth", and the unit is bit. For example, if the bit depth is 2 bit, that is (2^2), divided into 4 layers, then there are only 4 situations at most for the volume of each sampling point. The sound from our phone calls and broadcast speakers has a relatively low bit depth and sounds rough. The bit depth of music in CDs can reach 16bit, which sounds much more delicate. After
is sampled and quantized, we can obtain the magnitude of each discrete point. The process of recording these quantities as 0 and 1 is called "encoding".
So to put it bluntly, sound quality is also very simple: the denser the sampling and quantization points, the better the sound quality.
But the problem is that no matter how precise it is, sampling and quantization only record the information of each point. Compared with the smooth original sound, there will definitely be information loss, and it cannot be lossless in a strict physical sense.
But it doesn’t matter~because your ears cannot receive 100% of all audio information. Even if Jay Chou were brought to your territory to sing "My Territory" to you, you wouldn't be able to hear every detail of his pronunciation.
So "lossless music" does not need to be lossless in the physical sense. As long as your ears can't tell the difference between the original sound and the recorded music, it's lossless.
So, how accurate does the sampling have to be to not be able to tell the difference?
Generally speaking, the upper limit of human hearing frequency is 20kHz. According to the famous "Nyquist-Shannon sampling theorem", when the sampling frequency is greater than twice the highest frequency of the analog signal, no information will be lost. So if the sampling frequency reaches 40kHz, which means at least 40,000 sample points are included in one second, it is enough to feed your ears.
This is also the reason why the CD frequency is 44.1kHz and the audio sampling rate of mainstream videos is 48kHz - anyway, more than 40kHz is enough, and they leave you a little margin ~
quantization is similar.The sound of human speech is usually between 40-60dB. Listening to sounds exceeding 90dB for a long time will damage your hearing. According to this formula of bit depth and dynamic range, the bit depth of CD is 16bit, and the dynamic range can theoretically reach 96dB; some classical concerts use 24bit bit depth storage, which has a higher dynamic range, which is enough to wrap your ears.
Generally speaking, if the sound quality of a song reaches CD level, it can be called lossless music. Lossless music contains a lot of information and is large in size. A CD has a capacity of about 600-700MB, which can only give Jay Chou time for a dozen songs.
Lossy compression formats such as mp3 reduce the file size by removing part of the information in the sound.
For example, this is the lossless version of "My Land", compared with the ordinary mp3 version. It can be seen that a large amount of high-frequency information disappears in the mp3 format file. But these high-frequency information are difficult to capture by the ears, and the algorithm design of mp3 is very excellent, so you may not be able to hear the difference.
Thanks to FiiO for providing the equipment, we organized a blind listening test within Diesel.
We selected two pieces of classical music and two pieces of pop music, each providing two different sound quality options. In the first round, everyone was asked to guess whether the music they heard twice had the same sound quality. The second time was for everyone to tell which of the music they heard was of higher quality and which was of lower quality.
The result is that the test accuracy of most of my friends is indeed the same as that of Xiameng. Four of the friends guessed it correctly, but three of them admitted that they were fooled. Only one friend said that she could really hear the difference in high frequencies. So we took her for another round of testing, and found... she still didn't stand the test.
Generally speaking, our test results show that most people cannot hear the difference between high-quality MP3 and lossless music. But there is no difference at all. After all, if it is purely blind, from a probability perspective, the probability of getting all wrong and all right should be the same. However, the actual result is that 4 colleagues got all right. Several people also said that the difference in listening feeling of this song "Yellow" is relatively obvious.
Our testing is relatively small. However, Mr. Zhao from QQ Music told us that they have conducted a test on a scale of 50,000 people, allowing everyone to blindly choose whether they prefer SQ (lossless sound quality) or HQ sound quality (high quality). The result is that 57% of users think SQ sounds better, and 43% think HQ sounds better. This shows to a certain extent that some users can indeed hear the difference, but for most users, the difference may not be significant.
But we are very curious, ordinary people can't hear it but we can understand it. But can audiophiles who have spent time and money really hear the difference between high-quality mp3s and lossless files?
Chai knows: Judging from your experience, can they bring about a clear difference in listening?
[Jiu Duanqi Talk] Yu Mengqi: I think the difference between 320k mp3 and lossless is actually not that easy to distinguish as you said. I originally tried it in our own circle of friends. For example, I prepared three songs, and then packaged them into wavs, but one of them was transferred to mp3, and the other was a real wav. Anyway, I admit that I can’t hear pop music at all, but I can hear classical music.
Chai knows: Then why is there such a problem that pop music cannot be heard but classical music can be heard? What's the difference between them?
Yu Mengqi: Relatively speaking, I understand that the various sounds of classical music will be more coherent, but the way pop musicians sing lyrics is not so coherent. For example, analog (signal) is a perfect sine wave, and digital (signal) is steps. The more incoherent your own loudness changes are, the harder it is for you to expose the sense of digital steps. And if you have a coherent sound, such as many string sounds, then it is easier to make this difference.
But in any case, the mp3 version does lose some audible information, which no matter how good the equipment is, cannot make up for it.Today, when hard drives and traffic are no longer valuable, if you are an audiophile who has high requirements for sound quality, it is indeed possible to hear more details when listening to lossless music. This cannot be said to be an IQ tax.
Lossless music has more details than ordinary music, which is easy to understand. But as I just said, the hearing range of the human ear is limited, and music that can achieve CD sound quality is lossless music. So why are there sound quality options on the music platform that are more lossless than lossless music? If you buy a luxury membership, you have to upgrade to a super membership to listen to it. Does it make sense? How to talk about
? From a purely theoretical point of view, it is not completely meaningless.
Although countless people, including us, have said that the human ear can only recognize sounds with a frequency of up to 20kHz, this refers to the upper limit of a single tone: that is to say, when only playing a single fixed frequency sound, you can only Can hear 20kHz.
But this does not mean that when you listen to music, you are completely unable to perceive sound waves with frequencies above 20kHz. This 2000 study, for example, found that subjects had no response if sounds above 22kHz were played alone. But when listening to music containing frequency information higher than 22kHz, the subjects' brain waves were more active, and they also preferred the music version containing high-frequency information.
So, although we can’t “hear” ultra-high frequency sounds above 20kHz, we may still be able to perceive their existence in some way.
On many music software, there will be a small "Hi-Res" gold mark next to the music that is "more lossless than lossless". This was originally an audio standard proposed by Sony, which means music with higher fidelity than regular CDs and lossless music, usually representing songs that can reach a 96kHz sampling rate and 24bit bit depth.
Of course, just because it’s better in theory doesn’t mean you can hear it. To hear the difference between Hi-Res and ordinary lossless, the level requirements are even higher.
Yu Mengqi: Hi-Res is an authentication method that I think has greater commercial effect than practical effect. If we speak objectively, I think it guarantees some bottom line. But it completely fails to express the upper limit. It can only ensure the bottom line for example.
To be honest, the improvement from "lossless" to "Hi-Res" is definitely not as big as the improvement from ordinary mp3 to lossless.
Compared to continuing to improve the sound quality on the basis of ordinary lossless, what everyone may be more concerned about is another question: Even if you spend money, if you hear lossless, does it have to be true lossless?
is indeed not necessarily the case.
A very simple example: For example, many old songs were not recorded to such high standards when they were originally recorded, and they did not meet the "lossless" standard from the source. If you want to turn them into "lossless music", you can only forcibly upgrade the specifications through post-processing methods.
For example, we can use some technical processing to force Zhou Xuan's old song to a near-lossless specification - but this is like forcibly outputting a 480P video into 4k. It does not increase the amount of information, but only wastes more information. Lots of traffic.
In fact, this is also a problem that music software has to face. Teacher Zhao from QQ Music told us that most of the sources of their lossless music are original files provided by the official. But among the large amounts of lossless music, there are occasionally some audio sources that do not actually meet the lossless standards.
[QQ Music] Teacher Zhao: After he comes over, we need to do a test on the lossless file. However, we found that there may be some songs that are unofficial, and some songs may not necessarily meet the conditions of lossless, but they are converted to lossless. At this time, we will reject them and ask them to provide truly lossless ones. However, this tool does not guarantee a 100% accuracy rate and may have some error rates. However, our tools have been developed for many years from 2012 to now, and the tools are becoming more and more perfect. Basically, there are very few fish that slip through the net. Our standards are set relatively high, other platforms may just be CD quality. We have two levels higher than CD.
So sometimes, audiophiles with a keen sense of hearing may find that although some music is marked "lossless", the listening experience is very average.The lossless sound quality on the
music platform and the rarer Hi-Res sound quality are all obtained directly from the original files. And these higher-standard "mastering" sound quality options involve the participation of AI.
Teacher Zhao: Nowadays, artificial intelligence technology is developing very fast, so we are trying to use some technologies to recover some of the higher information content of this song. So we made a master tape. Then all the audio sources we got did not have such quality songs, so we tried to find ways to recover this part of the information. So this is what we do through algorithms.
As for whether you choose to listen to the original sound quality or the AI-processed sound quality, it depends on your own choice.
If you get real lossless music, you must listen to it with professional equipment to achieve its full effect. If you are interested in this, you can jump to this video we did. Whether you can hear it or not depends on your personal nature.
But whether you can hear the difference or not, it doesn’t stop you from burning more expensive equipment and stocking up on lossless music with higher specifications.
Yu Mengqi: I used to be the kind of audiophile who doesn’t listen to music. That is, I may only have ten songs forever, and then I keep changing devices, and then listen to these ten songs forever. I have been such an enthusiast for a long time, but on the one hand I felt that I was wrong, and on the other hand I couldn't stop. But then I gradually became more and more interested in music, so now I am an audiophile who prefers music more.
In fact, players who have been involved in any field know that the fun of playing with equipment often lies in playing with the equipment itself, not in the final actual effect.
Just like people who spend hundreds of thousands of dollars to buy a computer, they may not really use it to cut videos and play games, just to get the pleasure of "the score is a little higher"; people who spend hundreds of thousands of dollars to buy a camera may not really be able to take pictures. Taking good photos is just to see if the "German taste" is a little bit heavier again; people who spend millions to buy a car will never reach the speed limit in their lifetime, just to see if the response speed is 1 millisecond faster.
In the final analysis, the fun of tossing lies in the tossing itself.
Okay, thank you for coming here. In order to explain lossless music clearly, we have made two videos, and also contacted manufacturers, audiophiles, and music platforms for interviews. We are very happy in this process and hope you can enjoy it.
See you next time!