Digital audio quality analysis-blind listening experiments with different precision and sampling

We compared and tested the sound quality of 16-, 20-, and 24-bit quantified digital audio data in the lab, testers with "golden ears" will use their feelings to tell you the difference between the quantization precision of different bit numbers. Through these test results, you can even see the differences between different sampling frequencies. The results of testing in well-equipped acoustic recording studios are very convincing. We can see what the audios of extraordinary audios have heard, but the more important question left behind is: can we hear this difference in our daily life? This is exactly what this article will answer you.

Although our experiments have a variety of requirements, we have never used any sound material that is used for listening tests. In fact, we want to test the real world situation of sound with different precision. Therefore, the selected materials are recordings in the real world, and the playing environment is also the real world environment, it also uses a variety of different audio systems. The most important thing is that we need to test whether these audio tracks of different precision will sound different after they are eventually burned to a 16-bit/44.1 kHz music CD.

The hardware and software materials we used in the experiment were: We invited some musicians (two classical guitar players, one violin player and one female singer) to brilliant studios in San Francisco ). We use the u87 and C 460 microphones from the akg company to perform recording. After the signal passes through a range microphone's front-power amplifier, after the DB 44-96 converter of DB technologies is converted to 16-bit/44.1 kHz, the 24-bit/44.1 kHz and 24-bit/88.2 kHz digital signals enter the sonic solutions digital audio system. (In the sonic solutions system, two sampling frequency converters are required to convert the 96 KHz sampling frequency to 44.1 kHz, so we use a sampling rate of 88.2 kHz, in this way, only one converter is needed .)

After recording a sound track, we convert some of the audio tracks into 16-bit/44.1 kHz format and burn them into a music CD.

Can these golden ears hear the differences?

The listener test is divided into two parts: the first part is carried out during the recording phase. The Listener does not change the sampling frequency and quantization accuracy of the original audio track. The method is to output digital signals directly from the sonic solutions system, by Meyer HD-1 monitoring speaker playback out. The test results can be said to be varied. It is difficult for some listeners to distinguish between 16-bit and 24-bit signals. In fact, when the same audio track is repeatedly played, the listener often changes their conclusions.

The difference between 44.1 kHz and 88.2 kHz is obvious. Most people can quickly make judgments when high-precision sound is played back.

One interesting thing is that some listeners prefer 16-bit/44.1 Khz audio tracks. Their point is that these low-precision sounds are warmer and smoother, and there is a sense of presence that is not heard in high-precision sounds. They think that high-precision sound is thin and the sound is brittle. Others disagree with this idea. They believe that the sound field of high-precision sound is wider and the details are better.

When the first part of the test is complete, we convert the 24-bit/44.1 Khz audio track to 16-bit/44.1 kHz format and then burn it into a music CD. (During the test, the sonic solutions audio workstation's 24-bit/88.2 kHz to 16-bit/44.1 kHz conversion function was not completed yet, there are no hardware devices for this operation .)

Then we get the 16-bit/44.1 kHz version and the version converted from 24-bit/44.1 kHz to the keyboard magazine's own music studio, use the Pro Tools digital audio workstation of digidesign for 30-second clip playback. There are 13 such clips and two full-length original audio tracks in the burned music CD. Then let the listener listen to these sounds and fill out the questionnaire form.

The final result is: although we can hear the differences between the recorded sound with different precision, the speaker cannot tell which Sound clip is accurate. When we carefully looked at these surveys, we found that some audients thought that the "sound is good" piece was 24-bit/44.1 kHz, but the truth is, they felt that the "sound" clip was recorded at 16-bit/44.1 kHz.

Another interesting conclusion is that when some listeners listen to a sound clip for 30 seconds, it is difficult to tell the difference between different precision, but when they listen to a full-length audio track, then there is no problem. For others, the result is the opposite.

Different listening results are also related to the listening system used by the listener. In the main music studio of keyboard magazine, audio tracks are played back with a pro tools audio system and 888/24 I/O (16-bit data ), played back from the active speaker of event 20/20. The speaker can easily hear different voices with different precision. Similarly, this is true when we use the listening systems of genelec and dynaudio acoustics. However, it is not so easy to draw a conclusion through the sound of a home stereo, a variety of headphones, in the car and other environments.

The final conclusion is: During the audio recording test phase, most audients think that the sound field of a 24-bit recording is wider, the space is better, and the details are richer, however, the voice recorded in 16-bit cannot do this. This is a good thing or a bad thing. The negative effects of high-precision sound (such as short-term pulse waveform interference and microphone manual processing) on auditionists are significantly greater than those on their own. In addition, in the first part of the listening test, some people think that 16-bit sound is better. Others disagree. They prefer the wide sound field and rich details brought by high-precision audio.

During the 13 30-second fragment listening test, the listener can draw a correct conclusion when the clip is played back to about 65% places. In the test of two full-length audio tracks, the listener can draw a correct conclusion when the audio tracks are played back to about 43% points. One thing to note is that all listeners can identify sound fragments with different precision, but they cannot accurately tell the accuracy of the sound they hear.

What do these results mean? First of all, through our test, we can tell the difference between the different quantization precision and the sound recorded at different sampling frequencies, even when the audio data is converted to 16-bit/44.1 kHz, this difference still exists. However, the speaker has different opinions on whether high-precision sound is better.

The final conclusion of the test is as follows: "24-bit does add something new to the sound, but I cannot say whether these new things are needed in my music ". Particularly, it is not suitable for high-precision audio systems for dance music and electronic music.

However, audio recording of human and mechanical instruments will benefit from high-precision digital audio. "I really want to use high-precision equipment for recording music played by mechanical instruments ". All listeners have reached a consensus that high-precision digital audio equipment should be used for mother band production or on-site recording with strict requirements.