Data lost due to compression is irreversible Part 2
[Quantization bit number (bit depth)]
◉ Unit: bit
◉ Audio: Resolution related to volume. The higher the value, the more faithfully the quiet sound can be reproduced and the wider the theoretical dynamic range (ratio of the maximum and minimum volume values). 16-bit, 24-bit, and 32-bit floats are used primarily in production.
◉ If you compare it with the video …: Conceptually, it corresponds to the number of gradation bits. In terms of feel, it is almost the same as the dynamic range of the video. The wider the range, the greater the gradation possible without overexposure and underexposure.
◉ Remarks: There is no concept of the amount of quantization bits in compression formats such as MP3.
◉ Image of the number of quantization bits
When a square is cut on the vertical (volume) axis, the volume change less than one step cannot be reproduced, resulting in noise. In other words, the finer the squares, the more accurately the low volume can be reproduced. The actual number of steps in the number of bits in common use is as follows.
・ 16 bits → 65,536 steps
・ 24 bit → 16,777,216 steps
It can be seen that the 24-bit, which is said to be high-resolution, can reproduce the volume change much more accurately than the CD-quality 16-bit. In other words, 24-bit has a “wider dynamic range” than 16-bit.
[Sampling frequency]
◉ Unit: Hz
◉ Audio: Temporal resolution. Involved in the reproducible frequency range. If the frequency is low, the treble range will not be reproduced correctly. As the frequency increases, it is possible to reproduce frequencies above the audible range. Those used primarily in production are 44.1 kHz, 48 kHz, 96 kHz, and 192 kHz.
◉ If you compare it to video …: In terms of temporal resolution, it is equivalent to frame rate. The higher the speed, the smoother the video will be (in the case of sound, it is perceived as treble reproducibility rather than smoothness).
◉ Remarks: The upper limit of the frequency that can actually be reproduced is half the frequency. For example, if the speed is 96 kHz, it can be played up to
48 kHz ◉ Explanatory sampling frequency diagram
If you compare it to a video, you may understand it in some way. As of 2018, I think the lowest line quality that can be used regularly is the “16 bit / 44.1 kHz” used by CDs. If each value gets lower than this, it will collapse more and more so that it can be heard. If the number of bits is small, small sounds are converted to noise, and if the sampling frequency is small, the aliasing noise (noise that is inevitably generated by digitization. Moiré sound phenomenon) falls into the audible range and is comes back jarring. And note that half the value of the sample rate is the upper limit of the actual recorded / played rate. In other words, in the case of “44.1 kHz”, the actual recording / playback is up to about 22 kHz. The human audible range is said to be 20Hz to 20kHz, so that’s a sufficient value in terms of specs. By setting the sample rate to twice the upper limit of this audible range, overlapping noise is removed from the audible range, and by cutting it with a digital filter, jarring noise, which is CD quality, is removed. From this, you can see that “16 bit / 44.1 kHz” is the lowest line.
The master file
must be of high quality
That said, it’s hard to understand how sound quality changes at low bits and low sample rates without actually experiencing it.
