![\"El](\"https:\/\/logisticaempresarial.es\/wp-content\/uploads\/2019\/06\/AdobeStock_101059968-1080x675.jpeg\")
<\/p>\nThe best posts are born from comments. And yes, there was talk of sound quality, digitization and more. After recently trying an in-depth study to understand what was the best optimization for an audio file, I’m ready to share my findings with you. For many, it will be hot water, I hope at least someone finds a decent explanation. I want to dispel the myths of loss-making formats, mp3 in my head and convince him that between an audio CD, a flac and an mp3 “good” nothing changes a tube. Listen to faith.<\/p>\n
<\/p>\n
First, let’s see what it means to digitize an analog signal.
\nDisplay the audio signal with a waveform. This is your analog signal. You don’t want to take the whole signal, but the bare minimum, so you can rebuild it later. You must select the value that the wave assumes at a specific time (sample). Therefore, the number of points it takes in a second corresponds to the sampling frequency and must satisfy a particular sentence so that there is no ambiguity (technical alias) in the waveform you would suggest again. The best picture that clarifies the concept is the following:<\/p>\n
![\"Aliasing\"](\"http:\/\/upload.wikimedia.org\/wikipedia\/commons\/f\/f6\/Aliasing-plot.png\")
<\/p>\n
The red signal has been sampled too low in frequency, and the stored dots (squares) are not sufficient to uniquely reconstruct the original signal. In fact, the blue line passes through the sampled points.<\/p>\n
For each point you mark, you must now select the precision or decide how much space (bits) you want to allocate to store the data. Mathematical comparison, decide how many numbers to keep after the comma. The more you hold, of course, the better it will be. However, for some signals, it is not enough to have very accurate precision. For example, 3 numbers after commas will suffice (precision to the thousand). In other cases, you need more precision because the variation can take place on the fourth or fifth decimal. In information theory, more bits are used for the least used values, and vice versa, fewer bits are used for the most used ones (as they transmit less information).<\/p>\n
Let’s see what the specifications of an audio CD are. The acoustic signal produced by musical instruments is digitized using a 16-bit quantization and a sampling rate of 44100 examples per second. Second: that is, 44100 examples are saved for every second of the song and each sample is saved using 16 bits of memory. The bit rate of an audio CD is 44.1 kHz * 16bit = 705.6 kbps.
\nAn mp3 encoder reduces the number of bits used to store the individual sample. In fact, the sampling frequency is a variable that is generally not played on mp3. However, it is of vital importance as it must satisfy the Nyquist-Shannon theorem: To avoid loss of information, an analog signal with finite bandwidth must be sampled at twice its maximum frequency. This allows us to eliminate the alias problem seen earlier. Used for our example, since the upper limit of the audible frequency of the human ear is around 20 kHz, a good sampling frequency is around 40 kHz. Typically, 44.1 kHz is used.<\/p>\n
Start a small digression – a good site for sound testing to check the frequency range you hear (or how your system plays them) is http:\/\/www.audiocheck.net\/. Personally, my ears (equipped with my now famous AKG headphones) are valued from 20Hz to 19kHz, while with iPod headphones I go from 40Hz to 18kHz (and it was over that AKG paid them 130 bitches). End of digression.<\/p>\n
However, at the same sampling rate, the “precision” with which the data is stored contributes significantly to the quality of the compressed audio signal. To summarize, I would say that the sample rate plus a high bit rate guarantees good quality. How high should the bit rate be? Well, very often it depends on the actual signal, so much so that some mp3 coding algorithms have a so-called variable bit rate (VBR), where more memory bits are used for the parts of the signal used with less frequency according to Huffman coding.<\/p>\n
In general, the 128 kbps files have good quality, but personally I find a noticeable improvement in the 256 kbps files, while between 256 kbps and 320 kbps (always in person) I hear no difference. And much less between 320 kbps mp3 and audio CD (706 kbps) …
\nTry it for yourself: (I didn’t want to influence you, but by the names you can also imagine what the mp3 coding specifications are).<\/p>\n","protected":false},"excerpt":{"rendered":"