This article was originally published in Cooking Ideas, a Vodafone blog where we collaborate weekly with the goal of creating stories that “feed the mind of ideas.”
A programmer named Jeff Atwood said some time and several entries from his blog, the always recommended Coding Horror, to a healthy entertainment he called The Great Experiment of bitrate in MP3. Its objective: to verify empirically if for ordinary people there are really qualitative differences when listening to music in various MP3 formats compared to traditional ones.
The contestants were the traditional formats called “no loss of quality”, basically CD (Compact Disc) and FLAC versus compression formats with loss of quality: MP3 with different bitrates. The bit rate, better known by its name in English, is a key feature because it basically determines how much information is transmitted per unit of time: in this case it is the waves that define the music and become human voices and instrument notes . In the world of MP3 encodings of 64, 128, 192 or 320 Kbps (kilobits per second) are usually used.
Like everything in life, music coding is a compromise between quality and quantity: a song stored in the best possible format – for almost all experts, that is the CD – can occupy about 50 MB (megabytes), maybe 40 or 35 only using some of the lossless compressors that save some space without loss of quality (FLAC, Apple Lossless, etc.). That same song in MP3 can vary between 4, 8 and 12 MB depending on the bitrate (64, 128 and 192 Kbps). To further complicate the matter, you can also choose between a constant (CBR) or variable (VBR) bitrate that is usually optimal when compressing different moments of the songs with various bitrates.
For many users, being able to store between 5 and 10 times more music in the same space is an important saving, easy to translate if one takes into account the price of hard drives, flash memories or storage on iPods, tablets and the like. But there have always been two schools confronted: that of audiophiles who believe that nothing can equal the maximum quality of the CD and that of those who, with a more practical sense, consider the differences between an MP3 and CD ridiculous, if at all there are.
Atwood’s experimental study sought precisely to shed some light on these theories based on the basics: listening to music, quantifying its “quality” and deciding which is the best format based on the various variables. For this, he prepared five different audio files: one of them uncompressed and another four tablets at different bitrates between 128 and 320 Kbps. He put them on his server so that people could listen to them and vote (with a quality “note” of 1 to 5) without knowing which was which. And in total he got more than 3,500 people to contribute to the results – hundreds more than for many of the “quality studies” mentioned in the TV commercials.
The results were analyzed with a spreadsheet and various statistical tools, which showed trends and conclusions quite clearly:
The only sample that could really be considered very different from the rest was the MP3 at 128 Kbps CBR, the worst quality. That quality is not enough to compare with the rest. The best simply ignore it.
The MP3 at 160 Kbps VBR is the highest quality sample, even better than the MP3 at 320 Kbps CBR. This indicates that the coding with a variable bit rate is higher than the fixed one even at those values, and that 160 Kbps VBR up is impossible to improve qualitatively.
Ironically, this would indicate that there are MP3s that are heard “better” than audio CDs. Several things can happen here: that the “artifacts” created by compression seem to improve the audio or that when testing people “imagine things,” which could also happen. The truth is that the data serves to feed the theory that from 160 Kbps people no longer distinguish one quality from another, as it is deduced from the data.
The conclusion of the study confirms the hypothesis that an MP3 at 192 Kbps VBR has such quality that not even the ultrasensitive and powerful ear of a dog would notice the difference with an audio CD. Wow!
In conclusion, we already know at what rate to code and compress if we want a good saving in storage without losing quality: a MP3 of 192 Kbps VBR, the winning format of the test.
