Music, like recorded through a microphone is a wild mixture of waves with different frequenca and amplitude. During digitalization, a specialized A/D converter chip measures the signal strength and stores this measurement. This process is repeated with high speed (and high frequency) and for every single measurement the results are stored in a file sequentially.
The higher the sampling frequency and the finer the A/D converter measures, the better. If for example in case of a CD this sampling frequency is 44.1 kHz, then 44,100 single measurements are stored during one second. The Nyquist-Shannon theoreme predicts, that with this sampling frequency a maximum real frequency of fs/2=22 kHz can be recorded . Professional studio gear typically works with a sampling frequency of 48 kHz. For high quality records like Audio-DVD, the sampling frequency can be up to 96 or 192 kHz .
The precision of the A/D converter can be described by the number of bits used for the result. For CDs this is a 16bit resolution, for professional audio equipment 34 bits are typically used and then converted down as needed . Calculating the amount of data is quite easy. 44,100 times 2 and times 16 results in a so called bitrate of 1411.2 kbit/s for a CD. For a three minute Tango this calculates to around 255 MBit. In reality this amount is bigger, because a CD also stores data for error correction and structure information.
The amount of Data stored on a CD is big and even until the late 90ies, it was very difficult to handle this amount of data with a personal computer. A solution is to simply see the data file as a set of numbers and use compression algorithms from general computing. These algorithms (lossless compression algorithms) are able to compress the set of data and later restore that without loss of quality. F.i. to compress the sequence (1000000001111001) you simply can write (1, 8*0, 4*1, 001). The resulting compressed files typically have a size of 40-70% of the original one .
Lossy data compression was developed from 1982 by Hans-Georg Musmann and Karlheinz Brandenburg in Erlangen (Germany), as part of the MPEG-1 standards implemented and finally in 1995 as MP3 (ISO MPEG Audio Layer 3) published.This method uses psychoacoustic assumptions, similarities between both canals and statistical predictions to increase the compression quality, a good introduction can be found at wikipedia [3,4]. With this method, the original file size can be reduced by a factor of up to 7. The main factors for quality is the bitrate. For the classical MP3 method, the bitrate may be between 8 and 320 kbit/s. Generally bitrates above 96 kbit/s will provide good quality. The software, which actually is used for compression and decompression is called the CODEC. Apart from the bitrate, the quality of the CODEC, the complexity of the music and the quality of the digitalization process is determining the final audio quality. Since the late 90ies, the quality of CODECs has significantly improved . For MP3 the free LAME CODEC is very popular.
Audiodata today also contain so called Metadata. Here special information like title and artist can be stored to be used for searching and structuring of music collections .
For the generation of audio data, this working cycle is used:
a) Retrieving the raw data from CD (grabbing), alternativly digitizing of analogue recordings.
b) Compression with a CODEC
c) Providing Metainformation (tagging)
d) Playing during or after decompressionwith a compatible CODEC
For the choice of the audio file format there are a couple of options:
1. Lossless CODECS
Typical methods are:
– ATRAC (adaptive transforming acoustic coding) of Sony
– Apple Lossless (.m4a oder .mp4)
– FLAC (Free lossless audio Codec)
– MPEG4 Audio Lossless Coding (ALS)
– WMA lossless (Windows operation system)
FLAC is quit popular but has a low compression rate.
2. Lossy CODECS
– MP3 (.mp3)
– Ogg Vorbis (.ogg)
– WMA (Windows Media Audio, .wma)
here MP3 und Ogg Vorbis are popular.
A table with other compression methods can be found here . Some of these methods even contain digital content management.
3. Quality discussion
During the late 90ies, MP3s with a bitrate of 128 kbit/s have been of lower quality. Today, after a significant improvement of CODECS, they are given the attribute “transparent” by a majority of listeners. This means there is no audible difference between an uncompressed and a compressed version of the file. Even if some people say, they are able to discern between compressed and uncompressed (“golden ear”), and even if MP3 has problems with low bitrates, it is still a high quality procedure for higher bitrates. The true scientific approach to testing is a so called ABX test. For this, a program asks you to discern different samples of compressed and uncompressed music and rates your results on a statistical basis. For current CODECS, a difference is not detectable if the bitrate is above 128 kbit/s. You can do the test with the software foobar2000 and a special ABX plugin [7, 8], a special software is also available for apple computers.
4. Summary: What about Tango music ?
I admit, that this lecture was boring and we all ask what´s that to do with Tango music. Today, a Tango DJ has to collect music and to store that into a collection. Looking at todays price for storage (2015: portable disk 2TB 120$, fixed disk 4TB 140$) compression is not a big issue any more.
For todays DJs other criteria are more important. Two of them are: easy handling of Metadata (which is good for all file formats because the ID3 tag system is independant of the CODEC) and compatibility with professional DJ software and databases. In this regard, the compatibility is low for lossless formats. It has been recommended to stay with the MP3 format and use the highest possible bitrate of 320 kbit/s . I personally agree with this procedure, because it will provide the best mix of quality and procedural safety, but can understand the people who support lossless formats like FLAC.
-Richard (DJ Ricardo)
P.S.: In the next article, we will rip, code and tag a Tango CD.