1. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology IIIDigital Audio III.5 (F Oct 30) MP3 and other digital audio file formats

2. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology We deal with computer file formats for saving digitized audio (= acoustical) data Such formats contain not only the audio-related data, but also a number of additional technical information about the encoding method of the raw data, and also data concerning the composer and other poietic information. This additional information can be complex in function of the actual method. We shall see for MP3 that the encoded data stream is in itself a virtuosic achievement invented to enable playing a piece starting at any time moment of the piece. (Remember that Fourier transform needs all time moments!) The additional data strongly depend on compression methods invented to reduce the saved data with respect to the original audio data. Compression is the rate original data/file data. For example, if the original data is a 10 minute CD data and the file needs 10MB, what is the compression? Answer: 44.1×10 3 Hz × 16 Bit × 2 channels × 10 min × 60 sec / 10 7 × 8 Bit ≈ 10.58

3. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology There are two compression types 1.Lossless compression This means that the original data can be completely reconstructed from the compressed 1.Lossless compression This means that the original data can be completely reconstructed from the compressed ones. 2.Lossy compression This means that the original data cannot be complely reconstructed from the compressed ones. First Example of lossless compression: TV pixels in Black&White 100 pixels 10 pixels 1000 bits uncompressed B1*W1*B2*W2*… B, W = 1…. (≤ 10 bits for 2 10 = 1024) numbers of successive uninterrupted black or white pixel sequences RLE = Run Length Encoding used in MP3 RLE = Run Length Encoding used in MP3

4. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology There are two compression types Second Example of lossless compression: Huffman compression, used in MP3 Frequency Value

5. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology There are two compression types 1.Lossless compression This means that the original data can be completely reconstructed from the compressed ones. 2.Lossy compression This means that the original data cannot be complely reconstructed from the compressed ones. First Example of lossy compression: ADC (Analog to Digital Conversion, as opposed to DAC) from analog to CD with 44 100 Hz and 16 bit amplitude digitization CD with 44 100 Hz and 16 bit amplitude digitization 0100001010100100010001010…0100001010100100010001010… rate = ∞

6. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology There are two compression types Second Example of lossy compression: Quantization in MP3 compression: In the 16-bit representation of amplitude, the five last bits are neglegted: So, for example 1011 1100 0101 1110 is transformed to 1011 1100 0100 0000 and then to the 11 relevant left bits 1011 1100 010 In MP3, there is a much more important lossy compression that is generated by criteria from psychoacoustics. We shall come back to this after a first presentation of MP3 and its history.

7. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology Important terminological and historical facts MPEG (Moving Pictures Experts Group) is the code name for the standardization group ISO/IEC JTC1/SC29/WG11 (Int. Standards Organization/Int. Electrotechnical Commission). It was created in 1988 to generate generic standards for encodings digital video – and audio data. MPEG-1 is the result of the first work phase of the group and has been established in 1992 as standard ISO/IEC IS 11172. It contains Layer-1, Layer-2, Layer-3, which means three operation modes with increasing complexity. By MP3 one denotes Layer-3 of MPEG-1. MPEG-2 Advanced Audio Coding (AAC) is the result of the second work phase. In enhances Layer-3 in many details. We are not going to discuss this phase here.

8. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology Important terminological and historical facts (cont.) The development of these compressed audio formats goes back to research by Dieter Seitzer since 1960, at that time working at IBM, and his student Karlheinz Brandenburg, who is above all responsible for the psychoacoustical compression methods. It is remarkable that Seitzer’s patent was rejected in 1977, it was however accorded in 1983, but was then suspended because of missing interest from the industry! MP3 is above all based on research and development by Karlheinz Brandenburg at the Fraunhofer Institut für Integrierte Schaltungen (IIS) in Erlangen. It is an open standard, but it is protected by many patents (≥13 US patents, ≥16 German patents). We shall discuss legal aspects at the end of this topic.

9. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology Important terminological and historical facts (cont.) MP3 includes these options: Mono and stereo, in particular joint stereo encoding for efficient combined encoding of both stereo channels. Sampling frequencies inlude 32 kHz, 44.1 kHz, 48 kHz, for MPEG- 2 also 16 kHz, 22.05 kHz, 24 kHz and for MPEG-2.5 (Fraunhofer-internal extension) also 8 kHz, 11.05 kHz, 12 kHz. The compression’s bitrate (= the bits traversing the audio file per second) goes from 32 kbits/s (MPEG-1) or 8 kbits/s (MPEG-2) up to 320 kbits/s. For MP3 the bitrate can even vary from frame to frame (a frame is the unit package in MP3, we come back to it later) and, together with the so-called bit-reservoir technology, allows a variable as well as constant bitrate. Moreover, compression MP3 has the advantage to be platform-independent. This is also a strong reason for its popularity.

10. Guerino Mazzola (Fall 2015 © ): Introduction to Music Technology List of Audio File Formats uncompressed WAV Microsoft (extends RIFF = Resource Interchange File Format) AIFF Mac (Mac’s version of WAV) AU Simple SUN audio file format, UNIX sound standard PCM (Pulse Code Modulation) Raw digitao audio data, usually stored in WAV compressed lossless ALAC = Apple Lossless, stored in MP4 container, e.g. in iPods Monkey’s Audio, better than Shorten or FLAC and WavPack MPEG-4 (multimedia) SLS (Scalable Lossless), extended to AST (Audio Lossless), DST (Direct Stream Digital) based on Apple’s QuickTime Shorten, used to compress CD files TTA (True Audio) compression on multichannel 8, 16, and 24 bit data of wav files, GPL, cross-platform WavPack 8, 16, 24, and 32 bit wav files, also surround sound and high frequency sampling rates WMA Lossless developed for archival purposes lossy AAC (advanced audio coding) a successor of MP3, complies with MPEG-2 and MPEG-4, up to 96 kHz ATRAC (adaptive transform acoustic coding) Sony, developed for MiniDisc 1992 MP3, to be discussed Musepack, also known as MP+, for PC, Linux, and Mac OSX, violates patents from MP3, similar to MP3 Ogg Vorbis, popular in free software circles WMA (Window Media Audio), part of the Windows Media framework, originally competitor to MP3