Tech for Teaching

Audio Formats

Audio is stored in a computer as a list of samples. Each sample is a measurement of the signal’s voltage at a specific moment in time. The signal will be sampled many thousands of times a second. When the files is played back these measurements are used to generate the original voltages. The signal is then smoothed out a little and sent to the speakers or headphones.

A digital audio recordings at the moment of digitization is characterized by the following parameters:

• The number of bits which are used to record each voltage measurement. For CD’s, video DVD’s, and most computer files 16 bits must be recorded.
• The number of “samples” or measurements which are recorded per second. 8000 samples per second is used for telephone calls. CD’s use 44,100 samples per second. A video DVD’s soundtrack (most often) records 48,000 samples per second.
• The number of channels recorded. Spoken word recordings are generally monaural which means that they record one channel. Stereo recordings have two channels, one for the left speaker, one for the right. Suround sound will have more channels for additional speakers which may be in the rear or center or for a subwoofer.

CD audio is not compressed, but audio files shared on the World Wide Web and through streaming services are compressed so that they can be downloaded more quickly. Audio compression methods can be characterized as either lossless or lossy.

Lossless and Lossy Formats

Because digital audio files record thousands of samples per second, they get big very quickly. Ten minutes of music in CD-quality stereo takes 100 megabytes.

Audio files, like other digital files can be compressed. A general purpose data compression algorithm such as DEFLATE could shrink an CD-quality audio file to 80 to 90 percent of its original size. An algorithm such as FLAC which is designed specifically with the patterns of audio data in mind can shrink it by 50 to 70 percent of its original size. During decompression the original audio data is reconstructed perfectly. For this reason algorithms such as DEFLATE and FLAC are described as lossless.

Losslessly compressed audio files are still pretty big. Much higher compression can be achieved if we are willing to accept a slightly less than perfect reproduction of the original audio. If this is done carefully, the audio file can be reduced to less than ten percent its original size while sounding so close to the original that few can hear the difference.

If you are making and distributing audio recordings, it is best to use lossless formats such as WAV or FLAC for your master copies. When you are ready to distribute the recordings, you can convert them to a lossy format such as MP3, AAC, or Ogg so that they will download more quickly. If you later have to make changes, you should go back to the original lossless masters. If you edited the lossy file instead and recompressed them, you would lose a little quality each time.

More information about digital audio concepts and digital audio compression can be found in the article Digital audio concepts at Mozilla.org.

LPCM WAV Files (Lossless)

WAV is an audio container format developed by Microsoft and IBM. WAV files with LPCM format audio inside are frequently used as a master format. These files are uncompressed, so they can be quite large.

As of January 2020 95% of web browsers can play LPCM WAV files. This is convenient for testing, but on a public website it is better use MP3, AAC, or Ogg files.

FLAC Files (Lossless)

FLAC files contain audio which has been compressed in a way which allows full reconstruction of the original audio. Because nothing is thrown away, the space saving is seldom more than 50%. FLAC is a popular format for high-quality music collections.

As of January 2020 over 90% of browsers will play FLAC. FLAC files are smaller than PCM WAV files, but are still overkill for almost all applications.

MP3 Files (Lossy)

MP3 stands for “MPEG-1 layer 3 audio”. MPEG-1 is a now-obsolete standard for compressing videos. MP3 is the third of three options for encoding the audio track.

MP3 files compress CD-quality audio to about 10% of its original size by throwing away details which most people cannot hear. The degree of compression is expressed in bits per second. 128 bits per second produces acceptable music reproduction. 64 bits per second is frequently used for voice recordings. If an MP3 file is compressed too heavily, a ringing noise that sounds a bit like squeaking mice can be heard behind the speaker’s voice.

Like many other lossy compression formats, MP3 defines how the file should be decoded and played, not how it should be encoded. The author of the encoder has leeway to choose between various possible encodings. MP3 encoders have gotten better over the years. The opensource LAME encoder is widely acknowledge as the best. You can download and use it directly, or you can use a tool which includes it such as FFmpeg or Audacity.

Like other MPEG standards, MP3 was covered by patents, but it appears the last of these expired in 2017. This means that this already very popular format can be deployed everywhere. Though it has been surpassed technically, its ubuquity gives it a decided advantage. As of January 2020, about 97% of web browsers can play MP3 files.

Advanced Audio Coding (Lossy)

The Advanced Audio Coding (AAC) is an MPEG standard which is intended to replace MP3. The most popular form, AAC-LC, can provide near-CD quality stereo sound at 128 kbps (about 30% less than MP3 would require). The AAC-HE v1 and provides medium quality stereo audio (suitable for Internet radio and podcasts) at around 64 kbps. AAC-HE v2 provides the same at about 32 kbps.

Though AAC is still covered by patents, as of January 2020, about 97% of web browsers can play AAC audio as long as it is in an MP4 container. Because of the patents, the Firefox web browser can play AAC only if the underlying operating system has an AAC decoder it can use.

OGG Vorbis (Lossy)

When in 1998 it became known that the MP3 was covered by patents, programmar Christopher Montgomery created and a container format called Ogg and an audio encoding format called Vorbis. Today the Vorbis specification is maintained by the Xiph.org Foundation. Thos less popular than MP3, Vorbis is techically superior and so an Ogg Vorbis file can be smaller than an MP3 file while keeping the same audio quality.

As of January 2020 about 90% of web browsers will play Ogg files with Vorbis audio. Internet Explorer will not and Safari may not depending on whether a Ogg Vorbis decoder is installed in the operating system.

Opus (Lossy)

Starting in 2010 the Xiph.org Foundation, together with Skype and Mozilla created a new audio coding standard called Opus. This new format was published as Internet Engineering Task Force (IETF) standard RFC 6716.

Opus is intended as a successor to Vorbis and Speex, another audio encoding from the Xiph.org Foundation. Opus produces excellent results in a wide range of applications from high fidelity surround audio to Internet telephone calls. As of January 2020 it is considered superior in quality to all other popular lossy audio codings including MP3, Vorbis, and AAC. A 64 kbps Opus file has about the same quality as a 128 kbps MP3 file.

Opus is one of the two formats (the other is Vorbis) which may be used in Webm video files. It is used extensively in videos streamed from Youtube and other sites.

As of January 2020 about 90% of web browsers will play Ogg files with Opus audio. About 90% of browsers will also play Opus audio from a Webm video file.

Quality Comparison

In listening tests with the best encoders the various audio formats are are frequently ranked in this general by ascending quality:

1. MP3
2. AAC and Vorbis
3. Opus

At bitrates above 128 kbit/s few people can tell the difference.