Gods Of Music

Music product guides and reviews.

You are here: Home / gom / The Layman’s MP3 Guide: Installment 2

The Layman’s MP3 Guide: Installment 2

By

The Layman’s MP3 Guide: Installment 2

By Greg McLaughlin

Welcome to the second installment of the Layman’s MP3 guide. This
time around we’ll look at some of the more technical aspects of MP3’s
(and for that matter similar technology). Topics covered in this installment
are:

    • When and why audio compression is necessary
    • Lossy versus lossless compression and how they compare when simply
      modifying a sound file to be smaller.
    • What type of compression is used in MP3’s and its relative effectiveness
    • The primary factors that determine the size/quality of an MP3 file
      and how best to exploit them

Why is compression needed?

The primary purposes of using compression in music files are to save
disk space and to make music content easy to share with others. Uncompressed
sound data, typically stored in Wave (.WAV) files on PC machines, can
fill up a hard drive very quickly, particularly when working with CD-quality
sound.

Suppose you want to store and distribute your latest song without using
compression. At CD quality the wave file will have 16-bit resolution and
a sample rate of 44,100 samples per second. Each sample must be recorded
onto disk, and since there are typically 2 channels (stereo) that means
that 88,200 16-bit values must be recorded per second. Now assume your
song is 5 minutes long: the total disk space required will be 44,100 [samples
per second] * 2 [channels] * 2 [bytes, since there are 8 bits in a byte] * 60 [seconds in a minute] * 5 [minutes]. This is equivalent to 52,920
kilobytes, or nearly 52MB. Storing a large number of these files on any
media short of a monstrous hard drive would be troublesome. Streaming
a file of this size from an artist’s website would be impossible
for the majority of users (for that matter most web hosts won’t even
store files this large).

What Options are available to reduce file-size?

There are three common ways of reducing file size, and the term “compression”
is actually an umbrella term for two of these methods. All three methods
are briefly compared in the chart below:

Reduction Method

Description

Effect on sound quality

Effect on file-size
Non-Destructive (lossless) compression

Examples: ZIP, RAR, ARJ, ZOO

 File compression that, once decompressed, produces
output that is identical to the original. Typically these methods
of compression are general-purpose, not geared to any particular file
format. Files usually must be decompressed entirely before they can
be played (ie no streaming capabilities) 		 None

 

 Minimal. It is rare to achieve even 2:1 compression
with most sound or video data. 
Destructive (lossy) compression

Examples: MP3, RealAudio, Ogg
Vorbis (OGG), Windows Media Audio (WMA)

 File compression that generally relies on removing
multimedia data from the original file -with minimal perceptible change
to the user. Most lossy compression algorithms are geared toward specific
formats (ie MP3 is for digital audio only, not video or text). Files
created with lossy compression are typically decompressed on-the-fly,
meaning they can be streamed over networks. 		 Negligible to severe. Depends mainly on the format
and encoding parameters (covered in later sections of this installment) 		 Moderate to heavy. Depends mostly on the compression
algorithm and its internal setting. With MPEG audio it is currently
possible to attain near-CD quality sound while maintaining roughly
a 12:1 compression ratio 
Wave decimation

Examples: Cutting the sample rate from 44.1khz
to 22khz, converting wave from stereo to mono

 Essentially one can simply change the parameters
of the original wave so that it requires less disk space to record.
Referring to the calculation example above, one could for instance
convert the Wave file from stereo (2 channels) to mono (1 channel).
This would remove a factor of 2 from the file-size calculation, effectively
halving the amount of disk space required. 		 Severe. Lowering the bitrate from 16 bits to 8 bits
will save a lot of disk space, but will make the recording sound like
it came from AM radio – or worse. Similar effects will occur
when the sample rate is lowered. The effects of changing from stereo
to mono are obvious. 		 Moderate. Cutting the sample rate in half will halve
the file size. If that file were then switched to mono it would be
halved again etc. 

From this chart two important observations can be made:

    1. Lossy compression has a distinct advantage over wave decimation in
      that it can achieve heavy compression ratios without seriously damaging
      the quality of the sound.
    1. Lossy compression has a distinct advantage over lossless compression
      in that it can preserve the quality of sound with a much higher compression
      ratio

It is therefore no surprise that many parties ranging from Internet news
services to the mighty Microsoft have invested millions of dollars in
adopting and/or developing lossy audio compression algorithms.

So What Exactly is an MP3?

Mp3 is actually a shortened term for the output created by MPEG 1 Audio
layer 3 (referred to as M1AL3 from here on) – a lossy audio compression
algorithm developed by the Fraunhofer IIS-A in the late 80’s and early
90’s. The original goal of the project was to improve on the 8:1 compression
ratio required by existing algorithms (namely the MPEG 1 Layer 2 standard)
to attain CD-quality sound.

Using quantization (specifically with Huffman coding) and nested iteration
loops, the M1AL3 algorithm transforms digital audio data into a much more
compact form called a bitstream. This process is referred to as encoding,
which was covered in installment 1 of this guide. Before a bitstream can
be heard it must be changed back into standard digital audio, or decoded.
This is what WinAmp and similar programs do on-the-fly when you load and
play an MP3 file.

What determines the size and quality of an MP3?

The size and quality of an MP3 file is affected by several factors, some
having to do with the parameters set in installment 1 of this guide and
others set long before encoding takes place. Bitrate is the most intuitive
and recognized of these factors but does not alone determine the output.
A summary of the more important factors follows

Factor

Description

Notes/Suggestions
Bitrate Mode Bitrate refers to the amount of data from the bitstream
that must be transferred from the medium (internet, CD, etc) to the
decoder in a given period of time. This is typically measured in kilobits
per second, or kbps. The L.A.M.E. encoder offers three modes of encoding
with respect to bitrate:Constant Bitrate (CBR)

The most common of the three modes, CBR encoding keeps the bitrate
at the user-defined value (e.g. 128kbps) throughout the song. This
means that any similarly-sized section of a song requires the same
amount of disk space.

Variable Bitrate (VBR)

Variable bitrate allows the user to choose a number between 0 (highest
quality) and 9 to define the quality of the MP3. The encoder then
tries to maintain this quality by encoding different sections of
the song at different bitrates (e.g. quiet parts of the song and
solos would be encoded at a lower bitrate than louder or more complex
sections). Resulting file-sizes can vary wildly, depending on the
quality setting and the overall complexity of the song.

Average Bitrate (ABR)

A compromise between CBR and VBR, ABR allows the user to set an
average bitrate (and in turn a total number of bits allocated to
the song). The encoder then operates on the same principle as VBR
by using a higher bitrate to encode more complex sections of a song,
only with extra bits “loaned” by the lower
    1. Mp3.com and many other song-hosting portals stipulate that CBR
      be used to encode music. Uploads of ABR and VBR songs are usually
      rejected. Check with your hosting service to make sure you are
      able to upload ABR/VBR songs before using these modes.
    1. A good rule of thumb is to remember that constant bitrate =
      variable quality throughout the song and vice-versa
    1. Variable bitrates tend to work best on music with varying dynamics,
      like Classical and Jazz. Most commercial and Pop music is heavily
      compressed, leaving little dynamic difference throughout the song.
      As such there will be very little increase in quality when using
      ABR/VBR
    1. As VBR has only quality-based restrictions files can get very
      large depending on the song’s complexity. Avoid using this
      mode if you are concerned about disk space or the ability to stream
      your music properly on low-bandwidth mediums (like a 56k modem
      connection)
    1. If you are absolutely concerned about quality avoid using CBR.
    1. ABR provides a nice compromise by allowing for a varying bitrate
      while maintaining a predictable file size. This mode might be
      your best choice for home listening.
    1. Be advised that some playback software/hardware has trouble
      scanning through VBR/ABR-encoded MP3’s. Experiment before
      you commit.
Bitrate Value After selecting a bitrate mode, a bitrate
value must be set (except with VBR). Typical values for bitrate range
between 64kbps and 192kbps, but can be as low as 16kbps or as high
as 320kbps. Using higher bitrates will result in larger file sizes
and will require media with faster data transfer rates. Bitrate values
are also correlated to the frequency response of the output. The bandwidth
for some of the more commonly used bitrates are as follows:

32kbps 7.5kHz
96kbps 11kHz
128kbps 16kHz
160kbps 20kHz
256kbps 22kHz

 
    1. Mp3.com and many other song-hosting portals require a bitrate
      of 128kbps. Check with your hosting service for restrictions.
    1. If you ripping directly from CD or encoding a studio master,
      192kbps will produce near-perfect sound. 160kbps is adequate for
      well-pressed vinyl recordings while 128kbps is adequate for vinyl
      of lesser quality.
    1. If you are absolutely paranoid about lossy compression, or
      cannot avoid using lossy compression while editing/transporting
      a studio recording, consider using 256kbps or even 320kbps.
    1. If you plan to play your MP3 files in a DVD player or portable
      MP3 player be sure to check the hardware specifications. Some
      portable players will not play all bitrates (particularly bitrates
      below 64kbps). A large number of DVD players will not play MP3
      files encoded with bitrates higher than 128kbps.
Channel Mode Channel mode simply refers to the number of channels
used during encoding. Aside from the usual stereo and mono selections
there is an additional mode called Joint Stereo. In this mode the
encoder uses “mid” streams (sound that is the same through
both channels) and “side” streams (sound appearing only
in one channel) instead of two separate (and possibly identical) streams.
    1. Joint Stereo is ideal when low bitrates (ie less than 128kbps)
      are required but stereo output is still desired
Original recording No MP3 can sound better than the original recording.
Wave file parameters that must be considered before encoding include:Sample Rate

The most common sample rates are 22kHz, 32kHz, 44.1kHz, 48kHz and
96kHz

Bit Resolution

Sound files can be recorded in 8, 16, 24 or 32 bits.

Number of Channels

Typically sound files are in 1 (mono) or 2 (stereo) channels but
with the rise of DVD audio and high-fidelity gaming it is not uncommon
to encounter recordings with 4 or 5 channels

The red book audio standard (upon which the audio CD is based)
requires a 16-bit stereo recording sampled at 44.1kHz. DAT recordings
are typically 48kHz/16-bit, while it is common for studio recordings
to be 96kHz/24-bit.
    1. Be mindful of the sample rate when deciding
      what bitrate to encode with. The Nyquist Theory states that
      a waveform must be sampled twice in order to get a true representation.
      As such you should divide the sample rate of the original recording
      by 2 in order to get the best possible frequency response (referred
      to as the Nyquist rate). The Nyquist rate can be matched with
      the frequency table above (under “Bitrate Value”)
      to determine the optimal bitrate.
    1. Don’t waste high encoding bitrates on 8-bit recordings.
      The quality of 8-bit recordings is similar to that of AM radio,
      which suggests a bitrate of at most 96kbps. Similarly there is
      no dividend from encoding a mono recording in stereo.
    1. If you want to encode a 4 or 5 channel recording without losing
      or combining any channels (ie preserving the surround sound) there
      is an unofficial “multi-channel” MP3 format. This format
      is supported almost exclusively by movie-players, like Cyberlink’s
      PowerDVD, and the format itself is usually created as the sound
      portion of an AVI file. Those seriously interested in multi-channel
      recording or surround-sound for movies should enquire about the
      multi-channel AAC format.

 

Putting this to work in CDex

Most of the factors mentioned thus far (and more) can be manipulated
in CDex through the configuration window, which can be selected either
through the menu for by clicking on the “setup” button along
the right panel. Specifically the bitrate mode, bitrate value and channel
mode can be selected under the “Encoder” tab. What CDex cannot
change is the quality of the original sound file. This can be manipulated
to some extent through software packages like Sound Forge, Cool Edit or
Goldwave, butthe best advice one can give would be to make sure the initial
recording is as flawless as possible.

What’s Next…

As most net-savvy folks know, MP3 is not the only kid on the block when
it comes to sound compression. The inventors of the original MP3 format
have now released MP3Pro, which promises improved sound quality at standard
bitrates. Microsoft has also entered the multimedia race with their Windows
Media format which, among other things, promises improved security to
prevent unauthorized duplication (MP3 contains no inherent copy protection
scheme). On the opposite end of the copyright/freedom argument is Ogg
Vorbis, an open-source project that boasts multichannel output and no
royalty fees whatsoever. How do they compare to MP3? Find out in the next
installment…

Filed Under: gom