Audio compression has been around for as long as recorded sound, in its simplest form it can be a recording engineer watching the level meter and continually adjusting the recording level to fit within the medium; or a automated action which can take account of a range of factors. The objectives of audio compression are 1), to reduce the dynamic range of a signal so that it will fit into a transmission system without distorting; and 2), to improve the audibility of a signal so that it may be clearly heard. AM radio is a good example of both of these objectives being met by good compression. The medium wave band will only support around 30dB dynamic range and most listeners wish to hear their radios in relatively noisy situations.
The important thing to note is that this type of compression does not remove any of the information and reversing the process can restore the original information. For example, the Dolby noise reduction process used in tape cassette systems first compresses the signal then expands it back to its original form. All these compression systems operate in the analogue domain.
With the introduction of digital technology have come many low bandwidth formats which use digital compression (or bit reduction) to make the signal fit into the new format. Unfortunately, in most systems such as MP3/DAB/Mini Disc, the data loss introduced cannot be replaced and therefore some of the original is lost forever. (Note, there are some so-called lossless digital compression systems where there is no data loss, but these achieve much lower compression ratios than is available from MP3 etc)
To put this into perspective a CD uses 16 bit coding and requires around 2M/bit per second to store up to 74mins of recorded good quality sound. On the other hand, the best BBC Radio DAB signals restrict their bandwidth to 192K/bits per second. This means that around 90% of information is lost compared to a CD which is judged to offer reasonable quality sound. Or put another way, DAB is providing a signal whose resolution has been reduced to about 2bits.
When audio is digitally compressed, a clever trick is played. Before encoding, the signal is analysed according to psycho-acoustic rules and the areas of music that the brain may not detect are removed. This is done in such a way as to preserve the original frequency response and volume so on an A/B test you do not notice the degradation of the original signal. Only after a period of listening to the bit-reduced signal do you begin to notice that something is wrong. The improvement in quality when you revert to a CD can be very dramatic. This state of affairs is confusing because when you make the transition from CD to the compressed source degradation seems very slight.
If you need further convincing try listening to just the difference channel (put a loudspeaker across the L&R speaker connections) of a mono recording or broadcast. Instead of hearing nothing, or at worst very slight cross talk, you will hear a very phasey and distorted signal that increases in volume with the main channels. This simple test will give you an insight into the artefacts that are concealed in the compressed music. Incidentally, this problem gets worse as bit rates are reduced. Of course, if you do this with a CD the artefacts are non-existent.
Unfortunately, these compressed signals are now everywhere. Digital broadcasts of all types are bit reduced and the organisations which supply the service are incapable of recognising the loss of quality. Instead they wish to provide even more channels at poorer quality. It is clear that we were lucky to get 16bit uncompressed CDs simply because the coding equipment was not available at the time they were developed.
Ganymede Test & Measurement
14th January 2002