Analogue Audio as Continuous Data
Music is naturally an analogue format. When a singer sings or you hear a pianist play the piano the resulting sound wave that reaches your ear is continuous. In analogue music formats such as Tape or Vinyl the players reproduce a continuous electronic signal without breaks or interruptions.
Here are some non audio examples of continuous Data;
Height measurements, weight measurements, - Notice how these can be a wide range of numbers
that can increase depending on the accuracy and the detail that you go into.
Digital Audio as Discrete Data
When music began to be recorded into the digital domain it was done in a way that we describe as discretely.
Discrete data contains specific values rather than a wide range of values that are available with continuous data. For example, in the terms of music recorded discretely we measure the electrical signal at set points in time.
Here are some non audio examples of discrete data
People or animals – As these have to be counted as whole individuals
The roll of two dice – There are set numbers available in this instance
Physical money – You can't give some one 25.5pence physically.
The rate at which audio data is recorded digitally is called the sample rate and is analogous to frame rate in moving picture. It is measured in Hertz or Kilohertz (1000 Hertz) which means how many samples are recorded per second.
Similar to moving picture where in order for the picture to be represented usefully there must be 24 frames per second. Sound has it's own rules for comprehensive reproduction – for speech a sampling rate of 8000Hz (8Khz) is necessary and this sampling rate is still used for telephone communications.
Lower sampling rates than 8Khz start making phonemes (Sounds that contain high frequency energy) hard to distinguish from each other. In fact, the lower the sample rate the less higher frequency sounds can be replicated.
Nyquist Theory And Aliasing
This theorum states that a sinusoidal function (sine wave) can be reproduced correctly if the wave is sampled at a frequency that is higher than double the bandwidth. For example; an 800Hz sine wave would need to be sampled at above 1.6Khz.
If sampling is less than this then aliasing can occur. Aliasing is the creation of unwanted audio artefacts when there is not enough sampled information to correctly reproduce the original signal. See the example below;
Diagram from https://zone.ni.com/reference/en-XX/help/371361R-01/lvanlsconcepts/aliasing/
Sample Rate Standards
Within the digital sphere there have become industry sample rate resolution standards such as 44.1khz (44, 1000 samples per second), 48Khz (48, 000 samples per second).
The reason for 44,1Khz is that the average human range of hearing is 20hz – 20Khz and so to accurately reproduce 20Khz we would need to sample at a higher frequeny than double this rate which gets us to 44.1Khz.
In film and video, 48Khz is the current standard. This sample rate was partly chosen as audio had to sync with the standardised video frame rates of 24, 25 and 30 frames per second. A multiple of 600 Hz worked as 600 is easily divisible by these figures.
This is the name for sampling at a rate that is significantly higher than is depicted by the Nyquist Theorem for coherent sound reproduction.
Two oversampling standards are 96Khz and 192 Khz which are multiples of the film and video audio standard sample rate. 96Khz is becoming a standard within the recording studio environment.
Benefits of Higher Sample Rates
- By sampling at a rate above that required from the Nyquist Theorum, sounds that are outside of the human range of hearing are then recorded and reproduced. We would not be able to hear these phantom sounds but this process then introduces the likelihood of more unwanted artefacts made from aliasing. To get around these problems a low pass (anti-aliasing) filter is added and at 44.1 Khz this filter must be very steep and can cause harshness in the high frequency range. With higher sampling rates this filter can be less extreme and produce a more rounded sound.
- Higher sampling rates record more detailed information about the wave form and allow quick transients to be reproduced more accurately.
- Latency and signal to noise ratio are also improved with higher sample rates.