In the real world sound is analogue, not digital. There are no distinct steps between variations in air pressure produced by a sound source. In the days of analogue recording, the sounds were recorded as a continuously changing electrical audio signal, usually onto magnetic tape provided in a cassette or on a tape reel. Generally, reel-to-reel tape decks were of better quality than their cassette deck counterparts.
A major drawback of the analogue system was that interference from the recorder’s electronics and storage medium created audible background hiss or hum known as ‘noise’. It was more noticeable in cheap, domestic recorders, and less noticeable in expensive professional equipment.
Signal and Noise
The relationship between the recording equipment’s ability to provide a good, clean recording of what was wanted (called ‘signal’) to it’s ability to minimise the ‘noise’ that wasn’t wanted, was known as its signal to noise ratio, expressed in decibles (db). Cheap gear had a low signal-to-noise ratio; higher quality gear had a higher signal-to-noise ratio.
Some manufacturers offered built-in electronic noise reduction circuitry such as Dolby and DBX. To get the best from those systems, special and expensive recording tape had to be used. The machine used for playback also had to support the same noise reduction method used in the recording.
With the advent of digital recording and reproduction, popularised by CDs, the analogue sound waves are analyzed by the recorder’s circuitry and converted into discrete values called ‘samples’. Generally, the more samples taken each second and the finer the sample (its resolution), the better the reproduced sound will be. Unlike analogue recording, digital recording and processing is not subject to the same sort of noise interference. The result is a much cleaner reproduction.
Two major aspects of digital recording and sound quality:
Sample rate and resolution
The two graphs below show a representation of an analogue sound in red, and the digital reproduction in green; the idea is to reproduce the analogue red line as faithfully as possible with the green samples. The finer the digital samples (i.e looking less ‘blocky’ in the diagrams), the better the reproduced sound will be.
Below: More samples per second (higher sampling rate) reproduce the analogue sound (red line) more faithfully
Below: Less samples per second (lower sampling rate) reproduce the analogue sound less faithfully.
CD audio is sampled at 44.1 Khz. This is because 44.1 Khz is roughly twice the maximum frequency most humans can hear… in their younger years! Most people can hear sounds between 20hz and 20Khz, but this range diminishes with age.
Video audio is typically sampled at a higher rate of 48Khz, but higher rates are possible, depending on the software and hardware combination of the equipment.
Resolution determines the accuracy of the vertical digital measurement of the analogue wave. Together with how often it is measured (the sample rate), this determines how closely the overall sample adheres to the original analogue sound wave.
On the graphs below, resolution is shown by vertical divisions. The divisions are points where the analogue wave is measured at a particular point in time, and given a value to the nearest, and next lowest point (because this is digital, no fractional values are allowed). In the representation below, only the value of the green blocks count; fractional values (the grey blocks) can not be used.
Imagine an individual sample (like that in the above diagram) is like a stack of toy building blocks; you have to measure how many blocks high an analogue waveform is at that point – and you’re only allowed to give your answer in whole blocks. You can’t say ‘so many blocks and a half, or so many and a third’.
That limits your accuracy to the number of whole blocks below the red analogue line. But if you replace large blocks with a stack of smaller ones, the more accurately you could measure the height of the waveform. In this analogy, big blocks would represent a low resolution sample (low quality); smaller blocks represent a higher resolution sample (better quality).
8 bit resolution, 16 bit resolution… what the?
Everything a computer does, such as sound sampling, revolves around manipulating data represented by numbers, using the binary number system where there are only two digits (instead of our usual ten digits, 0 – 9, in the decimal system).
Binary digits are called bits. There are only two of them – one and zero.
This numbering system works well with the electrical nature of computers and can be easily represented by a particular voltage (for a one), or a lower voltage (for a zero).
So, what do bits have to do with sampling resolution?
As we saw earlier with the analogy of using green wooden blocks to measure the height of an analogue waveform at a specific point in time, the more blocks, or the more measuring points we had, the finer the accuracy and better the result.
In computer speak, the number of individual green blocks you had piled up would be expressed as a the number of bits of the sample. As we saw earlier, the smaller the blocks, the more you could pile up and the finer your measurements could be, because you have more measuring points.
In the same way, the more bits you have, the finer your sample resolution can be because you have more vertical measuring points, meaning better quality. It’s like having more, smaller green blocks, which gives you more measuring points.
When talking about sampling bits, the total number of vertical measuring points (the sample resolution) is a mathematical power of the number 2 (because in binary, there are only 2 values: ‘0’ and ‘1’).
8 bit: 28 = 256 total vertical measuring points per sample
16 bit: 216 = 65,536 total vertical measuring points per sample
24 bit: 224 = 16,777,216 total vertical measuring points per sample
It’s just like Photoshop… more bits = a larger number of colours available in the palette.
Good digital audio quality takes high processing power!
When sound sampling took off in the early 1980s, comparatively low computing power meant that it was a tough job for most PCs to crunch the numbers for processing sound sampling. To balance sound quality with acceptable processing time, low resolution 8-bit sampling was common. As computing power increased, higher resolutions and increased sampling rates became practical.
CD audio became a standard from the early 1980s. The engineers decided on a standard sampling resolution of 16 bit. Samples were taken at the rate of 44,100 times per second (44.1Khz).
The graph below shows how a low resolution of say, 8 bit, could compare with 16 bit.
Combine higher resolution with higher sampling rate = better quality again
Generally, sampling rates and resolution are limited by a combination of the software you’re using and the ability of the capture hardware, such as what ever capability is built in to your PC, camera, audio recorder, phone etc. The overall quality will also depend on the quality of other equipment such as microphones and connection leads.