Simon Peppercorn probes the process of sound production.
Let's start with some basics. Sound comes in the form of a continuous analog wave, with a frequency range, which has an infinite number of points. The shape, length, amplitude etc. of this wave, determines exactly how it will sound. A digital medium, however, does not have an infinite number of points, and is made up of a number of 1s and 0s.
To get this analog signal onto a digital medium, the signal is split into thousands of pieces (samples) by an 'analog-to-digital converter' (ADC). Information about each piece, such as its shape and its relationship to other pieces, is stored. The more pieces the signal is split into, the more information that can be understood about it. The rate at which these samples are collected per second is called the sampling rate, and is generally expressed in kHz (thousands per second).
Using a process called 'quantisation' the converter looks at the maximum height or amplitude of a sample and splits it into bits. By increasing the number of bits, it can be reproduced with increasing accuracy.
During playback, a 'digital-to-analog converter' (DAC) creates an analog wave, based on all the information it has collected and digitally stored, about the pieces. Unfortunately, this wave will never be an exact replica of its original. It will be very close, but some of the original signal would have been lost.
The way it was
However, sound reproduction wasn't always so fancy. Electronic sound devices originally produced a purely synthesised sound by creating its own complex waveform. This was achieved by mixing a pure sine wave, called a 'carrier' with a second wave, called a 'modulator'. By interfering with the shape of these waveforms, variation in the sound could be achieved, creating a crude version of the instrument each sound was meant to represent. This process was known as 'frequency modulation' (FM).
It was discovered that by merging a number of different carriers and modulators, with careful manipulation, even more complex waveforms were achieved. This added depth and timbre to the artificial sounds, and a much closer approximation of real life instruments could be created. However, the sounds were still very artificial.
In the late 60s, the Beatles, the Beach Boys and other chemically affected musicians started producing some cutting edge sounds, using FM synthesis combined with traditional instruments and recording techniques. This trend continued through the disco era and into the 80s. If the name Harold Faltermeyer means anything to you, then you know the type of sounds I am talking about.
The later development of 'wavetable synthesis', meant that PC sound cards, electric keyboards etc. could now create realistic sounding acoustic instruments from electronic devices. Wavetable synthesis involves the use of 'samples', (similar to the analog-to-digital conversion process) which are miniature multiple recordings of actual sounds, converted to a waveform.
As revolutionary as this was, it actually presented a few problems.
The realism and quality of the sound was impacted by things such as the recording techniques of the samples and the nature of the compression used when storing them. The more samples taken when recording a particular instrument, the more accurate its reproduction. However, in the days when 8MB of RAM was all your computer would ever need and sound cards and CPUs didn't have the grunt found in today's equipment, heavy compression was required to fit the samples into memory, and precious CPU load was needed to produce the required output. This ultimately affected the quality of the reproduced sound.
Stop it, that Hz!
The compact disc was developed in the 1970s. Seventy-four minutes of audio on a standard 120mm disc was achieved by using a sampling rate of 44.1kHz, with 16-bit quantisation. The later development of DVD allowed a much higher sampling rate.
However, for reasons that are still not completely understood by us-who-are-not-driven-by-the-marketing-dollar, the standard sampling rate for DVD was set at 96kHz. Indeed, it would make more sense to use a sampling rate of 88.2kHz. This is double the sampling of the CD. It would be easier to achieve, and simpler to filter back down to 44.1kHz, for those without the higher end sound equipment. In fact, achieving 96kHz is quite complex, involving mathematical equations that would have had Einstein reaching for the paracetamol.
With the human ear detecting sounds from about 20Hz to 20kHz, it would be rare for you to hear any noticeable difference between 44.1kHz and 96kHz recordings, yet DVD Audio can produce sound from 4Hz to 100kHz. But getting 24-bits to wobble 96,000 times per second is so tricky that it isn't always done properly, often with results that sound worse than 44.1kHz.
At the end of the day, it is commonly believed that the reason 'they' chose 96kHz was so that DVD technology was truly set apart from previous standards, making it a 'new' technology, not merely an upgrade. In other words, a marketing decision more than a technical one.
Still, nothing is perfect.
When you push a signal through all the components that make up your sound hardware, various errors are created. These errors can be defined as linear distortion, non-linear distortion and noise.
'Non-linear distortion' errors (deviations in the signal) are those which essentially create signals at frequencies within the signal that were not present in the original input.
'Harmonic distortion' is a type of non-linear distortion error, producing frequencies which were not present in the input signal. These new frequencies, known as 'harmonics', are frequencies at integer multiples of their input signal. Whether a harmonic is audible is dependant on its frequency and amplitude.
Maths freaks will love this one: 'Total harmonic distortion' (THD) is measured as the square root of the sum of the squares of the amplitude of each of the harmonics. 'Linear distortion' errors are those that do not actually create frequencies within the sound that were not already present in the original signal. Instead, these distortions affect the timing, size and relationships of the various frequencies within the signal.
You often see 'frequency response' listed in the specs of sound hardware. This is a form of linear distortion, and can basically be described as the ability of the device to respond to variation in pitch without affecting the relative loudness. 'Frequency response' errors are the most obvious ones in terms of what your ear can detect, and is specified by indicating a particular frequency range and its tolerance. Simply put, the wider the frequency range, and its given tolerance, the better the sound device is performing.
Both linear and non-linear distortion is related directly to and affects the output frequency of audio. Noise is a different type of error, in that it has nothing to do with the sound frequencies and is specifically not related to the input signal at all.
An important factor, when working with sound is the 'signal-to-noise ratio' (SNR), referring to the level of sound that can be produced as a ratio to the distortion or noise (not the intended sound) it produces. Noise or distortion is created by errors or impurities added to the signal as it moves through the circuitry of the device. The better the components used, the less noise created.
Noise/distortion can be measured by using analysers, which separate the errors from the signal. These analysers use a fast Fourier transform (FFT) technique, breaking down complex
Issue: 137 | June, 2012