Breaking it down
How spectral synthesis works
So what exactly is spectral synthesis, and how does it work? Well, we mentioned spectroscopy on the preceding page, and it represents the simplest way to describe the techniques we’ll be exploring in our tutorials. The most familiar form of spectroscopy will be familiar to anyone who has ever shone a light through a prism or, for that matter, anyone who’s ever seen the cover of Pink Floyd’s The Dark Side Of The Moon!
As any school kid knows, a prism breaks up incoming light sources and splits out a rainbow of coloured light. Indeed, a rainbow is nature’s own spectrographic display, with atmospheric moisture acting as a prism. It takes in the light of the sun and separates it according to the length of the waves that make up that light. Wavelengths are also referred to as ‘frequencies’.
Sounds can similarly be broken down into their constituent frequencies by using one of two techniques: band-pass filters or Fast Fourier Transform (FFT).
The former is what drives the most familiar incarnation of spectral synthesis: the vocoder. Here, a complex incoming signal (the modulator – say, a voice) is plumbed through multiple band-pass filters, each allowing a specific frequency – or range of frequencies – to pass through. The amplitude of the signals produced at the output of each filter is used to drive the amplitude of the carrier signal – usually a synth tone.
The concept behind Fast Fourier Transform is a little bit more difficult to understand. It’s based on the idea that any sound may be broken down into a collection of individual sine waves, each with its own amplitude and phase. There are a number of algorithms that can be employed, but the important thing for you to remember is that we are left with a visual representation of that sound as it occurs over time. This will be our spectrogram.
The resulting spectrogram may take any one of a number of forms, but it will always tell us what frequencies are in play, how loud they are, and when they occur in time. Commonly, the vertical axis of the image will represent the frequency, while the horizontal axis will indicate the passage of time. The amplitude of the frequencies in play is represented by the colour displayed and/or its intensity.
Analysing the spectrum of a sound allows for a deeper understanding – and deeper tweaking – of your audio