In conjunction with learning MIDI, we will be learning some simple analog synthesis to produce sounds for the final project. I’m not going to go too far into the history of analog synthesis, other than to mention that when we talk about today, we are actually talking about virtual analog synthesis. Virtual means that we are recreating the techniques and electronic hardware via digital software packages. If it happens in our computer, with software, there is nothing really analog about it. It is a digital system imitating an analog system.
digital audio processing
Up to now, we’ve been starting with sample digital audio recordings that we have manipulated in various ways to create transformed sounds. The transformations have themselves taken on a compositional aspect, as we lengthen things, shorten things, transpose things, etc., to produce music dependent in some way on the original sound sources.
synthesis in general
Synthesis starts with elemental sound components – waveforms – combining and processing them to create musical sound that is not dependent on any prerecorded sound. The two main synthesis techniques we will be using can be compared to painting and sculpture. Additive Synthesis starts with the most basic elements and creates sounds by combining these elements. It is like painting, in that every color must be added to a blank page, primary colors combined with other primary colors to produce new colors, drawn onto the page to make an interesting picture. Subtractive Synthesis starts with a rich block of sound, like sculpture starts with a large piece of material, and pares away (carves out) parts to find an interesting subset of sound.
Our basic elements of synthesis are:
oscillators – envelope generators – filters – modulators
Oscillators create fluctuating amplitudes over time, generally in basic geometric shapes or randomly. These basic shapes are referred to as elemental waveforms. The name of each waveform comes from its amplitude shape over time, and each waveform has its own unique spectrum**.
**For our purposes, we will refer to frequency content that is integer related to the fundamental, as well as the fundamental, as partials. You know about the overtone series. The overtone series starts with a fundamental, with the first overtone occurring next (the octave above the fundamental). When you use partials to describe the overtone series, you start counting with the fundamental as the first partial. The first overtone is the second partial, and so on. The difference is slight but important in understanding the amplitude relationships in elemental waveforms.
Sine waves are the simplest waveforms, containing only the first partial (fundamental frequency). Sine waves are usually the basis for additive synthesis, because if you have enough sine waves with control of frequency and amplitude for each, then you can theoretically recreate any sound. The issue with additive synthesis is the usually massive amount of control needed to produce interesting sounds.
Noise exists at the other end of the continuum from sine waves. White noise is a random fluctuation that produces all frequencies at equal amplitude. Since there are more frequencies as you progress upwards through the pitch range, white noise sounds unbalanced towards the high pitch range.
Sawtooth waves contain all partials with the relative amplitudes of 1/pn (pn = equals partial number). The first partial is 1/1, second is 1/2, third, 1/3, etc.
Square waves contain all the odd partials with relative amplitudes of 1/pn. 1/1, 1/3, 1/5, etc.
Triangle waves contain all the odd partials with relative amplitudes of 1/pn-squared. 1/1, 1/9, 1/25, etc.
We will generally use rich waveforms (sawtooth, square, triangle, and noise) as our sound sources, and filter them to produce interesting results.
Filters are named according the frequencies they let pass unchanged. The most common types are low pass, high pass, band pass, and band reject, spelled in a variety of ways (lopass, hipass, etc.)
Low pass filters have a cutoff frequency, allowing frequencies below the cutoff to pass unchanged, and frequencies above the cutoff to be progressively reduced in amplitude. Low pass filters are our most common tool for creating sounds that respond like natural sounds, as increased amplitude usually involves adding more frequencies to a sound. We can control the cutoff frequency to increase when the amplitude of the sound increases.
High pass filters perform the opposite function of low pass filters. They are most useful for percussive sounds, especially sounds like cymbals, that have a lot of high frequency content without low frequencies.
Band pass and band reject filters have a center frequency. Frequencies around the center either pass through (band pass) or are rejected (band reject, or notch) progressively as you move out in both directions from the center frequency. Band reject filters are mostly used for noise reduction.
Analog filters generally have a slope designation that indicates how much amplitude reduction is applied as frequencies move away from the cutoff. The gentlest slope is 6 dB per octave, meaning that one octave away from the cutoff is 6 dB less in amplitude, two octaves 12 dB, etc. You will find 12 dB and 24 dB slopes in most analog synths. Band pass and reject filters add both sides, meaning the minimum slope is 12 dB. While it may seem logical to use the steepest slope available to you, doing so can create unwanted artifacts. Filters are actually delay lines. To increase the slope, you increase the size of the delay line. Increasing the delay line causes phase problems (timing issues), which are especially noticeable with transient sound elements.
Filters also often have resonance controls. Resonance comes from the delay line being fed back into the original signal, and subsequently delayed/filtered again. This feedback produces an amplitude increase around the cutoff/center frequency.
An envelope is simply a function that changes over time. Typically we think of amplitude (loudness) envelopes, but envelopes can be applied to pitch, filters, and any other parameter. The most common envelope is called an ADSR envelope, named after its four segments:
- Attack time
- initial Decay time
- Sustain level
- Release Time
They often look like this in generic picture form:
The most typical modulator is an LFO, or Low Frequency Oscillator. LFO’s operate in the range of frequencies below audio hearing (less than 20 Hz), although most go up to 50 – 100 Hz maximum to produce some interesting modulation effects. LFO’s can provide pitch vibrato by modulating the frequency of an oscillator, amplitude tremolo by modulating the amplitude of a sound, or other effects like spectrum vibrato by changing the cutoff frequency of a filter. The elemental shapes can provide a variety of effects. Triangle and sine waves are usually employed to produce smooth fluctuations. Sawtooth waves have a sharp change at the start or end of a cycle, and a smooth continuous signal between. Square waves are like on/off switches, while noise can produce a random fluctuation that has many interesting uses.