Simulating FM with a Wavetable or SamplerWritten 2001/01/05
Just in case some of you don't know, FM synthesis involves using oscillators modulating the frequency of another oscillator(refer to glossary in issue 4 for more). While many wavetable synths and samplers don't have this sort of frequency modulation as part of their architecture, you can still simulate the FM sound using the proper samples and subtractive synthesis.
First you should know that FM synthesis produces a series of harmonics and subharmonics known as sidebands. The amplitude of these sidebands is changed when the modulation index (the amplitude of the modulator) is affected by envelopes or LFO's. Once you know how FM works, you can then proceed to emulate it.
For our raw materials we need two parts: 1. a sample of a static FM waveform rich in sidebands, and 2. a way to apply filter envelopes to the sample. Many wavetable synths (digital boards with samples stored in ROM) have a few such FM samples built in. And of course, if you're lucky enough to own a sampler, you can easily make your own waveform, whether you sample from an FM synth a friend lends you, or by scouring the web's resources for the right sound. Once you've got a sample with that metallic quality characteristic of FM, you can start applying filter envelopes to it. The filtering out of the harmonics simulates the changing of the modulation index on a real FM unit. You can also use a little highpass filter to get rid of some of the subharmonics as well, just be careful not to overdo it, and experiment with keyboard scaling of the filter to keep things sounding right all the way across the keys. As you start changing parameters, some very convincing sounds should start to emerge. For fun you can add additional modulators to the filters, like velocity, alongside the envelope.
Why would you go through all this trouble? This technique lets you approximate a kind of synthesis that might otherwise be impossible for you if you don't actually own a machine with true FM architecture built in. It's also more memory efficient than sampling a long modulation index sweep.
Now if you add a little delay you can pretend you're Ray Lynch - if you're so inclined. Soda pop anyone?
Return to CREON's Nightwaves Page
Contact me via email at 
