Sometimes you want a synthesizer to sound exotic and fresh —
totally unlike anything ever heard before. And sometimes you want it to
imitate the sound of an existing instrument, such as a guitar or
clarinet, with as much realism as possible.
Fig. 1. PianoTeq is a physically modeled grand piano. Controls include hammer hardness, the tuning
width of the multi-string unisons, the impedance (absorption) of the soundboard, and so on.
Electronic instruments can produce realistic sounds in two main ways. We can play a sample (a digital recording) of the original instrument, or we can attempt to synthesize the sound from scratch, without using samples. Both approaches have strengths and weaknesses.
The great strength of sample playback synthesis is that a sample can sound exactly like the instrument being emulated. The weakness of sample playback is that it’s quite tricky to control samples with the kinds of expression and nuance that instrumentalists such as violinists and trumpeters routinely apply. Trying to play an expressive solo line with samples can feel a little like trying to hit a home run when the baseball is encased in a block of cement.
Physical modeling is a technology that strives to be as realistic as sampling while preserving the unique performance characteristics of the instrument being modeled. In physical modeling, the synthesizer uses a mathematical model that digitally duplicates the physical characteristics of a given instrument.
Physical modeling typically requires a lot more computer power than sample playback, because of the number of mathematical computations involved in running the model. But a good software-based physical modeling instrument can be astonishingly realistic and playable. For instance, if you push the pitchbend wheel upward while playing a note with the trumpet model in Arturia Brass, you won’t hear a smooth synth-type pitchbend. Instead, the trumpet tone will “break,” splattering upward through the overtone series in a way that sounds a lot like what happens when a trumpeter adds lip pressure. With a physical-modeled guitar, you’ll be able to adjust the pickup position so that it’s closer to or further from the bridge, again in a way that sounds a lot like the result of moving the pickup on a real guitar (but without the messy woodworking and rewiring).
WHAT’S IN THE MODEL
An acoustic musical instrument contains an energy source of some sort, which produces vibrations. It contains one or more resonators, which amplify and shape the vibrations. And it contains bits of hardware that absorb or reflect the vibrations. In physical modeling, each of these components is modeled in software using filters, delay lines, and other components.
In a guitar, the energy comes from the pick striking the string. The string itself is a resonator, and the bridge and nut absorb and reflect the vibrations of the string. In an acoustic guitar model, the bridge will also transmit vibrations to a separate resonator, the body, whose characteristics are quite different from those of the string.
In a physical model of a guitar, then, you’re likely to see parameters for the hardness of the pick, the distance from the bridge at which the string is plucked, the thickness of the string, and so on. Adjusting any of these will change the tone of the preset.
BEHAVIOR MODIFICATIONS
Unlike a sample playback instrument, which may have hundreds of presets that can emulate anything from timpani to a piccolo, a physical modeling synth will usually include only one or two models. You may be able to do quite a lot of sound programming with these, however. If you want to hear the sound of a guitar whose strings are as thick as bridge cables and are being strummed with a pick made of cloth or glass, you may be able to dial it up. Sounds that would be quite difficult to get using conventional oscillators and filters can be programmed very easily.
When parameters are set to extreme values, however, a physical model can become unstable. It may produce no tone at all, or nasty distorted noises, or it may play out of tune. You might be able to find musical uses for the unstable behavior. Or not.
AVAILABLE MODELS
Physical models of brass, woodwinds, guitar, acoustic and electric piano, and tone-wheel organ are available from several manufacturers. These instruments are not too difficult to model, and there’s a demand for them. Other instruments (such as bagpipes) are so specialized that developing models would probably cost more than a manufacturer could expect to earn. Modeling drums is feasible (Korg did it with the WaveDrum), but there’s less need to develop physical modeled percussion because sampled percussion can sound extremely realistic.
JARGON JOCKEY
Sample: A digital recording of a sound.
Sample playback synthesis: A technology in which one or more samples are triggered from the keyboard (or, with percussion samples, for example, from an alternate controller).