Exploring the Uncanny: Yamaha VL Series

History of Yamaha's Physical Modeling Synthesizers

Robin Vincent · 07/17/26

Yamaha has been responsible for a couple of paradigm shifts in the synthesis. Perhaps the most obvious came in 1983 with the release of the DX7. It turned the synthesizer industry on its head and introduced a new level of synthetic realism through digital FM Synthesis. Analog synthesizers had only really pretended to emulate the sounds of real instruments and were far more suited to being an instrument in their own right. Synths were unique, spirited, expressive, but definitively electronic-sounding. Whereas FM Synthesis was capable of producing convincing piano, string, percussion and woodwind sounds that rattled real instrument players.

The second came in 1994 with the release of the VL1. It used what Yamaha called “Virtual Acoustic synthesis”, but to anyone else it is known as Physical Modeling. The level of realism possible through physical modeling was off the charts. It could potentially sound like any real-world instrument, as well as instruments that couldn't physically exist. It was a synth for the virtuoso and offered the sort of expressive playing that was only previously available on an acoustic instrument.

In this article, we're going to explore physical modeling and its importance in the history of synthesis. We'll look at some of the synthesizers that championed this technology and how they beat FM Synthesis and Sampling at their own game.

Physical Modeling

The genesis of physical modeling (sometimes called acoustic modeling) harkens back to the early days of electronic sound generation and attempts at speech synthesis. But let's not get bogged down in history and focus on the technological breakthroughs that brought us to the Yamaha VL-1.

The basic idea is that if you can analyze a real musical instrument and generate enough data points on how the physicality of that instrument produces a sound, then you should be able to recreate that sound using a computer model or simulation. Or, to put it another way, if you can work out the maths of how a string vibrates, how a pipe resonates, how air is excited, how a surface responds to a hit and how these things connect to bowing, blowing, strumming and striking then, with a powerful enough computer, you can simulate these processes in real time and produce a realistic approximation of the playing of that instrument.

That's the point, really. It's in the playing of the instrument that physical modeling really finds its genius. When digital sampling became popular with the Fairlight CMI in the late 1970s and early 1980s, it was often painted as a danger to real musicians. If you could sample a saxophone and play it on your keyboard why on earth would you need to hire a saxophonist ever again? The answer to that question was found in the sampler's inability to capture the performance of the instrument. In a sampler, you had velocity and perhaps multiple levels of sampling, but it was impossible to allow for the movement of the mouth, the subtle pressure of the fingers and the variation that a player imparted upon the instrument. All of those aspects can be calculated through physical modeling and then controlled through expressive interactive devices. Physical modeling was, potentially, indistinguishable from the instrument it was modeling.

The technology most responsible for kicking off the research in modern physical modeling was the discovery of the Karplus-Strong algorithm in the 1980s. Invented by Alexander Strong and developed by Kevin Karplus, it uses some simple synthesis modules to simulate the sound of a plucked string. If you have a modular synthesizer then you can do this yourself. All you need is a noise source, a digital delay effect and a filter. You then fire a short burst of noise into the delay set to a very short delay time. The result is a struck, twanging sort of sound that can have its timbre manipulated by the filter and the pitch altered by varying the delay time. With a bit of fiddling, you can get some quite convincing string sounds, which is what the very famous Rings module, from Mutable Instruments, is based upon.

So, how does Karplus-Strong have anything to do with computer models of real instruments? Well, the digital delay is sampling the input and essentially capturing a waveform as digital data. This led directly to the formulation of "digital waveguide synthesis" by Julius O. Smith III at Stanford University, which experimented with a series of digital delay lines to represent the geometry of waveguides such as tubes, bells, plates and membranes. The mathematical understanding of waveguides allowed them to model all sorts of shapes and materials in both resonators and exciters, also known as the body of the instrument and the thing that drives it.

Yamaha licensed the technology from Stamford in 1989 to begin developing their "Virtual Acoustic" technology at a time when DSP hardware was becoming powerful enough to run the models in real time. It was a risk investing in some scientific research with a view to turning it into a workable product, but Yamaha had to do something. Their dominance with the DX range had been annihilated by Korg and Roland with their "Sample & Synthesis" style machines, like the Korg M1 and Roland D-50. They offered much better and broader realism than FM-based synths. However, Yamaha could see the limitations in sampling and had a hundred years of instrument-making behind them to draw upon to shape this new form of synthesis.

The Yamaha VL1

Yamaha's first synthesizer to feature virtual acoustic tone generation (also known as Self-oscillating Virtual Acoustic or S/VA) was the 1994 VL1. It was expensive and an odd-looking machine. Like some sort of walnut-panelled vintage synthesizer surrounded by ostentatious gold-coloured futuristic plastic. It was distinct in form and most certainly unique in function. The VL1 was a largely monophonic (although occasionally duophonic) synthesizer designed for solo performance and focused on blown or bowed instruments. The very idea that someone would release a mostly monophonic synth in the golden age of polyphony seemed ludicrous, but then this was no ordinary synth.

The key to the VL1 was how it was controlled. Previously, expressive control came in the form of pitch bend, velocity and playing with the mod wheel. The VL1 came with three wheels, the BC2 Breath Controller and EC7 Foot Controller. Yamaha actively encouraged you to engage breath, feet and hands in the playing of the included instruments. The results were breathtaking. The nuance and level of control, deviation, authenticity and musicianship were staggering; there was nothing like it.

The Yamaha VL1 physical modeling keyboard synthesizer

The instrument sounds were factory-defined, and Yamaha kept the access to the complexity of the instrument models tightly locked up. However, there was a comprehensive set of parameters available to fiddle with. You could affect the pressure, the bow speed, tightening of the lips (embouchure), vibrato, tonguing, breath noise, overblown distortion, filtering, dampening, throat formants and absorption; terms you'd never have heard directed at a synthesizer before. Within the VL1, the instruments were split into two models, the "exciter", such as bowing and blowing and then the "resonator", such as a tube, pipe, string or membrane. These naturally go together to match their real-world counterparts, but what would happen if you made some unnatural couplings? It was completely possible in the VL1 to blow a string, strike a pipe and pluck a membrane, giving it some wonderfully different sounds that were still just as playable as the more conventional instruments.

The Yamaha VL7

By the end of the same year Yamaha had introduced the VL7. It was almost half the price of the VL1 but took on a nearly identical look and intention. It was purely monophonic, had half the sounds and could only run a single instrument at a time when the VL1 could manage two. It was pitched as a solo lead instrument where the performance and control were everything. When you try to emulate how a saxophone is played with breath and modification to each note, then translating that into polyphony is nigh impossible. And so, it was resolutely monophonic and proud of it.

The Fabled Yamaha VP1

While going to great lengths to extol the virtues of the monophonic VL1 and VL7, Yamaha also produced a polyphonic version called the VP1. It focused much more on the string elements rather than the woodwind and brass, which made sense in a polyphonic environment. It used a slightly different implementation of the virtual acoustic technology called Free-oscillating Virtual Acoustic (F/VA) and combined a "driver" model with a "string" component to build the sounds.

The two implementations are quite different. The Self-oscillating S/VA found in the VL1 has energy continuously injected into the model by the bowing or blowing, and then the effect on the sound waves is continuously recalculated. With F/VA the energy injection came from short, sharp events such as plucking or striking and all the calculations were then done in the resonant part of the model. The idea was that the VL1 (and VL7) would cover the solo instruments while the VP1 would fill in the rest of the orchestra around it.

However, it wasn't all going to plan. The physical modeling in the VP1 never quite matched up to that of the VL1. It sounded weird and synthetic, and with an eye-watering price tag of around $30,000, it was out of reach of the sort of customers who may have found the weirdness interesting. The VL1 and VL7 were also not selling as well as Yamaha had hoped. While the realism of the sounds was applauded, the musical skills required to achieve a believable performance were so high that only virtuosos need apply. You had to be a great wind player to get a great wind performance which seemed a bit self-defeating. Consequently, Yamaha pulled the plug on the project and never actually released the VP1 past a handful of prototypes.

The Yamaha VL70m

Probably the most successful version of the Yamaha virtual acoustic technology appeared as the 1U half-rack VL70m in 1996. At less than an eighth of the price of the VL1, this astonishing sound generation technology was within reach of regular musicians. It was awful to edit, but it had a bank of monophonic sounds that were superb. Yamaha had the foresight to place the breath controller port on the front along with a socket for the WX series of wind controllers. This gave the VL70m an entirely new lease of life as a believable sound source for digital wind players.

Even if you didn't blow into it, the VL70m offered a source of sounds unlike anything else around at the time and with a bit of MIDI editing, you could get it to sound fabulous without having to pass a Grade 8 music exam.

Contemporary Competitors

However, Yamaha wasn't the only company working in this area at the time—other large manufacturers tried their own hands and developing instruments that incorporated physical modeling concepts, both in hardware and software contexts. Here are some highlights from that era.

Korg Prophecy

Around the same time Yamaha was developing their virtual acoustic technology, Korg was working on the OASYS project which was designed to model every form of synthesis into algorithms to run on DSP hardware. One hardware protrusion of that technology was the weird and wonderful Prophecy. It was fun, expressive, monophonic and along with virtual analog and Variable Phase Modulation tones it had some physical modeling. It wasn't designed with the elegance or pretensions of the VL1, and so was extremely accessible and very successful.

Technics SX-WSA1

The Technics SX-WSA1 was an unusual but great-sounding polyphonic physical modeling synthesizer that suffered from being created by a company that didn't really understand the synthesizer market. It cleverly combined PCM sample-based waveform drivers with physically modeled resonators, making it much simpler than modeling the entire acoustic instrument. It was very versatile and offered a far greater depth of editing than the Yamaha VL series.

The internet tells me that very few were made due to lack of sales; however, when I started working at the Turnkey Music Store in London in 1996, they had just bought a whole container load and were selling them off at a fraction of the original price. At the time, there were plenty to go around.

Korg Z1

The Korg Z1 is a bit like putting a whole bunch of Prophecy's into one machine. Subsequently, the Z1 (released in 1997) is a massively multi-timbral and polyphonic hybrid physical modeling synthesizer. Various physical modeling options appear like pluck, bow, brass and reed within the Multi-Oscillator Synthesis System (MOSS), along with VPM and virtual analog. It marks a shift away from virtuoso playing of authentic acoustic models into using physical modeling as just another sound source to create something more than intended. Breath control is replaced by the XY Pad, modulation wheels and aftertouch, making the performance elements far more accessible for non-wind players.

Applied Acoustics Tassman

Tassman was the first piece of musical software I came across that focused on physical modeling. Released in 2000, it was designed around a modular synthesizer approach with various synthesizer modules you could chain together to create your patch. Along with regular synth modules, you had physically modeled reeds, membranes, breath, bowing, tubes, beams and all sorts of other appropriate factors. I remember it sounded very different from anything else in the emerging software synthesizer scene.

Companies such as Applied Acoustics have gone on to produce perfectly modeled electric pianos, analog synths, string synthesizers and a whole modular system; and of course, others like Madrona Labs continue to push boundaries exploring some of physical modeling's more unusual possibilities.

Current Status

After spending a long time being ignored, physical modeling in terms of modeling acoustic instruments has been making somewhat of a comeback. The truth is that the technology has always been present and shifted from modeling real instruments into modeling circuits, amps, tubes and the sound of vibrating electrons that we find in all the virtual analog synths that we have all over the place.

Sampling, which has massively evolved in the same time frame, tends to win out for its simplicity of use when dealing with orchestras, massive soundscapes and solo instruments. The complexities and detail available in physical modeling suits virtual analog, where you have masses of available parameters but no requirement to move them all manually as part of a performance. Even so, the modeling of real instruments still has its merits.

In modular synthesis, the draw of the Karplus-Strong is in the simple elegance of that struck string that penetrates a mix of analog sounds. In software, physical modeling finds itself on the roster of available oscillators in big software synths from Arturia's Pigments to Native Instruments Reaktor. Dedicated hardware synths are still a rarity, but we've had recent foray with the Erica Synths Steampipe, which offers a lot of interesting possibility. Otherwise, big workstation synths tend to keep exciters and resonators knocking around as part of their sonic arsenal.

On the whole, physical modeling has been largely absorbed into the everyday fabric of synthesizer sound sources. I think it's unlikely that we'll see anything quite like the Yamaha VL series again. It was dedicated to its core technology but required far too much of the player to get the best out of it. These days, there seems to be less of a desire to believably fake the sound of a real instrument. We care less about where a sound comes from, how it was played and whether the technology is up to the task. Instead, we put our energies into whatever part of the musical process interests and thrills us.

Tutorials, news and Reviews from Perfect Circuit