Physical modeling in synthesis started as a niche style of sound shaping a handful of years ago but has become quite the buzzword in the music production community (probably second only to the inescapable “powered-by-ai” jargon). It was all the talk at NAMM when we went in 2024, but the belle of the ball on this front was undoubtedly Baby Audio’s newest synth, Atoms, which they showcased.
Between talking with the Baby Audio crew at NAMM and also watching a recent YouTube video they uploaded (linked below), I was incredibly inspired and excited to dive a bit deeper into what the heck physical modeling in VSTs and synthesis is. But before I commit myself to doing a deep and comprehensive review of Atoms, I wanted to clear the air and ask many questions. I’m sure the great production community is wondering about physical modeling.
Learn More About Atoms Here
So we sat down with Silvin Willemsen, the designer and developer of Atoms, to pick his brain with every question I can’t seem to find elsewhere online. To say that Silvin went DEEP into the topic is an understatement, but that’s what you get when you ask an expert about something they’re passionate about and highly knowledgeable about.
Can you explain the basics of physical modeling synthesis and how it differs from sample-based synthesis?
The interesting thing about physical modeling synthesis is that all the sound you hear is generated from physical equations rather than recordings or samples of real-world instruments. These equations describe the relationship between various physical parameters of the “object” you want to model, such as its mass, density, shape and size as well as its motion (acceleration/velocity). In the case of Atoms, the “object” is a string which is broken down into connected masses and springs. The great thing is that you can adjust all of these parameters and can very finely tune the sound that you want, resulting in a much higher degree of control than sample-based synthesis. Nearly all of Atoms’ knobs and sliders directly control parameters in the physical model.
On the other hand, sample-based synthesis is quite static and usually takes a lot of time to get right. Imagine that you want to sample a real-world violin. In order to get the entire range of the sounds it can create, you will need to record every combination of bow force, velocity, position along the string, etc. Furthermore, the recordings will probably use the playing style of only a few artists and use only one (or a few) microphone position(s) for recording. Physical modeling on the other hand allows for control over all of these parameters and can be very finely tuned and automated to your liking.
The name “physical modeling” might imply that the synthesized sounds are necessarily realistic, but this is a common misconception. Although physical modeling is generally very suitable for creating sounds with a natural character, I think that if you are looking for realistic sounds of traditional musical instruments, you would be better off using a sample-based synthesizer. For a digital counterpart of a real-life instrument, you can not get more realistic than a recording of the actual instrument. I do think that when it comes to natural interaction with the virtual instrument, physical modeling outperforms sample-based synthesis as it allows for detailed and precise articulations due to its flexibility. For me personally, however, the true value of physical modeling is that it unlocks the possibility to actually go beyond what is physically possible, while still retaining the organic and natural character that this synthesis technique brings.
To sum up, if you are looking for a higher degree of control and natural/organic sounds that have (potentially) never been heard before, physical modeling is a very interesting alternative to sample-based synthesis. This has exactly been our vision for Atoms, to create organic but otherworldly sounds that go beyond the physically possible.
Can you talk of the significance of the visualizer in Atoms and how it represents the underlying physical simulation?
Atoms is based on a simulation of a mass-spring network, where the masses and springs are connected back to back. Think about it like a physical string, such as a violin or a cello string, subdivided into small nodes (masses), with connections between them (springs). The network is able to move in two dimensions, that is, all masses can move up and down (y-direction) and left to right (x-direction). The visualizer gives a sneak peek into how the masses move when you play Atoms.
There are a two key differences between the underlying simulation and what the visualiser shows. First of all, the network is wrapped around a circle, instead of shown back-to- back. The y-location of a mass is shown by its distance from the center, and the x- location rotates it forward or backward around the circle. Secondly, only 5% of the masses are visualised. As fewer masses are used for higher notes, fewer masses are also visualized for higher pitches. The maximum number of masses used is 100, so for low notes, a maximum number of 5 masses will be visualized.
Different settings for the “Profile” parameter result in different visuals, but how the masses and springs are connected in the underlying simulation does not change. The masses are shown as circles, except for the “Lo-fi” profile, where they are located at the vertices of the shapes. One interesting thing to note is that for the “Alternative” profile, not all masses have the same mass (weight), which is why the circles shown by the visualizer have different sizes. The lines drawn for all but the “Alternative” profile only enhance the visuals and therefore do not represent anything in the simulation.
The chaos parameter introduces nonlinearity to the system. Can you explain its impact on sound and how it goes beyond traditional physics?
Before we can talk about (non)linearity, we need step back into the real world.
A real-life musical instrument can generally be subdivided into two main parts: a resonator and an exciter, for instance, guitar & pick, trumpet & lips, or violin & bow. The field of physical modeling usually assumes resonators to be linear, which means that when exciting a resonator loudly, the only thing that changes is the volume. In the real world, this is nearly never the case. For example, when a piano is played loudly, the timbre of the sound changes, in addition to the apparent increase in volume.
In Atoms, the mass-spring network acts as the resonator, but we add nonlinearity to it through the Chaos parameter. Where the springs in the mass-spring network usually only pull the masses toward each other, Chaos allows the springs in the mass-spring network to “push back” (similar to a car suspension, which pushes back when compressed). This results in nonlinear behavior such as detuning and pitch glides because the network is not fully tense anymore. Although this nonlinear effect might be present in real-world strings to a very small degree, we can crank it up to ‘surreal’ values to go beyond what is physically possible.
Moreover, this push-back increases the effect that the two dimensions (x and y) have on each other. This is was not initially intended, but is instead a direct result of the underlying equations. As the two dimensions that the network are differently mapped to the left and right audio channels, the Chaos parameter can sometimes cause very interesting spatial behaviour, naturally panning the sound from side to side.
How does the physical modeling approach in Atoms allow for organic and natural-sounding excitations and articulations?
We have not yet talked about the exciter we use in Atoms, which is a virtual bow. The interaction between the bow and the network is implemented using a friction model, which is highly nonlinear. I think that the bow excitation is one of the main parts which gives Atoms its organic sound. The parameters that directly influence the bow are Force (the force with which the bow is applied to the network) and Overtones (where the bow is
applied along the network) and allow for very precise articulations and small changes in the values of these parameters can greatly influence the sound.
Furthermore, the nonlinearity of the bow causes many of Atoms’ parameters to influence each other. For instance, turning the Release parameter down introduces physical damping to the system, which in turn influences the way that the bow is able to excite the system. This differs from other synthesis techniques, where each parameter usually only affects one specific part of the sound.
With Atoms being highly innovative, how do you recommend new users start exploring the plugin to appreciate its capabilities fully?
I think the best way to start is to go through the many presets created by several talented sound designers and artists. The Factory Presets and Zardonic’s Radioactive expansion pack already show many aspects of Atoms’ sonic capabilities. Furthermore, the randomize button can help to explore the plugin and reveal some truly surprising sounds that Atoms are capable of generating. It might happen that for some of these random parameter settings, the plugin does not seem to generate any sound. This could be due to high values for Attack or Overtones in combination with low Force. This might be confusing to some new users, but this is a direct result of the physical model.
Finally, and most importantly, take your time exploring Atoms. Treat it like a musical instrument that takes time to learn how to play and interact with. When creating a new patch, I usually start at the “KEYS – Back to Basics” preset, as this sets all parameters back to their default settings. To really get to Atoms’ core, turn all of the knobs to 0 (except the filter, turn this to 100 and perhaps set the release to 25) and start exploring each parameter one by one. Although, as mentioned before, the parameters influence each other, you will get a basic idea of which parameter influences which aspect(s) of the sound. Once you have an idea, you can start using the internal automation and, for instance, start creating per-key parameter sweeps using the Retrigger and Hold LFO modes. Also, once you know what each parameter does, you will be able to intentionally tweak patches that the randomiser spits out.
For someone new to physical modeling synthesis, what key concepts or parameters in Atoms would you suggest they focus on to begin crafting their sounds?
It might help to keep in mind that the underlying model simulates a bowed string-like object. I found that the overtones (where the bow is applied along the network) can really affect the base timbre of the sound, so take your time exploring this. For more musical and harmonic sounds, I would advise keeping the Chaos low (but not 0), the Overtones not too high, and use the Lo-Fi Profile only for low notes, as it detunes for higher notes. For more atmospheric sounds, turn up the Chaos and explore some different Profiles.
Finally, expect the unexpected! Atoms is not your usual synth with loads of parameters for controlling the sound. Instead, lots of control is contained in only a few parameters, and it is up to you to explore them and learn how to wield Atoms’ powers 🙂
Will Vance is a professional music producer who has been involved in the industry for the better part of a decade and has been the managing editor at Magnetic Magazine since mid-2022. In that time period, he has published thousands of articles on music production, industry think pieces and educational articles about the music industry. Over the last decade as a professional music producer, Will Vance has also ran multiple successful and highly respected record labels in the industry, including Where The Heart Is Records as well as having launched a new label with a focus on community through Magnetic Magazine. When not running these labels or producing his own music, Vance is likely writing for other top industry sites like Waves or the Hyperbits Masterclass or working on his upcoming book on mindfulness in music production. On the rare chance he's not thinking about music production, he's probably running a game of Dungeons and Dragons with his friends which he has been the dungeon master for for many years.
