nlm: Real-Time Non-linear Modal Synthesis in Max
2026
Read nlm: Real-Time Non-linear Modal Synthesis in Max (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-10240.txt.
Introduction
Diaz, Constanzo and Sandler present nlm, a set of Max externals for real time nonlinear modal synthesis. The tools model strings, membranes and plates, with live control of physical parameters and support for custom modal data. Multichannel output is built in, and the code is released as open source for all to use. By placing physical modelling inside a familiar music tool, nlm lowers the bar for composers, performers and sound designers.
Catalog record
| YEAR | 2026 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 16025 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.10240
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
nlm: Real-Time Non-linear Modal Synthesis in Max Abstract We present nlm , a set of Max externals that enable efficient real-time non-linear modal synthesis for strings, membranes, and plates. The externals, implemented in C++, offer interactive control of physical parameters, allow the loading of custom modal data, and provide multichannel output. By integrating interactive physical-modelling capabilities into a familiar environment, nlm lowers the barrier for composers, performers, and sound designers to explore the expressive potential of non-linear modal synthesis. The externals are available as open-source software at https://github.com/rodrigodzf/nlm . 1 Introduction Physical modelling synthesis is a technique for creating sound by simulating the physical properties of real-world objects. It is a powerful tool for creating realistic and complex sounds, and is used in a wide range of applications, from music composition to sound design. Several numerical approaches have been used to implement physical modelling synthesis, including Finite Difference Time Domain (FDTD) methods [ 5 ] , Digital Waveguide Meshes (DWM), and a broad family of techniques often grouped under the term modal methods . These include classical spectral and pseudospectral approaches [ 4 ] , the Functional Transformation Method (FTM) [ 11 ] and to an extent the Finite Element Method (FEM). All of these techniques share the property of decomposing the system into modes that can evolve independently in the linear case and interact in the presence of non-linearities. Due to their compact representation, physical interpretability, and efficient time integration, modal methods are especially attractive in environments like Max and Pure Data (Pd), where real-time interaction is a core requirement. In this work, we present nlm , a unified and optimised set of Max externals implementing non-linear modal synthesis models previously described in the literature [ 11 , 1 , 7 ] . While these non-linear modal models have been extensively studied and implemented in offline simulations, their deployment within Max has remained limited. Our contribution is the creation of a unified and computationally efficient real-time implementation, using the Eigen [ 9 ] C++ library for matrix operations. The externals are designed to allow interactive exploration of the parameters of the system, with support for user-defined eigenmodes, and creative exploration of excitation profiles. 2 Background 2.1 Modal Synthesis in Max and Pd Modal methods have long been central to the development of real-time physical modelling tools within Max and Pd. Their decomposition into resonant modes naturally aligns with modular software architectures, making them particularly well-suited for real-time interactive synthesis. In this section, we provide a short overview of influential modal synthesis tools developed specifically for these environments, focusing on those with notable impact on the field. Early contributions to Max include the NVM externals [ 8 ] , a modular physical modelling suite demonstrating real-time performance with hundreds of modes. Later, IRCAM introduced the mlys and modalys ~ objects, integrating the Modalys synthesis framework into Max and enabling users to define complex resonator networks interactively [ 6 ] . Concurrently, Trueman and DuBois developed the PeRColate collection [ 12 ] , wrapping waveguide and modal components from the Synthesis Toolkit [ 10 ] into Max objects, facilitating synthesis of percussive and other instrument types. More recently, the Sound Design Toolkit (SDT) [ 2 ] provided a cross-platform C core compiled into externals for both Max and Pd. The sdt objects support modal resonators which can be excited by impact, friction, and other physical interactions, allowing efficient simulation of complex non-linear collisions. Similarly, the Synth-A-Modeler compiler [ 3 ] translates FAUST diagrams into optimised externals for Max, offering a high-level prototyping approach for modal synthesis. The implemented resonators in these cases, however, have been limited to linear modal synthesis. While non-linear plate models, such as VKGong 1 1 1 https://vkgong.ensta-paris.fr and VKPlate 2 2 2 https://github.com/Nemus-Project/VKPlate , have been implemented in C++ and MATLAB, real-time implementations of these non-linear models have not yet been fully integrated into Max or Pd. 2.2 Non-linear Modal Models Assuming isotropic, homogeneous material properties, the dynamics of non-linear strings, membranes, and plates with bending stiffness can be described by the general scalar partial differential equation: \rho\ddot{w}+\left(d_{1}+d_{3}\Delta\right)\dot{w}+(D\Delta\Delta-T_{0}\Delta)w=f_{\text{ext}}-f_{\text{nl}}, (1) where w(\mathbf{x},t) is transverse displacement, \rho is the material density (i.e., \rho=\rho_{\text{m}}A in the string case and \rho=\rho_{\text{m}}h in the other cases), d_{1},d_{3} are the damping coefficients, D is the bending stiffness, and T_{0} is the initial tension. The \Delta and \Delta\Delta operators are the Laplacian and biharmonic operators respectively, and the terms f_{\text{ext}} and f_{\text{nl}} represent external and non-linear forces. The parameters with their units are summarised in Table 1 . Equation 1 and its modal formulation have previously been presented and discussed in the literature for strings, membranes [ 11 , 1 ] and plates [ 7 ] . Here we provide a streamlined, unified presentation to ensure the paper remains self-contained. Non-linearities in these models arise from deformation induced tension, in the Kirchhoff-Carrier model for strings and the Berger approximation for membranes. For plates, described by the von Kármán model, non-linearities emerge through coupling between transverse and in-plane modes. The displacement can be decomposed as a modal expansion: w(\mathbf{x},t)=\sum_{\mu=1}^{M}\frac{\Phi_{\mu}(\mathbf{x})q_{\mu}(t)}{\left\|\Phi_{\mu}(\mathbf{x})\right\|^{2}}, (2) where each \Phi_{\mu},\lambda_{\mu} eigenpair satisfies the eigenvalue problem \Delta\Phi_{\mu}=-\lambda_{\mu}\Phi_{\mu} . Analytical solutions are limited to simple geometries and boundary conditions, and more complex cases require numerical approaches. In the von Kármán case, an analogous expansion of the Airy stress function yields in-plane eigenpairs (\Psi_{\nu},\zeta_{\nu}) . Truncating the modal expansion to M modes, the system is described by the modal coordinates \mathbf{q}=[q_{\mu}]_{\mu=1}^{M}\in\mathbb{R}^{M} . Substituting into ( 1 ) yields: \ddot{q}_{\mu}+2\gamma_{\mu}\dot{q}_{\mu}+\omega_{\mu}^{2}q_{\mu}=\bar{f}_{\text{ext},\mu}-\bar{f}_{\text{nl},\mu}, (3) with \displaystyle\omega_{\mu}^{2} \displaystyle=\frac{D\lambda_{\mu}^{2}+T_{0}\lambda_{\mu}}{\rho},\quad\gamma_{\mu}=\frac{d_{1}+d_{3}\lambda_{\mu}}{2\rho}. (4) Applying the impulse-invariant discretisation to ( 3 ) yields the explicit update scheme: \mathbf{q}^{n+1}=\mathbf{a}_{1}\odot\mathbf{q}^{n}+\mathbf{a}_{2}\odot\mathbf{q}^{n-1}+\mathbf{b}_{1}(\mathbf{\bar{f}}_{ext}^{n}-\mathbf{\bar{f}}_{nl}^{n}), (5) where \odot denotes element-wise multiplication of vectors, and the coefficient vectors are defined as: \displaystyle\mathbf{a}_{1} \displaystyle=-2e^{-\bm{\gamma}T}\odot\cos(\tilde{\bm{\omega}}T), (6) \displaystyle\mathbf{a}_{2} \displaystyle=e^{-2\bm{\gamma}T}, (7) \displaystyle\mathbf{b}_{1} \displaystyle=\frac{1}{\rho\tilde{\bm{\omega}}}\odot\sin(\tilde{\bm{\omega}}T)\odot e^{-\bm{\gamma}T}, (8) with \tilde{\bm{\omega}}=\sqrt{\bm{\omega}^{2}-\bm{\gamma}^{2}} and sampling period T . A similar second-order update scheme can be derived using centered finite differences (Störmer-Verlet) [ 7 ] . In modal coordinates, the non-linear terms for string, membrane, and plate models become: \displaystyle\bar{f}_{\mu,\text{string-nl}} \displaystyle=\lambda_{\mu}q_{\mu}\frac{EA}{2L\rho}\sum_{\mu}\frac{\lambda_{\mu}q_{\mu}^{2}}{\left\|\Phi_{\mu}\right\|^{2}}, (9) \displaystyle\bar{f}_{\mu,\text{membrane-nl}} \displaystyle=\lambda_{\mu}q_{\mu}\frac{Eh}{2L_{x}L_{y}(1-\nu^{2})\rho}\sum_{\mu}\frac{\lambda_{\mu}q_{\mu}^{2}}{\left\|\Phi_{\mu}\right\|^{2}}, (10) \displaystyle\bar{f}_{\mu,\text{plate-nl}} \displaystyle=\frac{E}{2\rho_{\text{m}}}\sum_{p,q,r}^{n}\frac{H_{q,r}^{n}C_{p,n}^{s}}{\zeta_{n}^{4}}q_{p}q_{q}q_{r}, (11) where H_{q,r}^{n} and C_{p,n}^{s} are third-order tensors containing coupling coefficients, where p,q,r index the transverse and n,s index the in-plane modes, respectively. More details on the derivation of Equations 9 , 10 and 11 can be found in [ 7 , 1 ] . 3 Max Implementation The package consists of a C++ core library and four externals: nlm.string ~ , nlm.plate ~ , mcs.nlm.string ~ , mcs.nlm.plate ~ , where the last two are multi-channel versions. Given the similarity between the membrane and plate models, both are encapsulated within a single external nlm.plate ~ . The type of model can be selected via an attribute, specified as berger for the membrane or vk for the von Kármán plate model. The string and membrane implementations follow previous work [ 11 , 1 ] , while the calculation of coupling coefficients for the plate follows [ 7 ] and the MATLAB implementation provided by the VKGong project. Both externals expose the physical parameters of the governing equation ( 1 ) as interactive attributes, summarised in Table 1 . The externals currently support only rectangular geometries with simply supported boundary conditions (i.e., zero displacement and bending at the