⚡ electrical engineering

Describing Functions and Phase Response Curves of Excitable Systems

This paper proposes a novel framework tailored to excitable systems that utilizes a discrete-event operator to overcome the limitations of classical describing functions and phase response curves, offering a new basis for analyzing and designing central pattern generators in neural networks.

Original authors: Robin Wroblowski, Rodolphe Sepulchre

Published 2026-07-20

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Robin Wroblowski, Rodolphe Sepulchre

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Rhythm of the Nervous System: When Sparks Replace Waves

Imagine the human brain not as a smooth, flowing river, but as a city of billions of tiny, hyper-alert sentinels. These sentinels are neurons, and unlike a drum that beats on its own, many of them are "excitable." This means they sit quietly, waiting for a specific knock at the door before they fire a single, sharp burst of energy—a spike. This is how your nervous system communicates: not by humming a continuous tune, but by sending a rapid-fire sequence of discrete events, like Morse code or a drummer waiting for a cue to hit the snare.

For decades, scientists have used two powerful tools to understand how these biological rhythms work: the Describing Function (DF) and the Phase Response Curve (PRC). Think of the DF as a way to predict how a system reacts to a steady, rhythmic push, and the PRC as a map showing how a tiny nudge at the wrong (or right) time can speed up or slow down a beat. These tools are fantastic for systems that oscillate naturally, like a pendulum or a heart that beats on its own. However, they hit a wall when applied to excitable neurons. Why? Because the math behind these tools assumes the system is already humming a smooth, continuous song. But excitable neurons don't sing; they wait, then pop. Trying to use a tool designed for a smooth wave to analyze a sharp, sudden spike is like trying to measure the height of a lightning bolt with a ruler meant for ocean tides. It just doesn't fit.

The Paper's New Map for Spiking Neurons

In this paper, Robin Wroblowski and Rodolphe Sepulchre propose a clever workaround. Instead of trying to force the old, smooth-wave tools to fit the jagged reality of spiking neurons, they reinvented the tools from the ground up to speak the language of "events." They call these new tools the Event Describing Function (eDF) and the Event Phase Response Curve (ePRC).

The Big Idea: Counting the Beats, Not the Waves
The authors realized that for excitable systems, the most important thing isn't the shape of the wave, but the timing of the spark. They created a simplified model that treats the neuron like a black box: you give it a specific input event (a knock), and it gives you an output event (a spike) after a certain delay.

The Event Describing Function (eDF): Imagine you are tapping a rhythm on a table, and a friend is tapping back. The eDF measures the "lag" between your tap and their tap. If you tap every 25 milliseconds, how long does it take them to reply? The paper shows that by plotting this lag against the time between your taps, you can predict exactly when a whole network of neurons will start dancing in sync. For example, they simulated a ring of inhibitory neurons (the "brakes" of the brain) and found that if you connect two of them, they naturally settle into a rhythm with a period of about 25.31 milliseconds. Their new math predicted this almost perfectly, matching the complex computer simulations that solve the full, messy physics equations.
The Event Phase Response Curve (ePRC): Now, imagine the friend is tapping along, and you suddenly give them a little push or pull. The ePRC maps out what happens if you poke them at different moments in their cycle. Does a poke make them tap sooner? Later? Or does it make them miss a beat entirely? The authors found that for these excitable neurons, the answer depends heavily on when you poke them. A poke right after a spike might do nothing, but a poke during the "recovery" phase could trigger a whole new rhythm.

What They Found (and What They Didn't)
Through detailed simulations of a classic neuron model (the Hodgkin-Huxley model), the authors demonstrated that these new "event-based" tools are incredibly good at predicting how networks of neurons will behave.

They showed that you can design a network of neurons to produce a specific rhythm just by looking at the eDF curves. If the curves cross a certain line, you know a stable rhythm will exist.
They proved that these networks are robust. Even if you change the speed of the input slightly, the neurons tend to lock into a rhythm, either marching in step or in a "half-step" (out of phase), depending on whether the input is excitatory or inhibitory.
They found that the "decay time" of the connection between neurons (how long the signal lingers) is a powerful knob to tune the rhythm.

The Limits of the Discovery
It is important to note that these findings are based on computer simulations, not physical experiments on real brains. The authors are careful to say that while their method works beautifully for the models they tested (like the Hodgkin-Huxley neuron), it might not work for every single type of neuron in existence, especially those with very complex, non-smooth behaviors. They also focused on a specific scenario where every input triggers exactly one output (a 1:1 lock), which is a common but not universal behavior.

Why This Matters
The ultimate goal of this work isn't just to understand biology; it's to build better machines. The authors suggest that by using these new event-based tools, engineers can design "neuromorphic" computers—chips that mimic the brain's efficiency. Instead of trying to simulate the slow, heavy physics of a neuron, these chips could just pass discrete "events" around, using the eDF and ePRC to ensure the whole system dances to the right beat. This could lead to smarter, more efficient robots and medical devices that interact with the nervous system in a way that feels natural, not forced.

In short, Wroblowski and Sepulchre didn't just fix an old tool; they built a new one that finally speaks the native language of the brain's sparks.

Technical Summary: Describing Functions and Phase Response Curves of Excitable Systems

Problem Statement
Classical analysis tools for rhythmic behaviors, specifically the Describing Function (DF) and Phase Response Curve (PRC), are widely used in engineering and neuroscience to characterize feedback oscillations. However, these methods rely on assumptions of harmonic balance, quasi-harmonic waveforms, and intrinsic limit-cycle oscillations. Excitable systems, such as neurons in the excitable regime, violate these assumptions: they do not oscillate autonomously but generate discrete action potentials (spikes) only in response to triggering stimuli. Furthermore, their dynamics are characterized by switch-like transients and fast-slow time-scale separations rather than smooth, continuous oscillations. Consequently, traditional DF and PRC methods lack predictive power and are not directly applicable to excitable networks, creating a gap in the design and analysis of Central Pattern Generators (CPGs) and neuromorphic control systems.

Methodology
The authors propose a novel framework that adapts classical DF and PRC methodologies to event-driven excitable systems by abstracting the continuous-time dynamics into a discrete-event map. This approach shifts the focus from continuous trajectories to the mapping between input event timings and output event timings.

Discrete-Event Modeling: The system is modeled as a sequence of events where a presynaptic voltage spike triggers a synaptic current, which drives the membrane voltage, ultimately triggering a postsynaptic spike. This is formalized as a mapping from event sequences to event sequences, assuming a 1:1 phase-locking mode where each input event triggers exactly one output event.
Event Describing Function (eDF): Analogous to the classical DF, the eDF characterizes the steady-state phase relationship between periodic input and output event trains. It is defined as the relative phase shift $\phi(T) = \Delta(T)/T$ , where $\Delta(T)$ is the steady-state event delay and $T$ is the input period. This metric quantifies how the system's output timing depends on the input frequency, disregarding amplitude/gain dependencies typical of classical DFs.
Event Phase Response Curve (ePRC): Analogous to the classical PRC, the ePRC quantifies the time shift induced in a nominal periodic event sequence by a periodic synaptic perturbation. Unlike classical PRCs that assume weak coupling, the ePRC accounts for strong perturbations and the specific effects of synaptic parameters (e.g., decay time constants). It maps the timing of a perturbation ( $t_p$ ) relative to a nominal input to the resulting shift in the output event timing.

The methodology is illustrated using the Hodgkin-Huxley (HH) neuron model with generic synaptic conductance, demonstrating that the framework is not limited to specific neuron models but relies on the general structure of event-to-current-to-voltage-to-event mappings.

Key Results

Characterization of Excitable Nodes: Simulations of HH neurons reveal that eDF curves for inhibitory and excitatory nodes exhibit distinct behaviors. Inhibitory nodes show significant rebound delays, while excitatory nodes show immediate responses. The curves plateau at a "resting period" ( $T_r$ ) and exhibit nonlinear deviations (due to after-hyperpolarization/depolarization) between the minimum locking period ( $T_{min}$ ) and $T_r$ .
Network Oscillation Prediction: The eDF enables graphical prediction of oscillation existence and period in ring networks. The condition for a stable rhythm in a ring of $N$ nodes is $\sum \phi_i(T) = 1$ . The authors successfully predicted the periods of homogeneous and heterogeneous ring networks (e.g., half-center oscillators formed by two inhibitory nodes) with high accuracy compared to full continuous-time ODE simulations.
Stability and Monotonicity: The paper observes that a monotonically decreasing eDF implies a unique solution for the network period, which correlates with local stability.
Entrainment Analysis: The ePRC successfully predicts entrainment behavior. For inhibitory nodes, the ePRC identifies stable phase-locking equilibria. The analysis shows that inhibitory networks entrain robustly to in-phase excitatory perturbations at higher frequencies and out-of-phase inhibitory perturbations at lower frequencies. The framework accurately predicted phase lags in a four-node ring network under frequency mismatch.

Significance and Claims
The paper claims to provide a basis for designing and analyzing Central Pattern Generators in networks of excitable neurons, with direct relevance to neuromorphic control. By adapting classical tools to the event-driven nature of excitable systems, the proposed framework preserves the simplicity and insightfulness of DF and PRC methods while offering strong predictive power for non-oscillatory, spiking networks.

The authors position this work as a necessary adaptation for systems where traditional harmonic assumptions fail. They highlight that the framework delivers actionable design principles for rhythmic neuromorphic networks built from excitable components. The work draws inspiration from recent findings on winner-takes-all topologies and aims to facilitate the intuitive design of CPGs. The paper concludes by noting that extending this analysis to burst-excitable systems and more complex CPG architectures remains a direction for future research, rather than claiming immediate application to those specific complex systems.

The Rhythm of the Nervous System: When Sparks Replace Waves

The Paper's New Map for Spiking Neurons

Technical Summary: Describing Functions and Phase Response Curves of Excitable Systems

More like this