Access Official Technical PDF

Share Summary:LinkedInEmail

A CMOS cellular automata circuit can translate simple local update rules into complex patterns that evolve across an array of connected cells. One experimental approach uses an extended threshold-coupled map (ETCM), implemented with pulse-width modulation (PWM). The same circuit architecture can support continuous nonlinear dynamics as well as discrete cellular-automaton behavior, offering a hardware platform for studying systems whose behavior emerges from local interactions.

The work demonstrates a fabricated CMOS array running one-dimensional cellular automata (1D CAs), including Rule 90 and Rule 110. These results are relevant to engineers investigating nonlinear hardware, parallel computation, and the implementation of dynamical systems in silicon. They are experimental demonstrations of a model and circuit, rather than evidence of a production-ready industrial control solution.

From coupled maps to cellular automata

A coupled map lattice (CML) consists of sites whose states are updated over discrete time steps. In a threshold-coupled map (TCM), each site first applies a nonlinear function to its state. If the result exceeds a threshold, the amount above that threshold—the excess—is passed to neighboring sites, and the originating site is reset to the threshold. This relaxation step redistributes state between cells without advancing the time step.

The extended TCM adds two features: mutual connections between neighboring sites and adjustable connection weights. In a one-dimensional array, a cell receives contributions from its left and right neighbors. The nonlinear map, threshold, and left and right weights together determine the update behavior. Selecting these parameters appropriately allows the network to reproduce the binary update rules of a 1D cellular automaton.

For a binary cellular automaton, each cell is either 0 or 1. Its next state depends on its current state and those of its two neighbors, producing eight possible three-cell configurations. A rule assigns an output to each configuration. The ETCM can encode these outputs by choosing a nonlinear function and connection weights so that each configuration produces the required binary result.

How the PWM-mode CMOS circuit works

The circuit represents state using both voltage and time. Its core includes capacitors, analog buffers, a switched current source, a comparator, and logic. A state is held as a voltage, while PWM encodes that state as a pulse whose width is determined through comparison with a ramp voltage.

During nonlinear transformation, the circuit samples a nonlinear voltage waveform onto a capacitor. Changing the waveform changes the map applied to the state. During relaxation, the circuit compares the output pulse width with a threshold pulse width. If the state is above threshold, the excess pulse duration is sent to neighboring cores. Switched current sources then implement the connection weights by controlling the charge delivered to those cores.

This time-domain approach is significant to the implementation: the nonlinear operation and threshold-based transfer are carried out through waveform sampling, pulse timing, and circuit logic. The number of relaxation operations performed during an iteration can also be set, affecting the resulting spatiotemporal behavior.

Demonstrating Rules 90 and 110

The researchers fabricated a 20-by-20 array containing 400 core circuits using 0.25-micrometer CMOS technology. The array shares ramp and nonlinear waveforms across its cores, while its connections can be configured to support one- or two-dimensional network topologies. Experiments used mutual connections arranged as a one-dimensional network.

Rule 90: a fractal pattern

For Rule 90, the left and right connection weights were set equally. Starting with one active cell, the circuit produced a pattern similar to the Sierpinski gasket, a distinctive fractal pattern also observed in numerical simulation. The experimental output was not binarized and appeared noisy, but its overall structure resembled the simulated result.

Rule 110: asymmetric coupling

Rule 110 requires different left and right weights. The array’s current- and capacitance-variation compensation circuitry was used to set those weights asymmetrically. In the Rule 110 experiment, 20 core circuits were used, and the resulting spatiotemporal pattern was similar to the numerical simulation.

Rule 110 is known as a universal cellular automaton: with suitable initial conditions, its behavior can represent universal computation. The circuit experiment therefore demonstrates hardware capable of realizing this rule’s dynamics. It does not, by itself, establish a general-purpose computing product or a specific industrial application.

Why this architecture matters

A central feature of the design is that continuous and discrete dynamics can be implemented on the same chip architecture. A logistic map, for example, can generate continuous-valued behavior and complex spatiotemporal patterns; digital waveforms can instead configure the array to reproduce cellular-automaton rules. The internal state need not be reset simply to change between these types of dynamics.

For automation and embedded-systems engineers, the broader point is architectural flexibility: local analog state, time-domain processing, and configurable coupling can be combined in an array to explore distributed dynamics. Potential areas of interest include nonlinear computation and systems that use spatiotemporal patterns, although the reported work does not validate a particular plant-control or commercial deployment.

Conclusion

The ETCM circuit shows how PWM-based CMOS hardware can implement local threshold interactions and reproduce cellular-automaton behavior. Its demonstrations of Rules 90 and 110 illustrate both pattern generation and the importance of configurable connection weights. For technical decision-makers, the practical takeaway is a research platform that brings continuous nonlinear maps and discrete cellular automata into one circuit architecture—an approach worth considering when evaluating specialized hardware for dynamical systems, while keeping its demonstrated scope distinct from proven industrial applications.

Access Official Technical PDF

About The Author