The development of wireless communication technology and the Internet of Things requires radiofrequency communication systems with higher frequencies and faster communication speeds. However, traditional digital processing platforms—which involve high-speed analogue-to-digital converters, intensive data movement and complex digital computation in software-defined radio systems—suffer from high energy consumption and latency. Signal processing in the analogue domain using non-volatile memristive devices can reduce data movement and energy consumption, but the development of system-level designs remains limited. Here we report a radiofrequency signal processing system that is based on analogue in-memory computing within a multicore memristive system-on-a-chip. With the approach, we demonstrate an analogue discrete Fourier transform for spectrum analysis, a mixer-free demodulator for in-phase and quadrature demodulation, and analogue neural networks for radiofrequency transmitter identification and anomaly detection. The memristive system-on-a-chip offers an identification accuracy of over 90% and is up to 6.8 times more energy efficient and up to 6.2 times faster than traditional digital processing platforms.
Read more:
https://www.nature.com/articles/s41928-025-01409-y
