The most power-hungry part of a wireless base station is not the antenna. It is the electronics behind it – the chips, amplifiers, and signal processors that handle every calculation the antenna cannot do on its own. A new design from engineers at the University of British Columbia proposes offloading some of that work to the antenna itself, using materials that process radio signals as they pass through rather than after they arrive.
Modern wireless systems work in a specific sequence. A signal arrives at an antenna. The antenna converts it to an electrical signal. Electronics process that signal — filtering it, focusing it, cleaning it up, making sense of it. Then the processed result goes somewhere useful.
Every step in that chain consumes power. Every step adds hardware. And as wireless systems move toward 6G, the volume of signals that need processing, and the speed at which they need to be handled, is pushing that hardware chain toward a wall.
The UBC research, published in IEEE Wireless Communications, proposes a different architecture. Instead of a single flat surface that just deflects signals, the system uses multiple stacked layers of specially engineered material.
Each layer modifies the signal as it passes through — adding a small change, nudging it in a particular direction. The layers work together the way neurons in a simple neural network work together: each one contributes a small computation, and the collective result is something meaningful.
What Makes This Different From Existing Antenna Technology
Reconfigurable intelligent surfaces (RIS) already exist. They are flat panels of metamaterial (engineered material that interacts with radio waves in programmable ways) that can steer or redirect signals without active electronics. They are considered one of the key enabling technologies for 6G.
The limitation is that conventional RIS is linear. A linear system can only do certain things: steer a beam, amplify, combine. It cannot distinguish between signals in the way a neural network distinguishes between patterns. It cannot make the kind of nonlinear decision that separates one user’s data from another’s in a crowded spectrum.
The UBC team’s contribution is to make the surface nonlinear. Each layer in the stack is designed with a unit cell* that has a step-like response curve rather than a smooth one. That nonlinearity is what gives the system its computational character. The stacked layers, taken together, can perform operations that a flat, linear surface cannot.
* having dimensions order of a wavelength – typically mm or cm
What This Actually Changes
The practical target is base station hardware. A 6G base station serving thousands of simultaneous users needs to constantly calculate where to direct signals, how to separate overlapping transmissions, and how to maintain reliability as users move. Those calculations currently happen in digital signal processors that consume significant power and generate heat.
If a portion of that signal separation and processing can happen physically, in the wave domain, before the signal ever reaches a chip, the electronics behind the antenna get simpler. The base station gets smaller, cooler, and cheaper to run. In a world where the mobile industry is under intense pressure to reduce the energy cost of networks — 5G base stations already account for a significant fraction of mobile operators’ electricity bills — that matters concretely.
There is also a security angle. Nonlinear wave-domain processing can scramble signals in ways that are physically very difficult to intercept, adding a layer of inherent security that does not depend on encryption software running on a server somewhere.
The Broader Context
The direction this research represents, moving computation into the physical medium rather than concentrating it in silicon, is part of a wider shift happening across engineering. Photonic computing, neuromorphic chips, and analogue AI accelerators all share the same core insight: digital processors are good at many things but they are not the only way to compute.
For wireless systems specifically, wave-domain computing offers one capability that silicon cannot match: it operates at the speed of light, with no analog-to-digital conversion step, and it adds no latency. For the ultra-low-latency applications that 6G is designed to enable, including real-time industrial control and augmented reality, that may matter as much as raw throughput.
What to Watch For
The UBC team is clear that hardware testing is ongoing. The nonlinear unit cell design they propose is a proof of concept, not a product. The gap between a working unit cell in a lab and a manufacturable, weatherproof antenna panel that performs consistently in a real network environment is significant.
The near-term test is whether the nonlinear response can be maintained stably across the environmental conditions a real base station faces: temperature changes, multipath interference, and the constant variation of a live network. If the hardware results match the theoretical performance, this architecture becomes a serious candidate for inclusion in 6G base station designs scheduled for standardisation later this decade.
Sources
Abbas, O., Zayat, A., Markley, L. and Chaaban, A. “Nonlinear Stacked Intelligent Surfaces for Wireless Systems.” IEEE Wireless Communications (2026). DOI: 10.1109/mwc.2026.3666521
Preprint (open access, full text) arXiv:2510.23780 (submitted October 27, 2025) URL: https://arxiv.org/abs/2510.23780 PDF: https://arxiv.org/pdf/2510.23780
Institutional press release “UBCO breakthrough could transform future wireless networks.” UBC Okanagan News, April 30, 2026. URL: https://news.ok.ubc.ca/2026/04/30/ubco-breakthrough-could-transform-future-wireless-networks/
More Like This
New System Uses Smart Surfaces and Room Reflections to Boost Future 6G Signals
ETH Zurich just showed that 6G networks may not need as much fibre cable as we thought
This AI Taught Itself to Solve a 6G Problem That Stumped Mathematicians
Sam Brown holds a BSc in Electronic Systems Engineering from University of Regina, a Masters from the University of Alberta and a PhD from the University of Toronto in Electrical Engineering. His reporting interests include Physics and Artificial Intelligence.