Reconfigurable Intelligent Surfaces (RIS) for 6G — The Complete Engineering Guide

For a century we have treated the radio channel as a fixed, hostile thing we design around. Reconfigurable Intelligent Surfaces (RIS) flip that assumption: they make a slice of the propagation environment itself programmable. A thin panel of hundreds or thousands of tiny tunable elements, bolted to a wall, can take an incoming radio wave and re-aim it — filling a coverage hole, dodging a blockage, or sharpening a beam — using almost no power and no transmit chain. It is one of the defining "smart radio environment" ideas heading into 6G. This guide goes well past the metaphor into the physics, the math, the 3GPP/ETSI/ITU reality, and the problems that still stand between RIS and a real deployment.

1 · From a hostile channel to a smart radio environment

Every link budget you have ever built starts from a number you do not control: path loss. The transmitter beams power out, the environment scatters and attenuates it, and the receiver makes do with whatever survives. Beamforming, MIMO and massive MIMO all improved the two endpoints of the link — but the medium in between stayed a given.

The "smart radio environment" thesis is that the medium need not be a given. If we coat selected surfaces — façades, indoor walls, lamp posts — with electronically controllable reflectors, the network gains a third actor it can configure: the channel itself. A RIS is the canonical realisation of that idea. It does not generate a new signal; it edits the one already in the air.

2 · What a RIS actually is

Physically, a RIS is a planar array of unit cells, each much smaller than a wavelength (sub-wavelength spacing, often around λ/2 to λ/5). Each cell is a small patch antenna or resonator backed by a tunable load — a PIN diode, a varactor, or in some designs liquid crystal or MEMS. Changing the bias on that load changes the cell's reflection coefficient, specifically its phase. Do that across the whole grid and you have programmed a phase profile onto the surface.

Crucially, most RIS designs are passive or nearly passive: there is no power amplifier, no mixer, no ADC/DAC, no conventional RF chain on the signal path. The only power the surface draws is the trickle needed to bias the diodes and run a small controller. That is what separates a RIS from every relay and repeater that came before it — and it is also the root of its hardest problem (see §4).

3 · The physics — phase gradients and the generalized Snell's law

A flat metal mirror obeys the ordinary law of reflection: angle out equals angle in. A RIS breaks that rule on purpose. By imposing a linear phase gradient across the surface — cell n gets a little more phase delay than cell n−1 — the surface re-radiates a wavefront whose constructive-interference direction is tilted away from the mirror angle. This is the generalized (anomalous) law of reflection: the reflected angle now depends on the phase gradient you program.

Because each element's phase is set digitally — and cheap hardware only offers a few states — practical surfaces use quantized phases: a 1-bit cell picks between 0 and π; a 2-bit cell between 0, π/2, π, 3π/2. Coarser quantisation is cheaper and lower-power but loses a few dB of beamforming gain and raises side lobes. The art of RIS control is choosing the discrete phase per element so the reflected beam peaks exactly at the served user.

Phase quantization — the price of cheap hardware

Real cells cannot take any phase; they pick from a few discrete states. Coarser quantisation is cheaper and lower-power but costs beamforming gain and lifts side lobes. The classic figures:

4 · The cascade channel and the multiplicative path-loss problem

Here is the equation that governs every honest RIS link budget. With a direct path h_d (often blocked), a transmitter→RIS channel g, a RIS→receiver channel h, and the RIS response as a diagonal matrix Θ = diag(β₁e^jφ₁, …, β_Ne^jφ_N), the received signal is:

The catch is in that cascade. A passive surface only re-radiates the power that lands on it. So the end-to-end path loss is the product of the two hop losses, not the sum. In the far field it scales roughly as:

That squared product is the famous multiplicative (or "double") path loss, and it is why early lab demos that looked spectacular at 1–2 m disappoint at tens of metres. The two ways out are visible in the formula: make N (the element count) very large — hundreds to thousands of elements give an N² array gain that claws back the loss — or place the RIS close to one end of the link (small d₁ or small d₂), which is why RIS panels are best mounted near the gNB or near the served hotspot, not floating in the middle.

5 · Near field and beamfocusing

Make a surface big enough to beat path loss and you create a second subtlety. The boundary between "far field" (plane waves, steer by angle) and "near field" (spherical waves, focus on a point) is the Rayleigh distance, 2D²/λ, where D is the aperture size. A multi-metre RIS at mmWave or FR3 can push that boundary tens or hundreds of metres out — so many users sit in the surface's radiative near field.

That is not just a nuisance; it is an opportunity. In the near field a RIS can do beamfocusing — concentrating energy at a specific 3-D location (range and angle), not merely a direction. That sharpens spatial multiplexing and, as we will see, underpins RIS-aided positioning and sensing. It also complicates channel modelling, which is exactly why the 3GPP Release-19 channel-model work (below) matters.

6 · The RIS family — it is not just one device

"RIS" is now an umbrella. The variants differ in what physical quantity each element controls and on which side of the panel energy goes.

7 · RIS vs relay vs Network-Controlled Repeater (NCR)

This is the comparison that decides real deployments, because the network-controlled repeater is the option 3GPP has already standardised — and it is RIS's nearest competitor.

The Network-Controlled Repeater deserves attention because it is the pragmatic stepping-stone. 3GPP studied it in TR 38.867 and specified it in Release 18. An NCR has two parts: the NCR-MT (mobile-termination), which maintains a control link (the "C-link") to the gNB, and the NCR-Fwd (forwarding) function, which does the amplify-and-forward. Over the C-link the gNB sends side control information — beamforming weights, on/off timing, and TDD uplink/downlink switching — so the repeater is no longer the dumb, always-on booster of the 4G era but a beam-aware node the network actually orchestrates. RF requirements live alongside the NR repeater specifications (TS 38.106 family). In effect, NCR delivers much of the "smart coverage" promise today, with mature hardware, while RIS matures for 6G.

8 · Where standardization actually stands

This is where a lot of marketing gets ahead of reality, so let us be precise. As of mid-2026:

• 3GPP Release 18 — standardised the Network-Controlled Repeater (study in TR 38.867). This is the deployable, network-orchestrated coverage tool available now. RIS itself is not in Rel-18.
• 3GPP Release 19 (5G-Advanced) — RIS was not adopted as a work or study item. However, Rel-19 advanced the channel-modelling foundations RIS depends on: a study on channel-model enhancements for the 7–24 GHz (FR3) range concluded in June 2025, and an ISAC channel model was folded into TR 38.901 v19.0.0. Both share RIS's core ingredients — reflection, multipath, near-field — so this is groundwork RIS work will reuse.
• 3GPP Release 20 — the 6G study phase (roughly 2025–2027), explicitly non-normative. This is the first realistic window for a 3GPP RIS study item, and widely seen as RIS's main chance to enter the 6G trajectory.
• 3GPP Release 21 — expected to begin the first normative 6G specifications, drawing on Rel-20 studies. Any standardised RIS feature would land here at the earliest.
• ETSI ISG RIS — the dedicated industry specification group already publishing RIS group reports/specs (use cases, requirements, architecture, channel models). It is the venue doing the most concrete RIS standardisation today, feeding 3GPP and ITU.
• ITU-R IMT-2030 — the M.2160 framework (approved Nov 2023) defines the 6G usage scenarios; RIS is repeatedly cited among the candidate enabling technologies. The RIS Alliance has also published a white paper on potential standardisation directions.

9 · Control architecture — who tells the surface what to do

A RIS only earns the "intelligent" in its name through control. Each panel has a small RIS controller (an MCU/FPGA) that drives the bias lines that set element phases. Above it sits a control path to the network. Three questions dominate the architecture debate:

• Who computes the configuration? A centralised gNB/RIC computes the phase profile from channel knowledge and pushes it down, or the RIS does lightweight local optimisation from a codebook. Codebook-based beam selection (sweep a set of pre-defined phase patterns, pick the best) is the pragmatic favourite — it mirrors NR beam management and avoids estimating the full cascade channel.
• What is the control link? Options range from an out-of-band wired/IP link to the RIS, to an in-band signalling channel analogous to the NCR's C-link. Latency and overhead here cap how fast the surface can re-aim — fine for slow blockage changes, hard for fast mobility.
• How does it map to the gNB-CU/DU/RIC split? In an O-RAN-style view a RIS is a new controllable node; an xApp/rApp in the RIC could own RIS beam decisions alongside scheduling, which is attractive for joint optimisation.

The reason 3GPP standardisation is non-trivial is precisely this control plane: the surface needs synchronised, low-latency configuration tied to the gNB's scheduling and beam decisions — and that interface, its signalling, and its timing all have to be specified before multi-vendor RIS becomes real.

A codebook beam-management loop (the pragmatic design)

Rather than estimate the full cascade channel, most realistic proposals mirror NR's P1/P2/P3 beam management with a RIS codebook:

1. Sweep — the gNB cycles the RIS through a small codebook of pre-stored phase patterns (each a candidate reflected beam), while the UE measures reference signals.
2. Report & select — the UE reports the best beam (RSRP/SINR); the gNB picks the winning RIS configuration — exactly like CSI-RS/SSB beam reporting.
3. Refine & track — a finer local codebook hones the beam, and the loop re-runs as blockage or the UE moves, with hysteresis to avoid thrashing.

Codebook control keeps signalling bounded and avoids the impossible task of a passive surface sounding its own channel — at the cost of some optimality versus full channel-state-based optimisation.

10 · Use cases and deployment scenarios

RIS is not a general booster; it is a precision tool. The scenarios where it genuinely shines:

• Blockage mitigation at high bands — mmWave (FR2) and the new FR3 mid-band are easily shadowed by a body, a bus, a wall. A RIS on a nearby façade provides a reliable engineered NLoS path, the headline coverage use case.
• Cell-edge & coverage-hole fill — extend a cell's reach into a courtyard, underpass or indoor corner cheaply, without backhaul or mains power for a full site.
• Indoor / venue coverage — walls and ceilings in stadiums, malls and offices become controllable reflectors, improving SNR and spatial reuse.
• Integrated Sensing and Communication (ISAC) — a RIS adds controllable, known reflection paths that sharpen radar-like sensing and imaging; the Rel-19 ISAC channel model in TR 38.901 is the modelling bridge.
• Positioning & NLoS localisation — a RIS at a known location is a controllable anchor; its near-field beamfocusing enables range-and-angle estimation even without a direct path, pushing toward 6G's centimetre-class positioning goals.
• Physical-layer security — steer a deep null toward an eavesdropper while peaking the beam at the legitimate user.
• Energy efficiency & sustainability — replacing some active small cells with passive surfaces is a core "green 6G" argument (and the framing ETSI ISG RIS leans into).

11 · The open challenges — what is really in the way

Stripping away the hype, these are the problems engineers and standards bodies are actually fighting:

• Multiplicative path loss — the structural one (§4). It forces large surfaces, near-endpoint placement, or active/BD variants, all of which erode the "cheap and passive" pitch.
• Channel estimation — a passive RIS has no receivers, so it cannot sound the channel itself. Estimating the full cascade (Tx→RIS and RIS→Rx, per element) is high-dimensional; practical systems lean on codebook sweeping, structured/low-rank estimation, and on-off element pilots — all with overhead.
• Control signalling & latency — re-configuring the surface in step with scheduling and mobility needs a fast, standardised control interface. Too slow, and the surface is always pointing where the user was.
• Mobility — keeping a focused (often near-field) beam locked onto a moving UE is harder than far-field angular steering and stresses the control loop.
• Deployment & ownership — who buys, sites, powers and maintains thousands of surfaces on third-party buildings? Site acquisition and the business model are as unsolved as the physics.
• Multi-operator & regulation — a surface alters the RF environment for everyone, not just its owner's UEs, raising coexistence, interference and regulatory questions a single repeater never did.
• Standardisation — without an agreed control interface, channel model and conformance framework (the Rel-20/21 + ETSI work), RIS stays single-vendor pilots rather than an ecosystem.
• Lab-to-field gap — the recurring story: striking gains in anechoic chambers and short ranges, far more modest numbers in real, cluttered, mobile deployments.

The honest takeaway

RIS is one of the most intellectually beautiful ideas in 6G: make the channel programmable and a whole new design dimension opens up — coverage, capacity, sensing, positioning and security all from a passive sheet on a wall. But beauty meets the product-distance law. The near-term truth is that the Network-Controlled Repeater already delivers network-steered coverage today, while RIS earns its place by being large, near an endpoint, and replacing genuinely blocked paths — and by maturing through the Rel-20 6G study, ETSI ISG RIS, and ITU's IMT-2030 framework toward possible normative status in Rel-21. Watch the channel models and the control interface: those, more than any lab gain figure, will decide whether the smart radio environment becomes real.

Quick glossary

Frequently asked questions