Phase Selection and Analysis for Multi-frequency Multi-user RIS Systems Employing Subsurfaces

This paper proposes a low-complexity RIS design where each user is served by a dedicated subsurface on a unique frequency band, deriving closed-form performance metrics that demonstrate optimal Line-of-Sight operation and remarkable robustness to non-Line-of-Sight conditions while significantly reducing channel estimation and processing requirements.

The Gist

Imagine you are at a massive, crowded concert. The singer (the Base Station) is trying to talk to hundreds of different fans (the Users) scattered around the stadium. But there's a problem: the stadium is huge, there are pillars blocking the view, and the singer's voice is getting lost in the noise.

Enter the RIS (Reconfigurable Intelligent Surface). Think of the RIS as a giant, high-tech wall of thousands of tiny, smart mirrors placed between the singer and the fans. Each mirror can tilt slightly to catch the singer's voice and bounce it directly to a specific fan.

The Old Way: The "Group Hug" Problem

Traditionally, engineers tried to make all these mirrors work together to talk to everyone at once on the same frequency.

• The Problem: It's like trying to conduct an orchestra where every musician is playing a different song at the same time. To make it work, the conductor (the computer) has to do incredibly complex math to figure out exactly how to tilt every single mirror. It's slow, expensive, and requires knowing exactly where every fan is standing (Channel Estimation).
• The Result: High complexity, high cost, and a system that gets confused easily.

The New Idea: The "Subsurface" Strategy

This paper proposes a clever, simpler way to run the concert. Instead of one giant wall trying to do everything at once, they split the wall into subsurfaces (smaller sections).

Here is how it works, using our concert analogy:

1. Frequency Separation (Different Radio Stations):
   Imagine the stadium has different radio frequencies. Fan A listens to 101.1 FM, Fan B to 102.5 FM, and so on.

   • The RIS wall is divided into sections. Section 1 is dedicated to 101.1 FM, Section 2 to 102.5 FM, etc.
   • Because they are on different frequencies, the mirrors for Fan A don't need to worry about Fan B. They can just focus on their own job.

2. The "Uncontrolled" Mirrors:
   What about the mirrors in Section 2 that are supposed to help Fan B, but Fan A is listening to 101.1?

   • In the old days, engineers would try to make those mirrors "disappear" or stay perfectly still.
   • The Paper's Insight: Actually, let them bounce the sound around randomly! It turns out that in a noisy stadium, having a few extra bounces (scattering) can actually help Fan A hear better, especially if the direct line of sight is blocked. It's like having a few extra echoes in a cave that help you hear the singer even if you can't see them.

3. The "Simple" Math (LoS vs. NLoS):

   • Line of Sight (LoS): If the singer and the fan have a clear view of each other, the math to tilt the mirrors is simple. It's like aiming a laser pointer. The paper gives a simple formula for this.
   • No Line of Sight (NLoS): If there are pillars blocking the view, the sound bounces off walls. The paper suggests a "best guess" method: look at the strongest path the sound takes, pretend it's a straight line, and aim the mirrors that way.
   • The Magic: Even when the sound is bouncing off walls (which is messy), this "best guess" method works almost as well as the super-complex math, but it's much faster to calculate.

Why is this a Big Deal? (The Trade-Off)

The paper compares their new method to the "Super-Complex" method (called TMSE in the paper).

• The Super-Complex Method: Tries to get the absolute maximum volume for everyone combined. It's like a genius conductor trying to make the whole orchestra sound perfect simultaneously.

  • Pros: Slightly higher total volume.
  • Cons: Requires a supercomputer, needs to know the exact location of every single fan, and is very slow to set up.

• The New "Subsurface" Method: Gives each fan their own dedicated mirror section.

  • Pros:
    • Super Fast: The math is simple enough to run on a basic chip.
    • Easy Setup: You don't need to know exactly where the fans are; you just need to know which section of the wall is for them.
    • Fairness: Every fan gets an equal share of the mirrors. The complex method sometimes ignores the "quiet" fans to make the "loud" ones louder.
  • Cons: Because each fan only gets a slice of the wall, the total volume isn't quite as high as the super-complex method. However, the paper argues that the speed and simplicity are worth the tiny drop in volume.

The "Mirror Arrangement" Surprise

The paper also looked at how to arrange the mirrors on the wall.

• Grouped: Put all mirrors for Fan A together in a block.
• Interleaved: Mix them up (Fan A, Fan B, Fan A, Fan B...).

The Surprise: The paper found that Grouped mirrors work better!

• Why? It sounds counter-intuitive, but when mirrors are close together, they "talk" to each other (correlation). In a noisy environment, this teamwork actually helps boost the signal. It's like a choir standing close together; their voices blend and reinforce each other better than if they were spread out across the stadium.

Summary

This paper says: "Stop trying to be a genius conductor for the whole orchestra. Just give every section of the choir their own sheet music and let them sing their own song."

By splitting the problem into smaller, independent pieces (subsurfaces) and accepting a little bit of "noise" (uncontrolled scattering) as a helper, we can build wireless systems that are:

1. Cheaper (less computing power needed).
2. Faster (easier to set up).
3. More Robust (work even when the signal is blocked).

It's a shift from "perfect but fragile" to "good enough, simple, and reliable."

Technical Summary

Here is a detailed technical summary of the paper "Phase Selection and Analysis for Multi-frequency Multi-user RIS Systems Employing Subsurfaces."

1. Problem Statement

Reconfigurable Intelligent Surfaces (RIS) offer a low-power method to manipulate wireless channels between a Base Station (BS) and Users (UEs). However, designing optimal phase shifts for Multi-User (MU) systems presents significant challenges:

• Computational Complexity: Existing optimal MU designs often rely on iterative algorithms (e.g., Interior Point Algorithms) with high computational complexity, making them impractical for real-time deployment.
• Channel Estimation: Estimating the full channel state information (CSI) for all RIS elements to all users is a major bottleneck.
• Receiver Processing: Optimal MU reception often requires complex Minimum Mean Square Error (MMSE) processing.
• LoS vs. NLoS: While optimal closed-form solutions exist for single-user (SU) Line-of-Sight (LoS) scenarios, extending these to multi-user or Non-Line-of-Sight (NLoS) environments is analytically difficult.

The authors propose a system architecture that drastically reduces complexity by dividing the RIS into subsurfaces, where each subsurface is dedicated to a specific user operating on a unique frequency band.

2. Methodology

System Model

• Architecture: A RIS with $N$ elements is divided into $K$ subsurfaces (one per user). User $k$ operates on a dedicated frequency band.
• Signal Processing: The system uses Matched Filtering (MF) at the receiver. Since each user has a dedicated frequency, the receiver does not need to separate users via complex spatial processing; it simply filters for the specific frequency.
• Uncontrolled Scattering: RIS elements designed for other users act as uncontrolled scatterers for the current user. In high-frequency systems (where scattering is naturally low), these "unused" elements provide beneficial additional scattering paths.

Phase Selection Algorithms

The paper proposes two specific design methods:

1. Subsurface Design (SD) for LoS Channels:

   • Adapts the optimal single-user LoS phase selection method (from prior work [1]) to the multi-user context.
   • For user $k$, the subsurface phases are set to align the reflected signal with the direct BS-RIS steering vector and the UE-RIS channel.
   • Formula: $\Phi_k = \nu_k \text{diag}(e^{j\angle a_{r,k}}) \text{diag}(e^{-j\angle h_{ru,k,k}})$, where $\nu_k$ aligns the BS steering vector with the direct channel.

2. Extended Subsurface Design (ESD) for NLoS Channels:

   • Addresses scenarios where the BS-RIS channel is scattered (Rank > 1).
   • Since no optimal closed-form solution exists for general NLoS MU systems, the authors approximate the channel using the leading singular vector of the BS-RIS channel matrix ($H_{br,k,k}$).
   • The design replaces the LoS steering vector with the dominant right singular vector ($v_k$) of the channel matrix.
   • Formula: $\Phi_k = \omega_k \text{diag}(e^{j\angle v_k}) \text{diag}(e^{-j\angle h_{ru,k}})$.

Analytical Derivations

• Mean SNR: The authors derive an exact closed-form expression for the expected Signal-to-Noise Ratio (SNR) in LoS scenarios. This derivation accounts for the direct path, the designed subsurface path, and the interference/scattering from other subsurfaces.
• Rate Upper Bound: Using Jensen's inequality, they derive an upper bound for the expected data rate based on the mean SNR.

3. Key Contributions

1. Low-Complexity MU Design: A novel phase selection method that transforms a complex MU problem into $K$ independent Single-User (SU) problems by utilizing frequency separation and subsurfaces.
2. Robust NLoS Extension: An "Extended Subsurface Design" (ESD) that leverages the dominant singular vector of the channel, performing near-optimally even when the LoS assumption is violated.
3. Closed-Form Analysis: Exact mathematical expressions for mean SNR and rate bounds in LoS scenarios, providing theoretical insights into how correlation and system dimensions affect performance.
4. Complexity Quantification: A rigorous comparison showing that the proposed method reduces RIS design complexity by a factor of roughly $(K + K^2)/2$ compared to TMSE (Total Mean Squared Error) minimization methods, and reduces channel estimation requirements by a factor of $K$.
5. Layout Optimization: Investigation into physical RIS element arrangement, concluding that grouping elements for a user (contiguous blocks) outperforms interleaving them, due to the beneficial effects of spatial correlation on SNR.

4. Numerical Results & Findings

• Performance vs. Optimal (IPA):
  • In LoS conditions, the proposed SD matches the theoretical optimum.
  • In NLoS conditions (scattering), the ESD performs remarkably well, achieving 84–90% of the performance of an optimal Interior Point Algorithm (IPA) solution, despite having significantly lower complexity.
• Impact of Correlation:
  • Counter-intuitively, higher spatial correlation between RIS elements improves the mean SNR.
  • The analysis shows that grouping elements closer together (increasing correlation) yields better performance than interleaving them.
• User Scaling ($K$):
  • If the total number of RIS elements ($N$) is fixed, adding more users reduces the SNR per user (fewer elements per subsurface).
  • However, if $N$ scales with $K$ (keeping elements per user constant), adding users is beneficial because the "uncontrolled" scattering from other users' subsurfaces acts as a diversity gain.
• Trade-offs:
  • Rate: The proposed method yields lower aggregate data rates compared to full MU schemes (like TMSE) because of the bandwidth restriction (each user gets $1/K$ of the total bandwidth).
  • Fairness & Complexity: The proposed method offers superior fairness (Jain's fairness index ~0.56 vs 0.42 for TMSE) and drastically lower complexity in design, receiver processing, and CSI estimation.

5. Significance

This paper provides a practical pathway for deploying RIS in 5G/6G multi-user networks where computational resources and channel estimation overhead are limiting factors. By accepting a slight reduction in aggregate spectral efficiency (due to frequency splitting), the system achieves:

• Drastic reduction in hardware and processing complexity.
• Robustness to channel estimation errors and NLoS conditions.
• Scalability that allows for simple, closed-form design rather than heavy iterative optimization.

The findings suggest that for high-frequency systems (where scattering is scarce), the "uncontrolled" scattering provided by subsurfaces dedicated to other users is a feature, not a bug, enhancing the overall system performance.