Ferkans — Interactive Telecom Tutor

The Inconvenient Truth of Cellular Networks

Every cellular network has a dirty secret: the performance you experience depends enormously on where you happen to be standing. A user 50 meters from a base station (BS) enjoys a strong, interference-free link. A user at the cell edge — equidistant from two or three base stations — suffers weak desired signal and strong inter-cell interference simultaneously. This chapter is about recognizing this problem clearly, understanding the partial fixes that have been tried (CoMP), and then rethinking the architecture entirely (cell-free massive MIMO).

Definition:
Cellular Network Model

A cellular network consists of $B$ base stations, each equipped with $N_t$ antennas, deployed over a coverage area $\mathcal{A}$ . The area is partitioned into $B$ non-overlapping cells $\mathcal{C}_1, \ldots, \mathcal{C}_B$ with $\bigcup_{b=1}^{B} \mathcal{C}_b = \mathcal{A}$ . Each user $k$ is associated with exactly one BS $b(k) = \arg\max_{b} \beta_{bk}$ , where $\beta_{bk}$ is the large-scale fading coefficient from BS $b$ to user $k$ . The downlink signal received by user $k$ is

$y_k = \underbrace{\sqrt{P_t} \, \mathbf{H}_{b(k),k}^{H} \mathbf{v}_{k} \, s_k}_{\text{desired signal}} + \underbrace{\sum_{j \neq k} \sqrt{P_t} \, \mathbf{H}_{b(j),k}^{H} \mathbf{v}_{j} \, s_j}_{\text{intra-cell interference}} + \underbrace{\sum_{b' \neq b(k)} \sum_{j \in \mathcal{C}_{b'}} \sqrt{P_t} \, \mathbf{H}_{b',k}^{H} \mathbf{v}_{j} \, s_j}_{\text{inter-cell interference}} + w_k$

where $w_k \sim \mathcal{CN}(0, \sigma^2)$ is AWGN.

The partition into cells is a design choice, not a physical law. Cell-free massive MIMO will abandon this partition entirely.

Definition:
Cell-Edge User

A user $k$ is a cell-edge user if its distance to the serving BS $b(k)$ is comparable to the inter-site distance (ISD), i.e., if there exist one or more interfering BSs $b' \neq b(k)$ such that

$\frac{\beta_{b(k),k}}{\beta_{b',k}} \lesssim \rho_{\text{edge}}$

for some threshold $\rho_{\text{edge}}$ close to 1 (e.g., $\rho_{\text{edge}} = 3$ dB). Cell-edge users simultaneously experience (i) a weak desired signal due to large path loss, and (ii) strong inter-cell interference from nearby BSs.

Cell-Edge User

A user located near the boundary between adjacent cells, where the ratio of desired-to-interfering signal power is small. Cell-edge users are the primary victims of the cellular architecture.

Theorem: Cell-Edge SINR Scaling

Consider a hexagonal cellular network with path-loss exponent $\alpha > 2$ and inter-site distance $d_{\text{ISD}}$ . A user at the cell edge (equidistant from the serving BS and one dominant interferer at distance $d_{\text{ISD}}/2$ ) with MRT precoding achieves downlink SINR

$\text{SINR}_k^{\text{edge}} = \frac{N_t \, \beta_{b(k),k}}{\sum_{b' \neq b(k)} \beta_{b',k} + \sigma^2 / P_t}$

In the interference-limited regime ( $P_t \to \infty$ ), this reduces to

$\text{SINR}_k^{\text{edge}} \to \frac{N_t \, d_{b(k),k}^{-\alpha}}{\sum_{b' \neq b(k)} d_{b',k}^{-\alpha}}$

which for a cell-edge user with one dominant interferer at the same distance gives $\text{SINR}_k^{\text{edge}} \approx N_t$ . Without massive MIMO ( $N_t = 1$ ), the cell-edge SINR is approximately 0 dB — barely decodable.

At the cell edge, the desired and interfering signals travel roughly the same distance, so path loss helps neither. The only lever is the array gain $N_t$ from the serving BS. Massive MIMO helps, but the geometry is fundamentally unfavorable.

Proof

System model

Under MRT precoding $\mathbf{v}_{k} = \hat{\mathbf{H}}_{b(k),k} / \|\hat{\mathbf{H}}_{b(k),k}\|$ , the effective desired signal power is $P_t \, |\mathbf{H}_{b(k),k}^{H} \mathbf{v}_{k}|^2$ . By channel hardening, $|\mathbf{H}_{b(k),k}^{H} \mathbf{v}_{k}|^2 \approx N_t \, \beta_{b(k),k}$ .

Interference characterization

The interference from BS $b'$ is $P_t \, |\mathbf{H}_{b',k}^{H} \mathbf{v}_{j}|^2$ . Since $\mathbf{H}_{b',k}$ is independent of $\mathbf{v}_{j}$ (different cell), $\mathbb{E}[|\mathbf{H}_{b',k}^{H} \mathbf{v}_{j}|^2] = \beta_{b',k}$ (no array gain for the interferer).

SINR expression

The SINR follows by dividing desired signal power by total interference-plus-noise. At the cell edge with the dominant interferer at equal distance, the ratio simplifies to $N_t \cdot (\text{path loss ratio})$ .

Cell-edge limit

When $d_{b(k),k} \approx d_{b',k}$ , the path loss ratio is approximately 1, yielding $\text{SINR}_k^{\text{edge}} \approx N_t$ . For $N_t = 1$ , this gives 0 dB. $\blacksquare$

Example: Cell-Edge SINR in a 7-Cell Hexagonal Layout

Consider a 7-cell hexagonal network with ISD = 500 m, path-loss exponent $\alpha = 3.8$ , and $N_t = 64$ antennas per BS. A user sits at the boundary between cell 1 and cell 2, at distance 250 m from each. The remaining 5 BSs are at distances approximately 433 m, 500 m, 500 m, 433 m, and 250 m (wrapping). Compute the cell-edge SINR in the interference-limited regime.

Solution

Path-loss computation

With $\alpha = 3.8$ : $\beta_{1,k} = 250^{-3.8} = 1.91 \times 10^{-10}$ , $\beta_{2,k} = 250^{-3.8} = 1.91 \times 10^{-10}$ (dominant interferer, same distance). For the other 5 BSs at larger distances, the path losses are significantly smaller.

Dominant-interferer approximation

Retaining only the dominant interferer (BS 2 at 250 m): $\text{SINR}_k^{\text{edge}} \approx \frac{64 \cdot 1.91 \times 10^{-10}}{1.91 \times 10^{-10}} = 64 = 18.1 \text{ dB}$

With all 6 interferers

Including all interfering BSs: $\sum_{b' \neq 1} \beta_{b',k} \approx 3.2 \times 10^{-10}$ . $\text{SINR}_k^{\text{edge}} \approx \frac{64 \cdot 1.91 \times 10^{-10}}{3.2 \times 10^{-10}} \approx 38.2 = 15.8 \text{ dB}$ Massive MIMO provides enough array gain to push the cell-edge SINR above 15 dB. Without it ( $N_t = 1$ ), the SINR would be $-2.2$ dB — requiring very low-rate coding.

CDF of Per-User SINR: Cellular vs Cell-Free

Compare the cumulative distribution function of per-user SINR for a conventional cellular network versus a cell-free deployment. Adjust the number of antennas, users, and APs to see how the cell-free architecture eliminates the low-SINR tail that plagues cell-edge users.

Parameters

Number of BSs / APs7

N_t

(antennas per BS/AP)64

K

(users)20

\text{SNR}

(dB)10

Architecture

Common Mistake: Average Throughput Hides the Cell-Edge Problem

Mistake:

Reporting the average per-user throughput across the cell as the performance metric. This hides the fact that cell-edge users may have throughput 10--100 times lower than cell-center users.

Correction:

The correct metric for fairness is the 5th percentile (or 95%-likely) per-user rate. This captures the worst-case user experience. In 3GPP evaluations, the cell-edge throughput is defined as the 5th percentile of the per-user rate CDF. A network with high average but poor 5th-percentile rate has unacceptable coverage.

Definition:
Inter-Cell Interference (ICI)

In a cellular network, inter-cell interference is the aggregate interference received by user $k$ in cell $b(k)$ from all base stations $b' \neq b(k)$ :

$I_k^{\text{ICI}} = \sum_{b' \neq b(k)} \sum_{j \in \mathcal{C}_{b'}} P_t \, |\mathbf{H}_{b',k}^{H} \mathbf{v}_{j}|^2$

ICI is the dominant performance-limiting factor at the cell edge. Unlike intra-cell interference, which can be mitigated by spatial precoding at the serving BS, ICI arises from base stations that have no knowledge of user $k$ 's channel.

The key insight of CoMP and cell-free architectures is that ICI can be converted into useful signal if the interfering BSs cooperate.

Inter-Cell Interference (ICI)

Interference caused by transmissions from base stations in adjacent cells. ICI is the fundamental performance limiter for cell-edge users in conventional cellular networks.

Historical Note: The Cellular Concept: From AT&T Bell Labs to the Cell-Edge Problem

1947–present

The cellular concept was conceived at Bell Labs in 1947 by Douglas Ring and W. Rae Young, who proposed dividing a geographic area into "cells" served by low-power transmitters. The idea was to enable frequency reuse: distant cells could share the same frequencies because the signals would attenuate sufficiently. This was a breakthrough that made mobile telephony scalable. But the price was inter-cell interference at cell boundaries — a problem that was recognized from the very beginning but considered acceptable given the alternative of no coverage at all. Nearly 80 years later, we are finally questioning whether the cell boundary is a necessary evil or a design artifact that can be eliminated.

Quick Check

In a two-cell network where a cell-edge user is equidistant from both BSs (each with $N_t = 1$ antenna), what is the approximate SINR in the interference-limited regime?

$-3$ dB

0 dB

3 dB

$N_t$ (linear)

Correction:

0 dB

Equal path loss to serving and interfering BS means $\text{SINR} \approx \beta_{\text{serve}} / \beta_{\text{interfere}} = 1 = 0$ dB. The desired and interfering signals are equally strong.

Key Takeaway

The cell-edge problem is not a minor nuisance — it is the fundamental performance bottleneck of cellular networks. The 5th-percentile user rate in a conventional cellular deployment can be 10 to 50 times lower than the median rate. Any architecture that claims to improve on cellular must demonstrate gains in this tail, not just in the average.

Cell-Edge Performance in 5G NR

Even with massive MIMO in 5G NR (up to 64 antenna ports at sub-6 GHz), the cell-edge problem persists. The array gain $N_t$ helps, but the fundamental geometry — cell-edge users are far from the serving BS and close to interferers — limits the improvement. 3GPP Release 16 introduced multi-TRP (Transmission-Reception Point) operation as a partial CoMP solution, but the gains are modest (2–3 dB at the cell edge in typical deployments).

The Cell Boundary Problem

The Inconvenient Truth of Cellular Networks

Definition: Cellular Network Model

Definition: Cell-Edge User

Cell-Edge User

Theorem: Cell-Edge SINR Scaling

System model

Interference characterization

SINR expression

Cell-edge limit

Example: Cell-Edge SINR in a 7-Cell Hexagonal Layout

Path-loss computation

Dominant-interferer approximation

With all 6 interferers

CDF of Per-User SINR: Cellular vs Cell-Free

Parameters

Common Mistake: Average Throughput Hides the Cell-Edge Problem

Definition: Inter-Cell Interference (ICI)

Inter-Cell Interference (ICI)

Historical Note: The Cellular Concept: From AT&T Bell Labs to the Cell-Edge Problem

Quick Check

Key Takeaway

Cell-Edge Performance in 5G NR

Definition:
Cellular Network Model

Definition:
Cell-Edge User

Definition:
Inter-Cell Interference (ICI)