Open RAN Architecture

From Theory to Architecture: Open RAN

The fronthaul strategies developed in Sections 14.1--14.4 are theoretical frameworks. Translating them into deployable systems requires a standardized network architecture that defines where processing happens, how data flows, and what interfaces connect the components. Open RAN --- promoted by the O-RAN Alliance --- provides this architecture, decomposing the base station into Radio Unit (RU), Distributed Unit (DU), and Central Unit (CU) with open interfaces between them.

Definition:
O-RAN Functional Components

The O-RAN architecture decomposes the base station into three functional components:

Radio Unit (RU): The remote radio head at the antenna site. Handles RF front-end, digital-to-analog conversion, and (depending on the split) low-PHY processing (FFT, cyclic prefix).

Distributed Unit (DU): Handles real-time baseband processing. In the 7.2x split, the DU performs channel estimation, precoding, MIMO processing, and scheduling. Connected to the RU via fronthaul.

Central Unit (CU): Handles non-real-time processing. Protocol stack (PDCP, RRC, SDAP), mobility management, and inter-cell coordination. Connected to the DU via midhaul.

In a cell-free context, the RU maps to an AP, and the DU/CU together form the central processor (CPU).

The O-RAN Alliance also defines the near-RT RIC (near-real-time RAN Intelligent Controller) and non-RT RIC for AI/ML-based network optimization --- these connect via the E2 and A1 interfaces.

Definition:
Functional Split Options

3GPP defines eight functional split options between the RU and DU, numbered 1--8 from highest layer (Option 1: RRC/PDCP split) to lowest layer (Option 8: RF/PHY split, equivalent to CPRI).

The most relevant splits for cell-free MIMO are:

Option 8 (PHY-RF): All baseband at DU. Fronthaul carries time-domain I/Q samples. Rate: $\propto W \cdot N_t \cdot 2b$ .
Option 7.2x (intra-PHY): FFT and CP at RU. Fronthaul carries frequency-domain I/Q. Rate: $\propto W \cdot N_t \cdot 2b / \alpha$ where $\alpha$ accounts for resource block pruning.
Option 7.1 (intra-PHY): RU also handles beamforming. Fronthaul carries beam-domain signals. Rate: $\propto W \cdot N_{\text{beams}} \cdot 2b$ .
Option 6 (MAC-PHY): Full PHY at RU. Fronthaul carries transport blocks. Rate: $\propto$ user data rate.

The split determines the fronthaul-computation tradeoff: lower splits (closer to RF) require higher fronthaul but less RU processing, while higher splits require lower fronthaul but more RU processing.

Functional Split Options for Cell-Free MIMO

Split	Fronthaul Data	Rate Scaling	RU Complexity	Cell-Free Suitability
Option 8 (CPRI)	Time-domain I/Q	$\propto W \cdot N_t$	Minimal	Impractical for large arrays
Option 7.2x (O-RAN)	Freq-domain I/Q	$\propto W \cdot N_t$	Low (FFT, CP)	Standard for O-RAN cell-free
Option 7.1	Beam-domain signals	$\propto W \cdot N_{\text{beams}}$	Medium (beamforming)	Good for hybrid arrays
Option 6 (MAC-PHY)	Transport blocks	$\propto$ data rate	High (full PHY)	Best for fronthaul but limited coordination

Fronthaul Requirements by Functional Split

Compare the fronthaul rate requirements for different functional splits as the number of antennas per RU and the system bandwidth vary.

Parameters

W

(MHz)100

N_t

per RU32

I/Q bits12

N_{\text{beams}}

Number of beams for Option 7.1

Historical Note: The Open RAN Movement

2018--present

The O-RAN Alliance was formed in 2018 by the merger of the C-RAN Alliance and xRAN Forum. Its mission was to disaggregate the traditional monolithic base station into interoperable components from different vendors. The 7.2x split was standardized in 2019 as the primary fronthaul interface. By 2023, O-RAN-compliant equipment was deployed by major operators including Vodafone, Rakuten, and Dish Network. The architecture is particularly relevant for cell-free massive MIMO because it provides a standardized framework for connecting distributed RUs to centralized DUs --- exactly the AP-to-CPU architecture studied in this chapter.

Example: Mapping Cell-Free Massive MIMO onto O-RAN

A cell-free massive MIMO network has $L = 32$ APs, each with $N_t = 4$ antennas, serving $K = 16$ users over $W = 100$ MHz. Map this onto an O-RAN deployment and compute the fronthaul rate for Option 7.2x with 12-bit I/Q resolution.

Solution

Map cell-free components to O-RAN

Each AP $\to$ one O-RAN RU with 4 antenna ports
The central processor $\to$ one or more O-RAN DUs
Each DU can serve a cluster of RUs (e.g., 8 RUs per DU $\to$ 4 DUs)

Compute per-RU fronthaul rate (Option 7.2x)

With Option 7.2x, the RU performs FFT and CP removal. The fronthaul carries frequency-domain symbols: $R_{\text{fh}} = 2 \cdot 12 \cdot 100 \times 10^6 \cdot 4 \cdot \frac{16}{15} \approx 10.2 \text{ Gbps per RU}$

Total fronthaul capacity

Total: $32 \times 10.2 = 326$ Gbps. With resource block pruning (active PRBs typically 60--80% of total), the effective rate is approximately 200--260 Gbps. This is feasible with a switched Ethernet fronthaul using 25G links (13 links total) or a fiber ring with wavelength-division multiplexing.

Compare with estimate-and-forward

Using EF (Section 14.2), each RU forwards $K = 16$ complex scalars per subcarrier instead of $N_t = 4$ antenna signals. Since $K > N_t$ in this example, EF does not reduce fronthaul here. EF's advantage requires $N_t \gg K$ , motivating larger per-AP antenna arrays.

🚨Critical Engineering Note

Fronthaul Latency Constraints in O-RAN

The O-RAN 7.2x interface imposes strict one-way latency requirements:

Downlink: The DU must deliver I/Q data to the RU with less than 100 $\mu$ s latency to allow for RU processing before the air interface slot boundary.
Uplink: The RU must forward received I/Q data to the DU within 100--250 $\mu$ s for timely HARQ processing.

These latency constraints limit the fronthaul distance to approximately 10--20 km over fiber, or require dedicated low-latency Ethernet with time-sensitive networking (TSN). For cell-free deployments covering large areas, the DU must be placed centrally with fiber reach to all RUs.

Practical Constraints

•
One-way fronthaul latency: < 100 us (DL), < 250 us (UL)
•
Maximum fiber distance: ~10-20 km at 5 us/km propagation
•
TSN (IEEE 802.1CM) required for Ethernet fronthaul

📋 Ref: O-RAN WG4, Fronthaul Transport Requirements

⚠️Engineering Note

Security Considerations for Open Fronthaul

The open fronthaul interface exposes I/Q samples --- the raw physical layer data --- on a potentially multi-vendor network. This raises security concerns:

Eavesdropping: Unencrypted fronthaul allows interception of all user data before physical-layer encryption.
Tampering: A compromised RU can inject malicious signals.
Availability: Fronthaul disruption takes out entire cells.

The O-RAN Alliance specifies optional MACsec encryption on the fronthaul, but encryption adds latency (5--10 $\mu$ s) and processing overhead. Most current deployments rely on physical security (dark fiber in owned ducts) rather than cryptographic protection.

Practical Constraints

•
MACsec encryption adds 5-10 us latency
•
Physical-layer security insufficient for shared infrastructure

Theorem: Rate-Fronthaul Tradeoff Across Functional Splits

For a cell-free system with $L$ RUs, each with $N_t$ antennas, serving $K$ users, the achievable sum rate $R_{\text{sum}}(C_{\text{fh}})$ under optimal processing satisfies:

Split 8 (full centralization): $R_{\text{sum}}$ achieves the centralized MMSE rate when $C_{\text{fh},l} \geq N_t \log_2(1 + \text{SNR}_{\max})$ per RU.
Split 6 (full distribution): $R_{\text{sum}}$ is limited by the local processing quality but requires only $C_{\text{fh},l} \geq \sum_k R_k$ per RU.

The gap between these extremes is bounded by: $R_{\text{sum}}^{\text{Split 8}} - R_{\text{sum}}^{\text{Split 6}} \leq L \cdot N_t \cdot \log_2\!\left(\frac{\text{SNR}_{\max}}{\text{SNR}_{\min}}\right)$ which quantifies the cost of distributed processing in terms of the dynamic range of the received signals.

Full centralization (Split 8) achieves the best rate but requires enormous fronthaul. Full distribution (Split 6) minimizes fronthaul but sacrifices cooperation gain. The gap between them is largest when the dynamic range across users is large (near-far effect), because centralized processing can jointly optimize across all signal levels.

Proof

Split 8 achievability

With sufficient fronthaul, the CPU has access to all raw observations and can compute the centralized MMSE receiver. The required fronthaul rate is bounded by the mutual information $I(\mathbf{y}_l; \mathbf{x}) \leq N_t \log_2(1 + \text{SNR}_{\max})$ .

Split 6 limitation

With full PHY at the RU, each RU independently decodes and forwards decoded bits. The achievable rate per RU is limited by the local SINR, without inter-RU cooperation gain.

Gap bound

The cooperation gain from centralized processing is bounded by the log-det of the ratio of centralized to distributed interference-plus-noise covariance matrices. The dynamic range factor captures the worst-case gap. $\blacksquare$

Quick Check

Which O-RAN functional split is most commonly used for cell-free massive MIMO deployments?

Option 8 (full CPRI)

Option 7.2x (intra-PHY with FFT at RU)

Option 6 (full PHY at RU)

Option 1 (PDCP split)

Correction:

Option 7.2x (intra-PHY with FFT at RU)

Option 7.2x is the O-RAN standard split. The RU handles FFT and CP, reducing fronthaul while keeping MIMO processing at the DU.

Quick Check

In the O-RAN architecture, which component performs MIMO precoding and channel estimation under the 7.2x split?

The Radio Unit (RU)

The Distributed Unit (DU)

The Central Unit (CU)

The RAN Intelligent Controller (RIC)

Correction:

The Distributed Unit (DU)

The DU performs all real-time baseband processing including channel estimation, precoding, and scheduling.

Key Takeaway

Open RAN provides the standardized architecture for deploying cell-free massive MIMO. The 7.2x functional split strikes a balance between fronthaul efficiency and centralized processing. The choice of split directly determines the fronthaul capacity requirement and the achievable cooperation level --- making the information-theoretic analysis of this chapter essential for O-RAN system design.

Why This Matters: Open RAN and the Path to 6G Cell-Free

The O-RAN architecture is widely considered the deployment platform for 6G cell-free networks. The standardized RU-DU-CU decomposition maps naturally onto the cell-free AP-CPU model: each RU is an AP, and the DU/CU complex forms the central processor. The fronthaul strategies analyzed in this chapter (QF, EF, compression-based precoding) can be implemented within the O-RAN framework, with the functional split determining the specific strategy.

The key 6G challenges include: (1) supporting sub-THz bands with much larger bandwidths (requiring higher fronthaul), (2) enabling AI/ML-based RAN optimization through the RIC, and (3) achieving the "cell-free experience" at scale with thousands of RUs.

O-RAN (Open RAN)

A network architecture promoted by the O-RAN Alliance that disaggregates the base station into interoperable components (RU, DU, CU) with open, standardized interfaces. Enables multi-vendor deployments and is the primary architecture for cell-free massive MIMO.

Related: Fronthaul, Functional Split

Functional Split

The division of baseband processing between the radio unit and the baseband unit. 3GPP defines eight options; the O-RAN 7.2x split places FFT and CP processing at the RU and MIMO/MAC processing at the DU.

Joint Fronthaul Load Balancing Chapter Summary

Open RAN Architecture

From Theory to Architecture: Open RAN

Definition: O-RAN Functional Components

Definition: Functional Split Options

Functional Split Options for Cell-Free MIMO

Fronthaul Requirements by Functional Split

Parameters

Historical Note: The Open RAN Movement

Example: Mapping Cell-Free Massive MIMO onto O-RAN

Map cell-free components to O-RAN

Compute per-RU fronthaul rate (Option 7.2x)

Total fronthaul capacity

Compare with estimate-and-forward

Fronthaul Latency Constraints in O-RAN

Security Considerations for Open Fronthaul

Theorem: Rate-Fronthaul Tradeoff Across Functional Splits

Split 8 achievability

Split 6 limitation

Gap bound

Quick Check

Quick Check

Key Takeaway

Why This Matters: Open RAN and the Path to 6G Cell-Free

O-RAN (Open RAN)

Functional Split

Definition:
O-RAN Functional Components

Definition:
Functional Split Options