Background: RAN Functional Splits
Several options for splitting RAN functions have been proposed by 3GPP, going from Option 1 to Option 8. Every split has its own pros and cons. It turns out that the 7.2x split offers the best balance between RU complexity and inter-cell cooperation — and has become the dominant choice for Open RAN deployments.
This post discusses RAN split 7.3 in detail. In the 7.3 split, modulation and demodulation functions are placed closer to the antennas than in the 7.2x split, while still upstream of the RF components.
What Moves in the 7.3 Split?
In a 7.3 split, the functional boundary between the DU (Distributed Unit) and RU (Radio Unit) is positioned such that:
- Downlink: Hard bits (modulated bit decisions) are sent from DU to RU. The RU then performs the modulation (mapping bits to constellation symbols) and RF transmission.
- Uplink: Soft bits — specifically LLRs (Log-Likelihood Ratios) — are sent from RU to DU. The RU demodulates the received signal and computes LLRs before forwarding them for channel decoding at the DU.
In the 7.3 split, hard bits are sent in the downlink while soft bits (LLRs) are sent in the uplink.
Downlink Fronthaul Capacity
The required fronthaul capacity for the 7.3 split in downlink is given by:
Where:
| Parameter | Description |
|---|---|
Nsc | Number of subcarriers |
Nlayers | Number of MIMO layers |
Qm | Modulation order (e.g., 8 bits for 256 QAM) |
Ts | Average OFDM symbol duration |
Example (Downlink): Single carrier, 100 MHz bandwidth, 30 kHz SCS (Nsc = 3276, Ts = 33.33 µs), 256 QAM (Qm = 8), 16 layers:
This is a huge gain in fronthaul capacity efficiency compared to split 7.2x and Option 8, which must carry I/Q samples instead of bits.
Uplink Fronthaul Capacity — The LLR Cost
In the uplink, instead of hard bits, LLRs are sent from the RU to the DU. The LLR representation requires more bits per coded bit than the hard decision, so the uplink capacity requirement is higher:
Where Sbw is the soft bit width. Common values: 6, 8, 12, or 16 bits.
Example (Uplink): Same configuration as above, with Sbw = 8 bits:
This is 8× higher than the downlink requirement — a direct consequence of using soft bits instead of hard decisions.
Key Properties of Split 7.3
✓ DL Advantages
Bits instead of I/Q samples dramatically reduces DL fronthaul bandwidth vs. Option 8. Capacity scales with modulation order and layers.
⚠ UL Consideration
Soft bits (LLRs) in the uplink increase fronthaul requirements proportionally to the soft bit width (Sbw). This can be 6–16× larger than DL.
✓ Modulation Flexibility
DL fronthaul capacity in 7.3 depends on modulation order — operators can trade off capacity and throughput by adjusting MCS.
≠ vs. Split 7.2x
In 7.2x, fronthaul carries I/Q samples and is modulation-independent. In 7.3, it carries bits — so the capacity depends on MCS choice.
Comparison: 7.3 vs Other Splits
| Split Option | DL Transport | UL Transport | Modulation Dependence |
|---|---|---|---|
| Option 8 | I/Q samples | I/Q samples | No |
| Split 7.2x | I/Q samples | I/Q samples | No |
| Split 7.3 | Hard bits | Soft bits (LLR) | Yes (DL) |
| Split 6 | Coded bits | Coded bits | Yes |
Why Not Everyone Uses 7.3
While split 7.3 offers better DL fronthaul efficiency than 7.2x and Option 8, its uplink cost can be prohibitive. For a 100 MHz carrier with 16 layers and 8-bit LLRs, the 100+ Gbps uplink fronthaul requirement exceeds what most optical fronthaul links can economically deliver at scale.
This is one reason the industry has largely converged on split 7.2x for Open RAN deployments — it offers a good balance of RU simplicity and practical fronthaul requirements without the UL soft-bit penalty.
References
- 3GPP TR 38.801 — Study on new radio access technology: Radio access architecture and interfaces
- Cloud-RAN functional split for an efficient fronthaul network (IEEE Xplore, 2020)