What processing does the baseband section perform?

The baseband section handles all digital signal processing between the transport network and the radio unit. Key functions: 1) Channel coding/decoding (LDPC for data, Polar codes for control in 5G NR). 2) Modulation mapping (QPSK through 256-QAM). 3) OFDM/FFT processing (IFFT for TX, FFT for RX). 4) MIMO precoding and beamforming weight calculation. 5) Resource scheduling (allocating time-frequency resources to users every 0.5ms slot). 6) HARQ retransmission management. 7) Higher-layer protocols (RLC segmentation, PDCP ciphering, RRC signaling). Modern baseband processors handle 100+ Gbps aggregate throughput.

How does the O-RAN CU/DU split work?

O-RAN disaggregates the traditional monolithic BBU into three functional units. The CU (Centralized Unit) handles non-real-time Layer 3 (RRC) and PDCP functions, can be virtualized on general-purpose servers, and centralized to serve multiple cells. The DU (Distributed Unit) handles time-critical Layer 1 (PHY-high) and Layer 2 (MAC/RLC) functions, requires low-latency processing, and is typically deployed at or near the cell site. The RU (Radio Unit) handles low-PHY (FFT/beamforming), DAC/ADC, and RF transmission. The DU-RU interface uses eCPRI over Ethernet, requiring <100μs one-way latency and 25 Gbps bandwidth for 64T64R massive MIMO.

What is the processing capacity of a modern baseband unit?

A single modern BBU card (e.g., Ericsson Baseband 6630) supports 6-12 cells with 100 MHz NR bandwidth each, processing 40+ Gbps aggregate user plane throughput. The processing is dominated by LDPC decoding (40-60% of compute) and FFT/beamforming (20-30%). For 64T64R massive MIMO with 100 MHz bandwidth, the baseband must process: 64 antenna streams × 100 MHz × 30.72 Msps × 2 (I/Q) × 16 bits = ~63 Gbps of raw IQ data per cell. This drives the transition from custom ASICs to GPU/FPGA acceleration in virtualized RAN (vRAN) deployments.

RAN Architecture

Baseband Section

Q: What is the processing capacity of a modern baseband unit?

A single modern BBU card (e.g., Ericsson Baseband 6630) supports 6-12 cells with 100 MHz NR bandwidth each, processing 40+ Gbps aggregate user plane throughput. The processing is dominated by LDPC decoding (40-60% of compute) and FFT/beamforming (20-30%). For 64T64R massive MIMO with 100 MHz bandwidth, the baseband must process: 64 antenna streams × 100 MHz × 30.72 Msps × 2 (I/Q) × 16 bits = ~63 Gbps of raw IQ data per cell. This drives the transition from custom ASICs to GPU/FPGA acceleration in virtualized RAN (vRAN) deployments.

/BAYSS-band SEK-shun/

The digital processing subsystem in a base station responsible for Layer 1/2/3 protocol processing: channel coding/decoding (LDPC, Polar), modulation/demodulation, OFDM FFT, beamforming weight calculation, resource scheduling, and HARQ management. In traditional architecture, implemented as the monolithic BBU. In 5G O-RAN, disaggregated into CU (Layer 3/PDCP), DU (Layer 1-2/MAC/RLC), and RU (low-PHY/RF), connected via eCPRI fronthaul.

Traditional: BBU (monolithic)

O-RAN: CU + DU + RU

Fronthaul: eCPRI (25 Gbps)

Understanding the Baseband Section

The baseband section is the computational engine of every base station. It converts the transport network's IP data into the precisely timed, coded, and modulated digital samples that the radio unit converts to RF. On the receive side, it reverses the process: decoding the received samples back into user data. The processing must happen within strict timing constraints, as 5G NR requires scheduling decisions every 0.5 ms slot (0.125 ms for URLLC).

For massive MIMO with 64 antenna ports and 100 MHz bandwidth, the raw IQ data rate between baseband and radio exceeds 60 Gbps per cell. This enormous data flow drives the architecture toward the O-RAN functional split, where low-PHY processing (FFT, beamforming) moves to the radio unit, reducing fronthaul bandwidth by 10x or more compared to the older CPRI approach.

Baseband Processing Load

Raw IQ Data Rate (CPRI):
Rate = N_ant × BW × f_s × 2(I/Q) × b_sample
64T64R, 100 MHz: 64 × 30.72M × 2 × 16
= 63 Gbps per cell

eCPRI (Split 7-2x):
Rate ≅ N_layers × 12 × N_PRB × symbols/s × bits
~5–10 Gbps per cell (10x reduction)

Processing Breakdown:
LDPC decode: 40–60% of compute
FFT/beamforming: 20–30%
Scheduling/MAC: 10–15%
Control/RRC: 5–10%

BBU Architecture Evolution

Architecture	Processing	Fronthaul	Flexibility
Monolithic BBU	All L1/L2/L3	CPRI (63+ Gbps)	Vendor-locked
CU/DU Split	CU: L3, DU: L1/L2	eCPRI (5-10 Gbps)	Multi-vendor
vRAN	COTS servers	eCPRI	Software-defined
Cloud RAN	Cloud/edge	eCPRI + xHaul	Maximum

Common Questions

Frequently Asked Questions

What does the baseband section process?

Channel coding (LDPC/Polar), modulation mapping, OFDM FFT, MIMO precoding/beamforming, resource scheduling (every 0.5 ms), HARQ, RLC segmentation, PDCP ciphering, RRC signaling. Modern BBUs: 100+ Gbps aggregate throughput.

How does O-RAN CU/DU split work?

CU: non-real-time L3/PDCP (virtualizable, centralized). DU: time-critical L1/L2/MAC/RLC (near cell site). RU: low-PHY/FFT/beamforming/RF. DU-RU via eCPRI: <100 µs latency, 25 Gbps for 64T64R mMIMO.

Processing capacity?

Single BBU card: 6 to 12 cells at 100 MHz each, 40+ Gbps throughput. LDPC decoding = 40 to 60% of compute. 64T64R at 100 MHz = 63 Gbps raw IQ. Driving transition from ASICs to GPU/FPGA in vRAN.

Related Terms

← Base Transceiver Station Baseband Transmission →

RAN Architecture

Waveguide for 5G Infrastructure

RF Essentials provides precision terminations and custom waveguide assemblies for base station RU testing, fronthaul verification, and antenna feed systems.

Request a Quote