gNB-DU

gNB-DU

gNB-DU implements RLC, MAC, and the High-PHY layers close to the radio, exposing a standard F1 interface to one or more gNB-CUs and an Open Fronthaul interface to the Radio Unit — enabling multi-vendor RAN sourced independently at each layer.

F1-C / F1-U (DU ↔ CU) Open Fronthaul (DU ↔ O-RU) E2 (DU ↔ near-RT RIC)

Key features

  • RLC / MAC / High-PHY processing with sub-millisecond scheduling
  • O-RAN Open Fronthaul (7.2x split) support for multi-vendor Radio Units
  • F1AP-based association with one or more gNB-CUs
  • Dynamic and static TDD configuration, massive MIMO beamforming control
  • Real-time RIC (near-RT RIC / E2) interface for xApp-driven scheduling
  • Cell and carrier hot-add without service-affecting restarts

Governing 3GPP & O-RAN specifications

These are the primary specifications this product implements or interoperates against. NgRAN tracks each release train and updates products as specifications evolve.

SpecificationTitle / scope
3GPP TS 38.401 NG-RAN architecture description — CU/DU functional split
3GPP TS 38.470 / 38.473 F1 general aspects & F1AP (CU ↔ DU)
3GPP TS 38.211–38.214 NR physical layer specifications
3GPP TS 38.321 Medium Access Control (MAC) protocol specification
3GPP TS 38.322 Radio Link Control (RLC) protocol specification
O-RAN.WG4.CUS.O-RAN Open Fronthaul Control, User and Synchronization plane specification

Specification titles are summarised for reference. Full texts are maintained by 3GPP (3gpp.org) and the O-RAN ALLIANCE and are not reproduced here.

Architecture & interfaces

gNB-DU hosts the layers with the tightest timing budget, so it typically runs on accelerated COTS hardware near the radio site. It exposes O-RAN Open Fronthaul south to any compliant Radio Unit and F1 north to gNB-CU, letting operators mix radios, DUs, and CUs from different vendors within one 3GPP/O-RAN compliant NG-RAN.

Reference interfaces

F1-C / F1-U (DU ↔ CU) Open Fronthaul (DU ↔ O-RU) E2 (DU ↔ near-RT RIC)

Deployment model

Bare-metal or virtualized at the cell site, with optional hardware acceleration (lookaside/inline) for High-PHY.