Prototyping of Next Generation Fronthaul Interfaces (NGFI) using OpenAirInterface

Prototyping of Next Generation Fronthaul Interfaces (NGFI) using OpenAirInterface

Raymond Knopp, Navid Nikaein, Christian Bonnet, Florian Kaltenberger, Adlen Ksentini, Rohit Gupta




There has been lot of R&D for C-RAN (Centralized RAN) deployments with centralized processing in Base-band units (BBU) and Remote Radio Unit (RRU) using CPRI. However, CPRI requires expensive fronthaul like optical fiber to carry RF samples from BBU to RRU due to high bandwidth requirements. This has limited deployments of C-RAN around the world. However, 5G poses additional challenges due to massive bandwidth requirements, low latency, etc and this requires re-thinking of C-RAN architecture for 5G. In this whitepaper, we provide an overview of industry developments around next generation C-RAN architecture, Next Generation fronthaul interface (NGFI) and how it can meet requirements of 5G and can be built using alternate transport technologies like Ethernet without using CPRI. The whitepaper also gives an overview of standardization activities being carried out within the industry around IEEE NGFI and splitting radio stack between BBU/RRU and different tradeoffs associated with it. We also give an overview of how this work is carried within OpenAirInterface Software Alliance (OSA) and community led-development with our member organizations towards building end-to-end open source cellular C-RAN architecture for 5G.


C-RAN offers several advantages in terms of network deployment, reduced operating costs, improved network performance only to name a few as highlighted by China Mobile Whitepaper on NGFI [1]. However, the traditional C-RAN architecture has relied on CPRI/OBSAI [5] [6] to carry transport and synchronization information from BBU to RRU. The current generation C-RAN architecture carries samples from BBU to RRU and this imposes very high bandwidth requirements on the fronthaul transport network. Hence, current generation CPRI deployments will struggle to meet the scalability and performance requirements for 5G. Future 5G networks will provision much higher radio interface bandwidth than LTE and this results in exponential bandwidth increase for conventional CPRI based RRU deployments. Future NGFI architecture will also allow building cost-effective network massive mimo systems by splitting radio functions between BBU/RRU

There is industry consensus within 3GPP/NGMN/IEEE SDOs to re-think existing cloud-RAN architecture and evolve it towards the needs of 5G by splitting different parts of radio stack between different network elements (BBU, RRU). In that spirit, recently IEEE has formed NGFI working group [2] to standardize transport fronthaul interface for future cellular networks. We show in this whitepaper some of the initial work being carried out within OpenAirInterface Software Alliance [5] for prototyping next generation fronthaul interface and carry out real deployment studies with commercial phones.

NGFI Architecture:

NGFI is the fronthaul interface between baseband pool and remote radio heads for the next generation of radio network infrastructure. In traditional C-RAN architecture, all the baseband processing is carried out at BBU which sends I/Q samples to RRU via CPRI fronthaul interface. NGFI redefines the baseband processing split between BBU and RRU, hence redefining the positioning of eNB stack components between BBU/RRU. According to NGFI terminology introduced by China Mobile [1], BBU is redefined as Radio Cloud Center (RCC) and RRU becomes Radio Remote System (RRS). NGFI architecture from China mobile envisions point-to-multipoint architecture from RCC-RRU, hence there is another element Radio Aggregation Unit (RAU) which interfaces with RCC and carries transport for several RRU. Fig .1 shows the C-RAN architecture based on NGFI [1].


Source: China Mobile – networking03.pdf

Fig. 1: Next Generation Fronthaul Interface (NGFI) Architecture

NGFI Whitepaper [1] describes several ways in which RRU/RAU/RCC can be split as described in Fig. 2 (a). Fig. 2(b) shows the maximum interface bandwidth, whereas Fig. 2(c) shows the latency requirements for each of these splits [1]

Fig. 2(a): Different splits of RRU/RCC

Fig. 2(c): Interface Delay

Fig. 2(b): Maximum Fronthaul Interface Bandwidth

Source: NGFI Whitepaper, China Mobile [1]

Fig. 2: Different splits of RRU/RAU/RCC [1]

OpenAirInterface Architecture for NGFI

Traditional base-station IP was designed as monolithic architecture meant to run within single equipment. However, due to emerging NGFI requirements, it is imperative that eNodeB design has to be modular architecture capable of running in different parts of network with different layers of the stack communicating over standard IP/Ethernet interfaces. This allows dynamic placement of baseband functions within RCC/RAU/RRU thus offering maximum flexibility in designing future networks. It is in that spirit, we have redesigned OpenAirInterface so it can be easily split amongst different parts of the network.

We have currently extended OpenAirInterface (OAI) to support the following node functionalities.

  • 3GPP eNodeB
  • 3GPP eNodeB BBU [NGFI IF5]

IF4p5 refers to the split-point at the input (TX) and output (RX) of the OFDM symbol generator (i.e. frequency-domain signals). According to NGFI, IF4 is “Resource mapping and IFFT” and “FFT and Resource de-mapping”. We currently do not try to exploit multiplexing gains for unused spectral components. So, IF4p5 is simply compressed transmitted or received resource elements in the usable channel band. It should be noted that the OAI software architecture for NGFI will be extended to implement RAU, which can host both conventional CPRI-based RRH and also RRUs based on NGFI specs.



Fig. 3: Functional splits already implemented in OAI

Fig. 4. shows the deployment architecture that we are currently implementing at OpenAir5GLab@EURECOM [7]. The cloud/BBU pool runs on 20 Core Dell T620 server. The EPC runs in another server connected with cloud BBU pool, but EPC can also run in virtual environment on same server running BBU pool. This server is connected to Gigabit Optical Switch (Cisco Catalyst 2960-X). The switch connects to different remote radio units (RRUs) via Gigabit Ethernet. We are using low power Commell LP-170 (Intel Quad-Core Atom) as RRU connected to USRP B200 via USB-3. The RRU is powered over Ethernet via Cisco Switch, which also distributes synchronization information via IEEE 1588 protocol or via some other mechanism developed by the industry. There are also other low-cost Intel/ARM SoC boards and they can be interfaced with other supported RF platforms, for ex. (LimeSDR, BladeRF and EXMIMO) [3]. We are currently testing the deployment with COTS phone, for example (Sony Experia M4, Samsung Galaxy SV, Huawei E398, Bandrich C501, etc). We are also extensively testing this platform with Ercom Mobipass Test UE capable of simulating multiple UEs and interfacing next generation test UE platform for C-RAN testing. We also plan to integrate OAI UE [10], which runs on same generic Intel/ARM platform for testing purposes. It should also be noted that we are developing RAU within OpenAirInterface capable of interfacing with both NGFI RRUs and also conventional CPRI based Alcatel-Lucent Radio heads. Recently, IT-Aveiro, one of our community members extended OpenAirInterface for interfacing with CPRI based remote radio head which was developed within the context of research project [8]. Such a community led-development aims at creating interesting playground for exploring 5G architectures and performance evaluation for NGFI in conjunction with existing CPRI based commercial remote radio heads.

Fig. 4: Cloud-RAN Deployment Architecture (OpenAir5GLab@EURECOM)

Fig. 4: Cloud-RAN Deployment Architecture (OpenAir5GLab@EURECOM)

We are also collaborating closely with several other open source communities, for ex. (OpenStack, OpNFV, ONOS, OSM [11][12][13][14]) on remote orchestration of this network within virtual environments, such as KVM or containerize deployments, such as Docker, LXC. OpenAirInterface provides a complete open source End-to-end solution comprising both RAN and EPC which can be used as a use case for vEPC/vRAN/vRRH deployment and interesting playground for 5G orchestration and slicing with the final goal of creating 3GPP/IEEE/NGMN Reference implementation platform for prototyping and research. We are currently collaborating with ITU FG IMT 2020 [9] [16], IEEE NGFI [2], ETSI [15] and several other standard organizations to develop open source prototyping solution to influence standardization and vice versa.


For the purposes of the demo, we have scaled down the setup. We have two Gigabyte Brix; GB-BXi7-4770R, where one runs OpenAirInterface (OAI) EPC and another runs OAI RCC. OAI eNodeB is connected via 1 Gb Ethernet to Intel Atom Quad-Core SBC, Commell LP 170 which acts as RRU as shown in Fig. 5. We have used USRP B200 mini as SDR radio front-end for the purposes of the demo. For the purposes of the demo, we show only one RRU, but it can be easily scaled to distributed RRUs with synchronization fed through IEEE 1588 protocol over Gigabit Ethernet via standard Cisco switches. The software integrates seamlessly with other SDR platforms such as BladeRF, EXMIMO, LimeSDR and various other RF platforms. All the software for the demo can be downloaded from OpenAirInterface software repository [3].

Fig. 5: Demo deployment of OAI RCC/RRU + COTS Smartphone

Fig. 5: Demo deployment of OAI RCC/RRU + COTS Smartphone


[2] IEEE NGFI P1914,
[3] OpenAirInterface Software Alliance,
[4] Ercom Mobipass,
[5] CPRI,
[6] OBSAI,
[7] OpenAir5GLab@EURECOM,
[8] FlexiCell- OAI Based Cloud Radio Access Networks,
[9] ITU-T FG – IMT 2020,
[10] OAI UE, Alliance Project 1B (Basic Functionality),
[11] OpNFV,
[12] OpenStack,
[14] OSM,
[15] ETSI,
[16] ITU FG IMT2020 F2F Meeting,