Dr. Christian Lanzani, Sr. Wireless Product Marketing Manager
Perminder Tumber, Wireless Product Marketing Manager
Gareth Edwards, Principle Engineer
Thomas G. Noergaard, Founder
Niklas Meyer Mortensen, Engineer
Rami Al-Obaidi, Engineer
In a connected world, mobile networks – today primarily driven by smart phones - will have to evolve and support a largely increased number other devices and services overall contributing to the exponential traffic growth toward 5G. To address data demand in a business sustainable way, the wireless industry is challenged by re-thinking base station architectures. Centralized and Virtualized RAN architectures will be the base for the mobile networks of the future sustaining more capacity as well as network flexibility and scalability with a more generic hardware footprint. In particular, the connection between baseband and radio elements – known as Fronthaul – will be highly impacted by the partitioning of choice between different radio processing functions. CPRI interface has been the de facto standard for primitive and mainly point-to-point fronthaul transport for macro base stations, but its bandwidth and flexibility limitations are driving a re-thinking on this critical connectivity interface. Massive MIMO, Carrier Aggregation, Multi-Band support, radio cell densification are impacting the bandwidth requirements, while the coexistence between macro, micro, pico and small cells along with a centralized / virtualized processing environment are impacting the flexibility requirements.
To solve the Fronthaul challenge, a number of activities recently emerged in the IEEE standardization consortium and gained significant interest from all of the key industry players. The aim of such initiatives is to define an Ethernet-based fronthaul network that is addressing the capacity and flexibility limitations along with timing / synchronization challenges of radio networks.
This paper will help to navigate through the key concepts of packet based Fronthaul and discuss the implications of the adoption of latest Time-Sensitive-Network (TSN) IEEE 802.1CM standard, the IEEE P1904.3 Radio over Ethernet (RoE) standard and latest Next Generation Fronthaul Interface, (NGFI) IEEE P1914.1 initiatives. Finally, a look at the PoC platforms enabled by Xilinx technology and IP offering from Comcores will be offered.
Cloud-RAN (C-RAN) is an emerging concept in evolving wireless Radio Access Networks (RAN), which has the potential to more efficiently utilize base station processing resources in the RAN, and thereby reducing associated CAPEX and OPEX in the network deployment and operations. C-RAN does this by physically centralizing Baseband processing in a Data Center-like environment to exploit a multiplexing gain of processing power while distributing the radio access modules (or remote radio heads) for a scalable, real-time network. Mobile network operators and vendors claim, that C-RAN will introduce power savings. In their CRAN white paper , CMRI has shown CPEX and OPEX can be reduced by 30% and 53% respectively for new CRAN sites. However, the separation of BBU and RRH introduces the Fronthaul network and sets tight requirements for the underlying transport. Mainly low jitter and latency, and increasingly high bandwidth in the race to 5G along with network flexibility and scalability. An efficient Fronthaul interface is critical to support key technologies for both 4.5G and 5G such as C-RAN, massive MIMO, UDN. Fronthaul has been a concern to CRAN large-scale realization supporting LTE-Advanced carrier-aggregation and massive MIMO radio configurations.
Common Public Radio Interface (CPRI) is the de facto, but semi-proprietary interface protocol used in 4G networks, to transfer modulated radio signals at a constant rate irrespectively of the active user traffic. CPRI was intended to connect the RRH and the BBU supporting distances up to 20 Km using a bi-directional point-to-point topology over fiber. CPRI carries the synchronization, I/Q antenna data and OAM information in a TDM fashion between BBUs and RRHs. The dedicated fibers make the Fronthaul portion of a large C-RAN deployment relatively costly, with notably shortcoming being an inflexible networking, low transmission efficiency and limitations in OAM& protection capabilities.
Therefore other options are being considered in the industry for implementing more efficient and flexible transport in mobile Fronthaul networks toward 5G. Reusing existing packet switched networks is one of the concrete possibilities considered addressing cost, flexibility and scalability of the network.
Ethernet technology has demonstrated steady, cost efficient speed and capacity growth driven by the enterprise connectivity, access, and data-center markets. As a result of that, standardization activities held by the IEEE P1904.3 Radio over Ethernet (RoE) aimed at taking advantage of the Ethernet developments and specify a scalable and streamlined packet based radio interface solution that complements, for example, the existing CPRI radio transport specification based on fixed time division-multiplexing.
The further re-thinking of Fronthaul has a long term goal to enable a traffic adaptive interface with support for statistical multiplexing and small cell coordination while being antenna and radio interface neutral. The key to achieve Fronthaul interface redesign lies in the function re-split between BBU and RRU and the new Next Generation Fronthaul Interface (NGFI) standardization effort as proposed by the new IEEE P1914.1 Working Group will further lead to re-design of underlying transport networks with packet switching capability.
Packet switched networks provide widespread deployment and an ability to carry numerous services. In comparison to a traditional CPRI over fiber connection, significant savings in CAPEX and OPEX can be achieved by either leveraging CPRI encapsulation over Ethernet or by revisiting the Layer-1 partitioning between baseband and RRHs modules traffic adaptive bit rates, which will be under study in the IEEE 1914.1, Next Generation Fronthaul Interface. This is done while introducing packet switching flexibility and scalability. Enabling CPRI over Ethernet will naturally require an additional framer/packetizer function which can encapsulate the CPRI frame in an appropriate manner together with e.g. sequencing. Such a procedure is under development by the IEEE 1904 Access Networks Working Group (ANWG) with its IEEE P1904.3 standard for Radio over Ethernet Encapsulations and Mappings.
Nevertheless, operators are faced with a new challenge as they look for ways to move away from CPRI and find alternatives that can handle the ever increasing bandwidth requirements for the evolving RAN. The packetized networks are non-deterministic and lack precise time synchronization, which is inherent and required for CPRI. Highly accurate clock distribution, and Time Sensitive Networking (TSN) services have to be implemented, in order to fulfill the strict requirements of the Future mobile Fronthaul network architecture.
Even though timing and synchronization enhancements such as 1588v2 PTP have been developed for Ethernet, additional improvements may be necessary to satisfy the requirements of Radio over Ethernet. In IEEE802.1, the Time-Sensitive Network Task Group is currently working on a set of standards aiming to allow time-synchronized low latency high bandwidth services over Ethernet networks. Projects such as IEEE 802.1CM Time-Sensitive Networking for Fronthaul are in progress, as an extension to the TSNTG, and is looking into TSN used in the mobile Fronthaul networks.
To summarize, a packet Fronthaul for future mobile networks is made possible through a set of timing and synchronization standards by TSNTG and a Radio over Ethernet (RoE) standard by ANWG.
Overview of TSN
Figure 1 illustrates the concept of using a packet switched network as the underlying transport network for the Fronthaul in a C-RAN deployment. As shown the packet switched network is used not only as a Fronthaul, but also as a metropolitan area network (MAN) for third party services such as subscriber Internet services.
Figure 1 – Conceptual diagram for a packet switched network for Fronthaul
When dimensioning the Fronthaul in a C-RAN network, strict requirements are defined between the RRH and the BBU. These requirements have to be accommodated to service Radio over any Fronthaul network. The requirement are presented in Table 1. Simulations of TSN functions have been performed  along with simulations of combined TSN features in a Fronthaul network , with promising results of latency and jitter reduction.
|Requirement ||Maximum value ||Comment |
|Line Rate ||614.4 to 24330.24 Mbps ||CPRI line rates |
|Latency ||100µs ||One way transmission |
|Frame Delay Variation ||5µs or 10% of latency || |
|Frame Loss Ratio ||10-6 - 10-9 || |
Table 1: I/Q data requirements of Time-Sensitive Networking for Fronthaul [802.1CM draft 0.1]
The challenge is being able to service deterministic Radio data streams between RRHs and the BBU pool without being adversely affected by unrelated background traffic. To accomplish this the network must be able to orchestrate data streams at a frame level resolution. Without doing so, background traffic may cause extensive latency and jitter for the Radio streams impacting the overall system performance. The network must therefore support strict network wide cooperative scheduling procedures. Such procedures are offered by TSN with Frame Preemption, Enhancements for Scheduled Traffic and Stream Reservation Protocol. However, TSN consists of 7 active projects of which all are listed below:
- 802.1AS - Timing and Synchronization for Time-Sensitive Applications
- 802.1Qbu – Frame Preemption
- 802.1Qbv – Enhancements for Scheduled Traffic
- 802.1Qcc – Stream Reservation Protocol (SRP) Enhancements and Performance Improvements
- 802.1CB – Frame Replication and Elimination for Reliability
- 802.1Qch – Cyclic Queuing and Forwarding
- 802.1Qci – Per-Stream Filtering and Policing
Of these 802.1Qbu, 802.1Qbv, 802.1Qcc and RoE will be briefly presented below as we see them as potentially being useful or necessary as components of a TSN-enabled Ethernet Fronthaul network, with the overall solution dependent on network packet jitter performance and buffering requirements at the endpoints.
802.1Qbu - Frame Preemption
Frame preemption is the suspension of a preemptable frame during transmission, to transmit one or more express frames. When all express frames are transmitted, the transmission of the preemptable frame is resumed. A preemptable frame can therefore be broken into two or more fragments and will be reassembled at the end receiver of the Ethernet frame.
Figure 2: Pre-emption concept
Figure 2 illustrates frame preemption, where a TSN enabled switch receives multiple frames with different priorities. Transmission of the preemptable frame is interrupted by arrival of express frames, and resumed when all express frames have been transmitted.
Preemption of an ongoing transmission can only occur if the final fragment is larger than 64 octets, and the non-final fragment is larger than the minimum non-final fragment size. The minimum non-final fragment size is can be set to 64, 128, 192 or 256 octets, and 60 octets mCRC are added to each non-final fragment. The worst case delay for preempting a single frame is 123 octet times, and occurs if a preemptable frame has just started transmission.
802.1Qbv - Enhancements for Scheduled Traffic
Nodes in a TSN network can support Enhancements for Scheduled Traffic, which can schedule time frames for transmission of certain priorities. To achieve this, transmissions gates are associated with each priority. The gate for each priority can either be closed or open, in each timeslot. Frames from a traffic class can only be transmitted if the associated gate is open.
Figure 3 - Gate Control List
Each port in a TSN switch contain a Gate Control List, which list the states for each priority class queue at each time slot. An example of this is shown in figure 3. In switches not supporting gating, all gates would be perceived as constantly open. The Gate Control List is a cyclic list and is repeated after each cycle.
A packet cannot be sent if the gate is opened for a shorter time period than the transmission time of the frame. This will reduce the efficiency if frequent state transitions occur. It is assumed that the switches has detailed knowledge of the transmission time for frames. A precise timing synchronization between involved bridges is need to schedule gates along a data stream.
802.1Qcc - Stream Reservation Protocol
802.1Qcc specifies how to enable end-to-end reservation of resources for data streams in a TSN network. This is a critical feature in a TSN network, since the time-sensitive applications needs a guarantee of a QoS, to meet their requirements.
SRP is enabled by three signaling protocols MMRP, MVRP and MSRP, which are able to reserve streams across a packet network. MMRP manages registration of end nodes. MVRP declares VLAN memberships to all involved entities in a stream. MSRP allows reservations of streams in the network based on requirements for the stream and capabilities of the involved network nodes. The network nodes will allocate resources for the data streams.
Fulfilling the requirements of latency, jitter, and bandwidth of streams in a TSN network requires a clear definition of resources and a distribution of resources across the network. When the requirements of a stream cannot be accommodated the stream will not be setup. However, the ability to prioritize streams, will reduce the likelihood of important streams to be dropped.
TSN will play a crucial role in enabling a packet switched mobile Fronthaul. The biggest challenge of a packet switched Fronthaul is the strict requirements of the CPRI data streams from the BBUs to the RRHs. An illustration of a simple deployment scenario is shown in figure 4.
Figure 4 - Traffic streams over TSN networks
The red and green streams represent CPRI data from the BBUs to the eNodeBs. The blue stream is a logical X2 interface between the BBUs which is used for signaling and data transmission. The illustration shows the challenge of supporting other services simultaneously, as this traffic can cause latency and frame delay variation for the CPRI streams.
Radio over Ethernet
While TSN is designed to ensure timely reception of time sensitive packetized streams such as CPRI over Ethernet it does not deal with the encapsulation of CPRI and similar RRH-BBU communication. This is instead being developed by the IEEE 1904 Access Networks Working Group (ANWG), with the IEEE P1904.3 Task Force working on a draft standard for Radio over Ethernet Encapsulations and Mappings. The work targets among others definition of a native Radio over Ethernet (RoE) encapsulation and mappers transport format for both digitized radio payload (IQ data) and management and control data. The Ethernet packet format itself is not changed and neither is the MAC under this Project Authorization Request (PAR), as showing in Figure 5 .
Figure 5 – RoE encapsulation overview – Ethernet Packet remains unchanged
There is also focus on enabling support for alien radio data where the transport structure is simply a container for the data.
Notably, the P1904.3 Task Force is considering implementing a structure aware mapper for CPRI with a clearly defined encapsulation procedure as shown in Figure 6 . One can therefore differ between a “simple tunneling” mapper for the former use case as shown in Figure 7 , and structure aware mapper for CPRI. There is also a “structure agnostic” mapper in the draft which represents a middle ground of efficiency and complexity between the two other mappers.
Figure 6 – RoE encapsulation: RoE Structure Aware Mapper
Figure 7– RoE encapsulation: Simple Tunneling Mapper
Several use cases of RoE can be considered, such as aggregation of several CPRI streams from a number of RRH to a single RoE link to the BBU pool, or a native edge-to-edge RoE connection from the RRH directly to the BBU pool. RoE will naturally add a new switching/aggregation node between the baseband pool and the radio resources.
Next Generation Fronthaul Interface (NGFI)
While Radio over Ethernet is for encapsulation of CPRI and similar RRH-BBU communication, it does not address the shortcoming of CPRI, namely low-efficiency, inflexibility and poor scalability. To support key technologies for both 4.5G and 5G, a rethink is required. The new technology must support traffic dependent configuration, support statistical multiplexing, support cell coordination, be antenna independent and have radio interface technological neutrality. NGFI is claiming to be that technology and the key lies in the function re-split between the BBU and the RRU with the underlined transport network based on packet switching as shown in figure 8 
Figure 8– C-RAN Radio Network Architecture based on NGFI
The proposed NGFI is targeted as an open interface that will work to address the short coming of current Fronthaul interface by redefining the functional split between the BBUs and RRUs. Some of the processing functions are shifted to the RRUs. Such a change allows the Fronthaul interface to move from a point-to-point connection to a multipoint-to-multipoint enabled Fronthaul and allows for advance features such as bandwidth adaptation in response to dynamic payload variation, maximum support for high-gain coordinated algorithms, at the same time allows for decoupling of interface traffic from the number of antennas. This way NGFI is capable of supporting current air interfaces as well as future ones based on 5G technologies.
The newly formed NGFI 1914.1 Working Group under IEEE will work to standardize the Fronthaul interfaces and address the inherent challenges, with the possibility of the 1904.3 Task Force coming under the NGFI WG. The scope of this NGFI working group is to define the Architecture for the transport of mobile Fronthaul traffic, including user data and management and control traffic. IEEE 1914.1 WG will also recommend different functional partitioning between the RRUs and BBU, with some of the baseband processing moving out of the BBU.
Proof of Concept Designs
As covered, a number of standardization efforts are underway under the supervision of the IEEE and elsewhere, such as the P1904.3 Task Force on Radio-over-Ethernet, the IEEE1914.1 Working Group on NGFI, and the underlying network transport being looked at in 802.1CM for improvements. Standardization efforts always take several years to finalize and be fully adopted. In the meantime, a number of Proof of Concept efforts are on-going to validate the benefits and performances. Xilinx offers a variety of Proof of Concept designs for an Ethernet-to-CPRI Gateway allowing experimentation with the challenges of NGFI, as well as an IQ switch platform developed by Comcores leveraging Xilinx silicon technology and the collective expertise on Fronthaul. Both platform are immediately available to test and further refine/optimize Fronthaul designs enabling the next generation C-RAN connectivity.
As data demand continue to accelerate on the mobile infrastructure, the technical challenges with Fronthaul becomes more severe, leading to several industry standardizations taskforces and working groups to come up with innovative, flexible and cost efficient solution. TSN, Radio over Ethernet, and NGFI are complementary industry efforts aiming at addressing the technical challenges and lowering the deployment cost of Fronthaul networks by establishing or reusing existing packet based network infrastructure. Significant challenges in ensuring deterministic traffic scheduling, packet encapsulation and radio level synchronization must be overcome for these efforts to succeed. Xilinx and Comcores are actively driving these standardization efforts by enabling advanced proof of concepts available for designers to further enhance, innovate, and build on their next generation of RAN solutions.
 CRAN_white_paper_v2_5_EN (CMRI)
 Radio over Ethernet Motivation, scope, timeline Jouni Korhonen (Broadcom) 17-Oct-2014
 NGFI-Whitepaper_EN_v1.0_201509291 CMRI Oct 4 2015
 Wen-Kang Jia, Gen-Hen Liu, and Yaw-Chung Chen. Performance evaluation of ieee 802.1qbu: Experimental and simulation results. In Local Computer Networks (LCN), 2013 IEEE 38th Conference on, pages 659–662, Oct 2013.
 S. Rumpf, T. Steinbach, F. Korf, and T.C. Schmidt. Software stacks for mixed critical applications: Consolidating ieee 802.1 avb and time-triggered ethernet in next-generation automotive electronics. In Consumer Electronics Berlin (ICCE-Berlin), 2014 IEEE Fourth International Conference on, pages 14–18
 De switch akro 6/0 tsn. https://www.tttech.com/products/industrial/switches/iot-switches/de-switch-akro-60-tsn/. Accessed: December 10 2015.
Xilinx Silicon and tools portfolio contain the main ingredients required to address network architectural challenges, including Fronthaul. Xilinx FPGAs are fundamentally programmable components that can be configured to deliver the high-performance and low-latency processing required for architectural proof of concepts and productized Fronthaul designs, to succeed.
Xilinx FPGAs are programmable both at-the-factory and in the field, allowing for innovative advance technology not only to be tested and optimized but allows for a very flexible and future proof solution that can be customized for any particular scenario at any time.
As with previous product families, the recently announced 16 nm UltraScale+ family further add value by even a great degree on performance and power saving over existing solutions including support for 25G connectivity. With every node Xilinx continue to push the performance per watt envelope while offering total BOM cost reductions.
Comcores is an expert in wireless communication systems and has a portfolio of digital IP’s containing key building blocks for communication systems. These range from the radio system to the network. A pioneering effort and experience with fronthaul designs and the offering of a wide range of digital building blocks that enable end-to-end solutions makes Comcores the ideal partner to address new design challenges in the wireless space if you want to succeed.
Hardware demonstrators enables early phase testing and proof-of-concept of key building blocks for fronthaul. The building blocks are realized in VHDL and are delivered with robust documentation and verification methodologies leveraging Xilinx technology.