Fronthaul transport has been a central topic at Light Reading's 5G Transport & Networking Strategies event since 2017. So it is not surprising that fronthaul transport was, once again, a central theme during the virtual iteration of Light Reading's flagship transport event in November. Yet, 2020 marked a distinct change in tenor from definitions and theory to more practical considerations.
Heavy Reading research shows that operator interest in fronthaul is increasing and solidifying. In a global operator survey fielded in 2020, nearly half of respondents (46%) reported that fronthaul connectivity will be needed to support their functional split expectations. Splits can include the radio units (RUs) connected to a physically separated distributed unit/central unit (DU/CU) or full functional split architectures of RU+DU+CU.
Converged fronthaul for CPRI and eCPRI
One of the notable changes in this year's fronthaul discussion is the focus on today's networks. As Ericsson transport head Shane McClelland noted in his keynote presentation at 5G Transport & Networking Strategies, despite all the talk about centralized RANs, distributed RAN is the predominant mobile access architecture today. And within those networks, there are a great deal of Common Public Radio Interface (CPRI) 3, 5, 7 and 8 interfaces. Additionally, most new low band radios will continue to use CPRI interfaces. A key question is: How should operators connect the various RAN elements — both current and new and CPRI and eCPRI — without sacrificing RAN performance?
To be useful, the new packet fronthaul products coming to market must be a bit of a Swiss army knife of fronthaul functionality. Here, the industry has defined a couple of ways to handle CPRI traffic in a packet fronthaul network. Using IEEE 1904.3 Radio over Ethernet (RoE), operators can encapsulate CPRI traffic directly over Ethernet (similar to circuit emulation for TDM over Ethernet). As a fully standardized approach, RoE will work for any CPRI traffic from any radio vendor and uses standard IEEE Ethernet transport links. On the negative side, RoE offers minimal bandwidth efficiency gains, so it can be bandwidth-intensive on the transport side — particularly when there are many CPRI streams to transport to a hub site.
Another approach for handling CPRI in a packetized fronthaul network is to apply CPRI to eCPRI conversion. Because eCPRI places some of the Layer 1 processing at the radio site (before traffic hits the transport network), eCPRI transport is up to 10x more efficient than serialized CPRI. These bandwidth efficiencies apply even as CPRI is converted to eCPRI. Converting CPRI to eCPRI also prepares the carrier for future migration to cloud RAN architectures in which the eCPRI traffic is connected directly to the virtual distributed unit (vDU). On the downside and unlike RoE encapsulation, CPRI to eCPRI conversion techniques are radio vendor-specific and, thus, not standardized across different suppliers.
Fronthaul network addresses timing and synchronization
In addition to handling legacy CPRI, another important area highlighted during the 2020 event was fronthaul network timing and synchronization. In past events, Heavy Reading has talked at length about fronthaul latency restrictions — primarily, how to address the 100µs one-way latency limit imposed by hybrid automatic repeat request (HARQ) processing between radio components.
Although sometimes confused with latency, synchronization is a different issue. Synchronization refers to the time accuracy requirements that must be enforced between different radio components in the network to maintain accurate and reliable transmission. 5G introduces multiple challenges. First, functional splits and RAN virtualization create physical separation across RAN components (i.e., RU, DU and CU), which introduces the possibility for timing errors between those components. Second, 5G New Radio (NR) introduces new synchronization requirements.
Based on frequency division duplex (FDD) radios, most 4G low band networks require only frequency synchronization. New 5G mid and high band radios, however, use time division duplex (TDD), requiring synchronization across frequency, time and phase.
Standards work continues on this front, but consensus has emerged to support ITU-T Class C telecom boundary clocks in the fronthaul network. Requirements include 1,000ns absolute time accuracy between the RU and the primary reference time clock (PRTC) and 100ns relative time accuracy between adjacent RUs to meet RAN performance requirements. Furthermore, as a best practice recommended by Ericsson and others, operators are encouraged to have both a primary and backup timing source, such as coupling a satellite timing source with network-based timing source, in the event that the primary becomes temporarily unavailable.
Now is the time for fronthaul infrastructure
While industry progress in fronthaul is encouraging, Heavy Reading notes there is an increasing sense of urgency to build out these networks. In a global Heavy Reading operator survey conducted earlier this year, 62% of operator respondents reported plans to launch mass-market 5G services in the 2021-23 timeframe. The time to build out the supporting fronthaul infrastructure is now.
Looking for additional information?
For more information, you can access the 5G Transport & Network Strategies replay here: www.lightreading.com/webinar.asp?webinar_id=1742.
— Sterling Perrin, Senior Principal Analyst – Optical Networking & Transport, Heavy Reading
This blog is sponsored by Ericsson.