5G needs tighter timing requirements than do 4G networks. The timing must perpetuate from the radio throughout the telecom network core.

5G New Radio (NR) networks pose a variety of engineering challenges. They bring significant changes to every part of the network, from the core clocking function to the Radio Unit (RU) air interface. System designers will need to engineer 5G NR units to meet new timing and cost requirements. That includes re-engineering 5G advanced network and radio services, synchronization architectures, and both fronthaul and core transport. Meeting these and other challenges requires a new set of best practices for selecting RU components, and a thorough understanding of how these decisions will affect the entire network.

Engineering the 5G radio unit
Mobile networks depend on synchronization between radios. Specifically, the time alignment error (TAE) between different frequencies at the transceiver array boundary (TAB) determines the synchronization, transport engineering, and components required for adjacent radios to connect to user equipment (UE) and operate without co-channel interference. This applies to both frequency division duplex (FDD) and time division duplex (TDD). NR will primarily operate on the latter, which is a new operating mode for most network operators. TDDs will imply re-engineering the timing network to meet the 5G requirements at the RU.

NR has stringent TAE engineering requirements that carried over from LTE-Advanced (LTE-A), creating additional challenges for RU engineers. NR introduces, for example, a new power-management schema. LTE Evolved Node B (eNB) remains active in idle state, with continual transmission of idle mode signals such as Synchronization Signal Block (SSB) and Cell-Specific Reference (CSR). The longer idle periods in NR reduce the NR network’s heat, power consumption and UE paging while improving overall performance with lower energy consumption than LTE [Ref. 1].

5G RU design engineers, therefore, need components with fast start-up, high-frequency on/off cycles, and high MTBF that integrate hardened Digital Front End (DFE) application-specific blocks for maximum power saving and high performance-per-watt. These components also need to scale from small cells to macrocells while delivering carrier aggregation and multi-band 400 MHz, multimode, and instantaneous bandwidth allocation over Frequency Range 1 (FR1) and FR2. In addition, they must support existing/emerging Gallium Nitride (GaN) power amplifiers (PAs) to future-proof the RU. mmWave (FR2) has well-known power challenges aggregating Multi-User-Multiple Input, Multiple Output (MU-MIMO) interfaces while managing beamforming, etc.

Maintaining RF timing stable enough to meet the network time error (TE) requirement also impacts the on-board oscillator used to provide timing to the radio’s transimpedance amplifier (TIA) and PA DFE blocks. This timing must ensure a stable TAE at the RU. Traditionally, oscillators on the Baseband Unit (BBU) ensured clock holdover should the radio lose its timing signal. This is no longer feasible for two reasons. First, the BBU is no longer a timing interface or demarcation point in 5G networks because it’s disaggregated into the Centralized Unit (CU) and Distributed Unit (DU) functional blocks in an Open RAN architecture. Second, oscillator choice is also constrained by cost, heat, and power considerations, which is why Temperature Compensated Crystal Oscillator (TCXO) or Micro-Electro-Mechanical Systems (MEMS) technology are replacing high-performance oscillators built with other technologies.

When choosing oscillators, you should understand that 5G fronthaul transport cannot use Common Public Radio Interface/Open Base Station Architecture Initiative (CPRI/OBSAI) over fiber with proprietary timing and high-frequency timing pulses. Instead, 3GPP mandates IEEE 1588 Precision Time Protocol (PTP) over Ethernet (Figure 1). PTP has implications for the behavior of the selected oscillators. Low-cost MEMS oscillators introduce severe constraints, react poorly to physical-layer rearrangement, and typically cannot sustain the bandwidth used in PTP G.8275.2 profile [Ref. 2]. The result: they must be engineered with the lower-bandwidth G.8275.1 PTP profile on the fronthaul network. This has a concomitant impact on engineering both the fronthaul (DU to RU) and the backhaul network.

Fronthaul engineering
In 4G LTE, CPRI fronthaul transport impacted the network’s capacity, synchronization, and cost. NR introduces the new Enhanced CPRI (eCPRI) over Ethernet fronthaul, which must be engineered such that time-sensitive radio control services operate effectively. 5G NR requires adjacent radios to adhere to both absolute and relative TE specifications [Ref. 3], which implies either Primary Reference Time Clock (PRTC), ITU standard G.8272, or Telecom Boundary Clock (T-BC) within 260 (nsec) of the RU. The removal of protocol constraints in NR compared to LTE, combined with moving timing to Ethernet, has added considerable flexibility to fronthaul engineering, with some caveats. The lack of operationally viable clock holdover on the BBU or RU requires the use of a high-performance T-BC on the fronthaul network’s switches. Component choices of RU design engineers drive the timing architecture on the 5G fronthaul network. Table 1, from the Telecom Boundary Clock ITU standard G.8273.2, shows the maximum cTE allocated to the various BC classifications.

As we have seen, the combination of tight TAE at the RF interface, the use of zero-holdover oscillators and the mandatory use of G.8275.1 for synchronization impact fronthaul engineering considerations. As in LTE (TS.36.104), 5G NR mandates the use of ±1.5 µsec absolute TE at the air interface (TS 38.104). To let operators meet these TE requirements on the transport network, the T-BC (G.8273.2) recommendation has changed four times, taking the industry from Class A — constant TE (cTE) ±50 nsec to Class D cTE @ ±5 nsec. These rapid changes have forced T-BC component redesign in core switches and a succession of onerous network upgrades for mobile operators. In other words, the increasingly tight TE requirements at the edge of the network, along with the way the RU is being engineered, continues to impact the overall switch/router fabric of the transport network.

Small cells and the return of G.8275.2
While the development of power-efficient high-bandwidth small RUs has created challenging network upgrades, it also facilitates highly distributed small cell Open RAN service architectures. These architectures will enable small cell-based 3GPP Release 17 applications such as integrated access and backhaul (IAB), cellular vehicle-to-everything (C-V2X), IoT, and new private networks with local PRTC (source clocks). These applications use 5G Core (5GC) or 5G-U unlicensed or leased spectrum, operator network slicing, and the UK Joint Operator Technical Specification (JOTS), [Ref. 4, 5] for a neutral host “gateway.” Small cell infrastructures of this type will rapidly replace proprietary distributed antenna system (DAS) which can’t compete with the TDD based sub 6 GHz C-band, or mmWave RU. 5G small cells will also allow deployment of IAB systems. IAB can be fixed or ephemeral, line of sight or meshed, and use FR1 or FR2 for both mobile termination (MT) and backhaul with configurable radio clusters.

Moreover, small cells will have a more flexible deployment profile than Option 7.2 (Open RAN macro cells), which is tied to G.8275.1. Option 6/Option 8 small cells with the DU integrated in the RU with the TAA and DFE can more easily support both G.8275.1 and/or G.8275.2 (Figure 2). The latter profile will be crucial in a non-engineered environment such as an existing LAN where the switches lack T-BC.

Simultaneous with the changes in the radio network to meet phase requirements, planners and synchronization engineers have been deploying high-availability clocks in the core network, replacing time-division multiplexing (TDM)-based primary reference clocks (PRC) and synchronization supply units (SSU) with Ethernet-based clocking functions including Enhanced PRC (ePRC)/G.811.1, ePRTC /G.8272.1 @ ±30 nsec maximum absolute time error (maxTE), and PRTC-A or PRTC-B/G.8272 @±100 nsec or ± 40 nsec maxTE respectively, to provide PTP and/or Synchronous Ethernet (syncE) timing. These functions must be comprehensively engineered into the core transport and timing networks.

Engineering the core transport
Two issues have forced a rethink about the most effective core transport layer for PTP timing:

• The need to avoid being dependent on Global Navigation Satellite System (GNSS) satellite-based timing by meshing ePRTC to create “GNSS failure-resilient” networks using land-based time transfer.
• The increasing need for stable high-performance extremely tight timing to the 5G RU. Until recently, Ethernet G.8275.1 (On Path Support) networks dominated timing deployments.

Operators, however, now deploy PTP engineered on the optical layer. Carrying PTP on the lambda or optical timing channel using boundary clocks designed for deployment with dense wavelength division multiplexing (DWDM)/coarse wavelength division multiplexing CWDM systems has brings extremely low TE of less than ±3 nsec (better than G.8272.3 Class D) and extremely high stability. With this implementation, a network can be engineered to PRTC-A (= ±100 nsec) at all service points, also referred to as a “virtual PRTC.” vPRTC enables synchronization engineers to push ±100 nsec TE to the edge of the network, nearer to the DU where fronthaul begins.

Such low TE in the core, coupled with equally low TE on the engineered fronthaul, provides the network timing and planning tools needed for significantly greater elasticity in engineering the synchronization network in the access/midhaul distribution network. Consider this simple calculation for a vPRTC with a fast fronthaul and a metropolitan area Ethernet network (Metro E) network between the vPRTC dropoff and the DU pool (Figure 3).

The vPRTC could span several hundred kilometers over ten or more DWDM hops with less than ±100 nsec TE to the aggregation router. The metroE may have any Class T-BC. As both vPRTC and fronthaul have low TE, there is a huge timing budget available to appropriately engineer the access network (±1.24 µsec). This gives the network engineer tremendous flexibility in how to engineer the end-to-end network.

Conclusion
The logical and geo footprints of core LTE networks will stay relatively stable as operators migrate to 5G, but subtle engineering challenges and the solutions applied in deploying 5G RUs will have repercussions throughout the network. The ability to reduce cost and form factor in the 5G RU while maintaining TAE will require proper component selection which, in turn, will impact the fronthaul network and core network transport architectures. There also will be a corresponding evolution of synchronization engineering with the development of high-performance source clocks and optical-layer boundary clocks for the next-generation transport network. In short, there will be growing reliance on meeting timing requirements at the RU and the way this is accomplished will have a significant impact on how the entire mobile network is engineered.

References
1. Pål Frenger and Richard Tano, More capacity and less power: How 5G NR can reduce network energy consumption, Ericsson, https://www.ericsson.com/en/reports-and-papers/research-papers/how-5g-nr-can-reduce-network-energy-consumption
2. Tim Frost, The PTP Telecom Profiles for Frequency, Phase and Time Synchronization, May 2013. https://www.microsemi.com/document-portal/doc_download/133481-ptp-telecom-profiles-for-frequency-phase-and-time-synchronization
3. TS 38.104 Section 6.2
4. Neutral Hosts on JOTS NHIB, Small Call Forum document 250.01.01 https://www.smallcellforum.org/neutral-hosts-on-jots
5. Joint Operators Technical Specifications, Mobile UK, https://www.mobileuk.org/jots

Jim Olsen is a solutions architect in the frequency and time systems (FTS) business unit at Microchip Technology. He has extensive experience in designing and implementing network synchronization and timing architectures in more than 50 countries. He first joined Microchip in 1984 and has since served in a wide range of service, sales and marketing roles. In 2000, he transitioned to an advanced technologies role, helping identify new technologies and investment opportunities, and is currently a solutions architect for the North America region. Olsen speaks regularly at industry seminars and events and his numerous articles and whitepapers on synchronization and timing have appeared in books and trade publications.