Microwave Journal
www.microwavejournal.com/articles/34412-the-growing-importance-of-oscillators-with-5g

The Growing Importance of Oscillators With 5G

August 13, 2020

Normally, the most talked about aspects of 5G in the RF front-end are the deployment and functionality of massive MIMO (mMIMO), mmWave transceivers for small cells and the power density requirements of power amplifiers (e.g., silicon versus GaN). However, the key performance indicators of 5G impact other components in the RF front-end transmit/receive chains, particularly oscillators, where timing and synchronization requirements drive the need for more precise oscillator design and fabrication.

Cooperative radio techniques such as inter- and intra-band carrier aggregation (CA), MIMO, downlink coordinated multi-point (CoMP) transmission and reception and uplink CoMP require much tighter synchronization than traditional 4G technologies. The entire synchronization chain, including the air interface at the remote radio unit (RRU) and introduced errors from fiber, switches and routers at various nodes must be considered. The end-to-end (E2E) latency for time-division duplex (TDD) 5G networks is 1.5 μs; this, however, is only a foundational latency requirement that gets progressively tighter with cooperative radio techniques. The newly introduced enhanced Common Public Radio Interface (eCPRI) protocol has made provisions for this, which makes official the increasing importance of a stable source for an effective 5G network.

5G xHAUL: CPRI CONSTRAINTS

Previous generation cellular base stations comprised a baseband unit (BBU) and remote radio head (RRH) connected to the antenna through a run of coax. The RRH handled the conversion between the digital and RF signals, while the BBU handled the bulk of the processing by providing the physical interface between the base station and the core network.

The LTE base station (eNodeB) improved this with an integrated antenna and RRH connected to the BBU through an optical fiber using a CPRI signal, eliminating RF cable loss and interference. The 3GPP new radio architecture now consists of a centralized unit (CU), distributed units (DU) and RRUs, where the 4G BBU functions are split into the DU and CU. This network architecture (see Figure 1) includes fronthaul, midhaul and backhaul infrastructure to handle the capacity, latency and reach requirements of 5G. Instead of the CPRI interface between the BBU and the RRH in 4G, 5G fronthaul architectures will likely leverage the eCPRI interface between the DU and the RRU. The eCPRI protocol, however, is not limited to fronthaul; it can service connections between the CU and DU.

Figure 1

Figure 1 5G Location of fronthaul, midhaul and backhaul in the network.

Wireline Solutions: CPRI vs. eCPRI

The CPRI protocol is a standard digitized format mainly used to transfer point-to-point data over fiber to separate the radio equipment (RE) from the radio equipment control (REC). This enables the 4G eNodeB configuration with a BBU (REC) separate from the RRU (RE), which is often integrated with the antenna. However, CPRI does not scale well with base stations that have a functional decomposition - specifically, a functional split within the physical layer (intra-PHY split). The intra-PHY split is necessary in 5G, enabling high data rate functions such as CA, network MIMO, downlink CoMP and uplink CoMP. This led to the release of eCPRI, with the goal of “decreasing data rate demands between eREC and eRE via a flexible functional decomposition.”

Figure 2

Figure 2 CU, DU and RRU functions in high layer (a), low layer (b) and cascaded split points (c), with the corresponding intra-PHY eCPRI split (d).1,2

Figure 2 shows the functional split in various 5G architectures described in ITU-T GSTR-TN5G, as well as the intra-PHY downlink splits (ID, IID) and uplink split (IU) specified in eCPRI.1 More often than not, an eRE corresponds to an RRU, while the eREC includes the CU and DU functions.

Table 1


eCPRI Latency and Synchronization

Figure 3

Figure 3 The UNI is the physical point defining subscriber/service provider responsibility. (Source: eCPRI).

This flexible functional decomposition is accomplished by laying the eCPRI protocol layer above the packet-based transport network layer, which could be IP- or Ethernet-based. The new eCPRI protocol also comes with updated timing requirements for handling multiple REs. Where the CPRI quality of service requires a maximum overall round-trip link latency of 5 μs (R-26), the asymmetry of eCPRI calls for more variation in this requirement. The eCPRI standard includes various classes of service where the maximum one-way frame delay can go as low as 25 μs for ultra-low latency performance (see Table 1). The one-way frame delay must include both the fiber propagation delay from ingress and egress of various user network interfaces (UNI) as well as the switching delay from the transport network (see Figure 3). The Ethernet virtual connection can contain several UNIs.2 This causes much more stringent delay requirements on the switches and routers within the transport network, in addition to the already strict air interface requirements on the eRE. Typically, the eREC or CU/DU, does not require the same stringent synchronization and timing requirements that an eRE or RRU needs, since the eRE will generate the frequency for air transmission locally.

Figure 4

Figure 4 Various clocks in the synchronization path, illustrating the timing accuracy requirements shown in the tables.

Two kinds of time errors are specified for eCPRI timing accuracy: absolute and relative time error. Absolute time error (|TE|absolute) is the difference in time between the primary reference time clock (PRTC) and the local clock (see Figure 4). Relative time error (|TE|relative) is the difference in time between UNIs of a local cluster, which can be as low as 20 ns to adequately provide the time alignment error (TAE), or time error between transmitter antenna ports as required by 3GPP (see Table 2).2 Ethernet- or IP-based transport network synchronization can be accomplished via several standard protocols, such as synchronous Ethernet (SyncE) or precision time protocol (PTP), so long as the timing accuracy between UNIs is met.

Table 2

Network Synchronization Chain

Depending on the architecture of a wireless network, there are generally three main clocks for time synchronization of packet-based time and phase synchronization methods (i.e., NTP, PTP, SyncE standards): the PRTC, the packet master clock and the packet slave clock. The synchronization chain synchronizes the highly stable master clock with the slave clocks down the line. The ITU defined PTP telecom profile renames these clocks as the PRTC, the Telecom Grand Master Clock (T-GM), the Telecom Boundary Clock (T-BC), the Telecom Transparent Clock (T-TC) and the Telecom Time Slave Clock (T-TSC). Timing support is typically accomplished in intermediate nodes (e.g., switches and routers) through the T-BC. As shown in Table 2, the T-TSC can either be integrated into the end application (e.g., the eRE or RRU) or be external, delivering a phase/time reference to the end application via a synchronization distribution interface (e.g., 1PPS or ToD) as shown in Figure 5. Timing requirements are much tighter where the PTP termination is at the UNI, or when the T-TSC is separate from the end application clocks. The maximum TE at the UNI listed for category C is the same as the maximum TE at reference point C or D. This is to meet the 5G TDD requirement of 1.5 μs E2E latency. Requirements become tighter with cooperative radio techniques where the maximum relative TE requirements exist within a cluster, when multiple RRUs are connected to the same DU.

WHAT THIS MEANS FOR OSCILLATORS

In the ITU-T G.8271.1/Y.1366.1 report,3 the network limits up to reference point C (see Figure 4) involve two types of noise generation from the PRTC, T-GM, T-BC or T-TC, which are constant and dynamic in nature. Noise generation is expressed in terms of TE, where the constant TE (cTE) is produced by the chain, and the dynamic TE (dTE) is attributed to the low and high frequency noise components of the chain. The low frequency dTE components, defined as below 0.1 Hz, can be measured with maximum time interval error (MTIE) and time deviation, while the high frequency dTE components, which are above 0.1 Hz, can be measured with peak-to-peak TE.

Figure 5

Figure 5 Time synchronization with maximum absolute time error per ITU-T G.8271.1/Y.1366.1.3

The master-slave synchronization chain relies upon the PRTC, the timing from the protocol itself and the holdover clock, which is meant to maintain phase/time information when the T-BC loses its input phase and time references. The holdover clock can consist of either a stable internal local oscillator (LO) or receive an assist from a primary reference clock traceable signal. Wander generation, or the slight differences in clock signals in a network over time, occurs intrinsically with white and flicker frequency modulation, as well as extrinsically through random walk frequency modulation from aging, power supply variations, temperature, vibration/shock and frequency drift during a switchover period to a holdover mode. Meeting the holdover requirements for the various classes of T-TSC and T-BCs in ITU-T G.8273.2 is necessary to ensure overall TAE requirements.

Oscillator Constraints for Local Clusters

t3.jpg

Phase noise and MTIE requirements beyond reference point C (see Figure 4) are also vitally important to maintain timing within a local cluster of RRUs. The various 5G radio techniques rely on clean and stable RF sources with tight individual phase noise requirements over the cluster. Some of these radio techniques and their respective effects on oscillators are discussed below and summarized in Table 3.

MIMO Systems - Phase noise is known to negatively impact channel state information (CSI) - information on the propagation path for a signal from transmitter to receiver, including scattering, fading and power decay parameters - and cause channel aging for multi-user (MU) MIMO systems. CSI is especially important for systems relying on linear precoders to mitigate the effects of MU interference. Any difference between the estimated CSI and the real transmit path (e.g., from channel aging) is detrimental to system performance. Phase noise can cause a time-varying and random phase difference between the oscillators at the base station and the user equipment, with unpredictable rotations of the transmitted data symbols. This will invariably affect future installations of mMIMO with synchronous or asynchronous frequency generation, where there may be motivation to use low-cost local oscillators.

High Level QAM Systems - Orthogonal frequency division multiplexing (OFDM) enables similar data rates and bandwidth compared to single-carrier modulation schemes, while offering more immunity to severe channel conditions. Often, OFDM systems use a high order quadrature amplitude modulation (QAM) for each subcarrier. Next-generation cellular installations continuously expand the symbol set to increase bandwidth efficiency. This increase in constellation size and, subsequently, decision points becomes more sensitive to the effects of forward path impairments such as phase noise, which often manifest as slight shifts in the position of the constellation points. Ultimately, poor phase noise performance of the LO impairs the signal bit error rate (BER).

5G mmWave Systems - Small cells relying on non-line-of-sight or line-of-sight microwave backhaul have particularly difficult synchronization problems due to the nature of packet timing. These problems are exacerbated using mmWave signals, as more small cells are needed to adequately cover an area. This, in turn, tightens the latency and synchronization requirements because of the increasing number of intermediate nodes for backhaul. Phase noise also generally increases with carrier frequency; for example, frequency multiplication increases the phase noise, and high frequency crystals have lower Q. This adds another layer of complexity to the already complex latency and synchronization needs of 5G.

CONCLUSION

Network timing and noise sources are major considerations with 5G installations, as the transport network and air interface must work in tandem for reliable latency and synchronization. The intrinsic and extrinsic wander and jitter of RF sources in the timing chain require serious consideration. Aside from latency, phase noise may directly affect the BER of a wireless installation and degrade its reliability. Using cost-effective sources, stable over temperature and vibration, is essential to the performance of the 5G air interface and xhaul. In some cases, a phase-locked loop may be necessary to stabilize the phase noise, although with the penalty of cost and complexity. Stable oscillators and frequency synthesizers for commercially viable mmWave communications add another level of complexity to meet the needs of 5G.

References

  1. “Transport Network Support of IMT-2020/5G,” International Telecommunications Union, February 9, 2018, Web: https://www.itu.int/dms_pub/itu-t/opb/tut/T-TUT-HOME-2018-PDF-E.pdf.
  2. “Common Public Radio Interface: Requirements for the eCPRI Transport Network – V1.2,” June 25, 2018, Web: http://www.cpri.info/downloads/Requirements_for_the_eCPRI_Transport_Network_V1_2_2018_06_25.pdf.
  3. ITU-T G.8271, “Network Limits for Time Synchronization in Packet Networks,” International Telecommunications Union, October 2017.