Next generation clock application and technology trends

Ultra-large-scale computing and LTE-A drive high-performance timing solution requirements

Author: James Wilson, Silicon Labs Timing Senior Director of Marketing

Data center and wireless network infrastructure continue to increase network utilization and reduce the cost of data transmission. Industry sequential component suppliers use high-performance clocks and oscillators to meet this market demand for optimal frequency. Flexibility and ultra-low jitter.

Ethernet (Ethernet) has come a long way since IEEE 802.3 was first released in 1980 . Ethernet was originally developed as a technology for connecting personal computers (PCs) and workstations, and then gradually evolved into a network technology for a wide range of applications in enterprise computing, data centers, wireless networks, telecommunications, and industrial applications.

Due to the popularity of Ethernet networks and the declining hardware costs of the required support, Ethernet networks will continue to gain greater penetration in these applications. Some of the most interesting technological changes are coming soon, such as 100G Ethernet networks being used in data centers and wireless access networks. These trends toward high-speed fiber-optic Ethernet networks continue to drive demand for higher-performance clock and frequency control products.

In response to the changes in the above market and application technologies, Mr. James Wilson , Senior Marketing Director of Timed Products of Silicon Labs (also known as “ Core Technology ” ) , recently wrote a technical article that will help the industry to grasp the next generation of clock applications and technology trends. .

data center

With the rapid migration of traditional enterprise workloads to the public cloud infrastructure, there is a huge global investment boom in data centers. In addition to the increasing demand for low latency, data centers face a unique challenge of keeping most of the data traffic in the data center and processing the data across multiple compute nodes.

Modern data centers are the best of its network infrastructure, interconnected with each other through so that each switch to support the virtualization of distributed computing, which is the so-called "ultra-large-scale operation" (hyperscale computing) trend. One of the fundamental technologies that make commercialization of ultra-large-scale computing is high-speed Ethernet, and the rapid transition of data center switches to 25G , 50G, and 100G Ethernet networks to accelerate data transfer and network efficiency.

The transition from 10G to 25/50/100G Ethernet networks is driving data center equipment manufacturers to upgrade switches and access points to higher speeds, which in turn require higher performance, lower jitter timing solutions. Program. In these applications, ultra-low jitter clocks and oscillators are necessary because high clock noise can result in unacceptably high bit error rates or communication interruptions. Table 1 shows typical timing requirements for Ethernet physical layer (PHY) , switches, and switch fabrics. A safe and reliable way to make high-speed Ethernet networks use ultra-low jitter clock sources to provide excellent jitter margin for these Ethernet specifications.

Wireless access network

The wireless network will transition from 4G/LTE to LTE-Advanced and 5G in the next few years and is expected to undergo significant changes. The next generation of wireless networks will be optimized for carrying mobile data. 2, 2021, mobile data traffic is expected to increase to 49 per month in Beijing one hundred bytes (exabyte), 7-fold growth over 2016. To support this exponential growth required for bandwidth, wireless networks are redesigning and optimizing data transmission. The widespread adoption of high-speed Ethernet networks in the Wireless Access Network (RAN) is expected to be a key part of the advancement of this technology.

In the 4G/LTE wireless access network, the radio frequency (RF) and baseband processing functions performed by the base station are divided into independent remote radio heads (RRHs) and baseband units (BBUs) . As shown in Figure 3 , each RRH is connected to the BBU via a dedicated fiber based on the Common Public Radio Interface (CPRI) protocol . This architecture allows the replacement of dedicated copper and coaxial cables between the radio transceiver ( usually located in the base tower ) and the base station ( usually located near the ground ) and allows the BBU to be placed in a more convenient location to simplify Deployment and maintenance.

Although this network architecture is more efficient than traditional 3G wireless networks, the bandwidth is limited by the speed of the CPRI link ( usually 1 Gbps to 10 Gbps) , which is limited. In addition, the CPRI connection is a point-to-point link. The delay caused by the fiber and its changes make the RRH and BBU usually deployed near each other (< 2km to 20km distance ) , which limits the flexibility of network deployment.

In eCPRI and other 5G pre-standards (pre-5G) , many time-critical processing will be done in the RRH , so eCPRI links can tolerate more delays, which makes network deployment more flexible, C- The BBU can be deployed further away from the RRH .

As part of 5G evolution of the wireless industry are rethinking base station architecture, as well as the connection between the baseband and RF unit - that is, through the "future Internet" (fronthaul), is a key area to achieve optimization.

A higher bandwidth wide-band network is sufficient to enable new LTE capabilities to support high-speed mobile data, including carrier aggregation (CA) and massive multiple input multiple output (Massive MIMO) . In addition, network densification and the use of small cell , pico cell and micro cell will bring additional bandwidth requirements to the forward network.

To minimize capital expenditures and operating costs, 5G will use the Cloud-RAN (C-RAN) architecture to centrally deploy the baseband processing (C-BBU) for multiple RRHs .

New standards for legacy networks are under development to support C-RAN evolution. The IEEE 1904 Access Network Working Group (ANWG) is developing a new Ethernet bearer wireless- Radio Over Ethernet (RoE) standard to support the encapsulation of CPRI over Ethernet . This new standard will enable CPRI traffic from multiple RRHs and small base stations to be aggregated over a single RoE link , increasing the utilization of the forward network.

Another working group, the IEEE 1914.1 Next-Generation Forward Network Interface (NGFI), is revisiting the first layer (Layer-1) between RF and the baseband to support more Layer-1 processing in the RRH . NGFI enables the forward-end network interface to evolve from supporting point-to-point connections to supporting multipoint-to-multipoint topologies, increasing network flexibility and enabling effective coordination between different base stations. The 5G Proactive Network CPRI Standard (eCPRI), released in August 2017 , defines the new functional partitioning of base station functions and supports CPRI over Ethernet transmission.

These new forward-end networking standards require a frequency-flexible timing solution to support both LTE and Ethernet clocking in RRHs , small base stations, and ultra-micro base stations , as shown in Figure 4 . These new solutions offer the opportunity for a hardware design to unify all clocks into a single small form factor IC .

Another key challenge is precise timing and synchronization. In general, 3G and LTE-FDD mobile networks rely on frequency synchronization to synchronize all network elements (NEs) to a very accurate and accurate primary reference clock, usually from GNSS satellite systems (GPS , Beidou ). The transmitted signal. These systems require frequency accuracy of the radio interface of 50ppb, and in the backhaul network to the base station interface is required 16ppb.

LTE-TDD and LTE-Advanced retain the accuracy requirements of these frequencies, but add very stringent phase synchronization requirements (+/- 1.5us) . This is a key requirement for implementing new functions such as enhanced inter-base station interference coordination (eCIC) and multi-point coordination (CoMP) to maximize signal quality and spectral efficiency.

These phase synchronization requirements are also expected to further enhance the LTE-Advanced network architecture in the upcoming 5G standard , where multiple RRHs are connected to the centralized BBU via packet-based eCPRI networks , and their phase / frequency synchronization is IEEE 1588v2/SyncE is available. Timing and phase synchronization can be supported by building IEEE1588/SyncE on the RRH and the centralized BBU . A more high-bandwidth 100GbE network is used to implement backhaul between each BBU and the core network. Simplify clock generation, distribution, and synchronization in LTE-Advanced applications with higher performance, more flexible timing solutions .