The promised high data rates and huge capacity of 5G place increasing demands on network fundamentals, of which timing and clocking are critical. Rakon has been working closely with standards bodies, chipset vendors, customers and service providers in the 5G ecosystem, to develop products that meet the synchronisation requirements for the new generation cellular technology.
Quick recap on Rakon
Rakon is a world leading supplier of oscillator components for wired, wireless and optical networks, ranging from crystals, XOs and VCXOs, lower cost VCTCXOs through to high spec OCXO modules that are embedded into base stations worldwide. Rakon products have an enviable reputation for premium quality, long lifetime and high performance.
Rakon’s involvement in the small cell industry was a critical factor in achieving the very low price point for mass-produced small cells, where it continues to remain the dominant supplier.
What’s happening with synchronisation in fronthaul for 5G?
A major change for 5G RAN (Radio Access Network) relates to the interface between the BBUs (Baseband Units) and the RRUs (Remote Radio Units), often called fronthaul. In 4G, we called these BBU and RRH respectively, whereas 5G uses a distributed architecture and the nomenclature of CU (Central Unit), DU (Distributed Unit) and RRU (Remote Radio Unit).
4G connects BBU and RRH through a CPRI link, which is a standard point to point physical layer connectivity. It includes a clear specification of the physical layer that is enough to allow fronthaul vendors to transport the data but at the higher protocol level. This has become so vendor-specific that the BBU/RRH almost always has had to come from the same vendor.
CPRI has standard data rates defined – typically several Gbps per radio sector – and embeds synchronisation signals within the physical layer. Each RRH can fairly easily recover this clock information and maintain it using fairly low cost VCXO oscillators.
5G evolves this fronthaul link to eCPRI (enhanced CPRI) but doesn’t specify the physical layer. This helps make it much more efficient and reduces unnecessary overheads. The standard only determines how to frame packets and encapsulate the datastream. It’s quite likely this will be run over Ethernet, because it’s cheap and widely available and has well defined synchronisation architecture for transport requirements.
The consequence of using a generic packet interface rather than defining a physical layer stream is that distribution of timing and synchronisation becomes much more challenging. PTP (Precision Time Protocol) is likely to be used to achieve frequency and phase synchronisation and it will inevitably place tighter timing constraints onto the RRU modules. This in turn means that vendors are being asked to provide higher specification oscillators such as a TCXO or OCXO for every radio.
Network operators are keen to exploit their existing fibre distribution networks and want to achieve similar long distances for eCPRI that they have enjoyed with CPRI. Existing packet synchronisation allows data to have 70−100ns time error per node and up to 10 or 20 nodes. There are stringent delay requirements on top of the air interface requirements between the baseband function and radio function. This places quite extreme demands on each router or intermediate transport node, reducing the tolerance for variability of the packet delay from 70 −100ns per hop for traditional transport down to as little as 5ns per hop for eCPRI.
The air interface requirements of 3GPP necessitates having 50ppb accuracy in 1ms measurement time. Depending on the number of nodes on the front-haul, adequate filtering is needed to achieve this requirement. TCXOs or OCXOs are therefore designed into the radios to achieve the air interface accuracy requirement.
How does higher spectral frequencies affect synchronisation requirements
Another key difference in 5G is the introduction of higher spectral frequencies compared with the previous generation. To maximise the spectrum for efficient communication, the noise contribution of reference clocks needs to be minimised. As the spectral frequencies go up, multiplication of references from a low frequency base adds intrinsic noise to the multiplier. Providing high frequency, low noise reference clocks is a challenge in the 5G radio designs.
The physical properties of the quartz crystal within an oscillator module constrain the fundamental frequency it provides. Typically quartz resonates at up to a few hundred megahertz. Higher clock rates are generated through hardware multipliers elsewhere in the system. Vendors have been tasked to provide oscillators with higher output frequencies in the range of 1.5GHz to 2.5GHz for three main reasons:
- The high frequencies of 5G, particularly when operating in the millimetre bands at say 30GHz rather than sub 3GHz for 4G.
- The high bandwidth allocated to each 5G channel which can be 400MHz compared with 20MHz for 4G.
- Higher modulation rates of 128 or 256QAM. LTE has evolved from 64QAM and we can expect 5G to develop even higher.
Overall, the purity of the clock signal has a direct effect on the system performance and achievable end-user data rates. Phase noise is particularly critical. Minimising the noise contribution to EVM (Error Vector Magnitude) from the frequency reference, which uses a higher quality oscillator output, directly results in higher speeds and total capacity.
Quartz based GHz (1.5GHz to 2.5GHz) XOs and VCXOs provide best close-in phase noise performances required for such applications. Special design techniques make it possible to achieve high frequency, low noise devices which is challenging to achieve with active multiplication techniques.
Who’s making these 5G products today and in the future?
So far the CU and DU are very closely coupled with interworking standards not fully baked yet, so the major RAN vendors are supplying the market. In the longer term, I would expect the market for DUs to become more open. Currently the five or six major companies in the market can’t satisfy every niche requirement. So far however, I haven’t seen any other new entrants with commercial products for sale.
Fronthaul vendors, including Ethernet switch/routers, are starting to claim their products are “eCPRI friendly”. Broadcom has just announced an Ethernet switch chipset specifically designed for CPRI and eCPRI fronthaul. Most of the major RAN vendors have developed their own ASIC chips for use within the DUs.
Increasing volumes and reducing cost
If 5G is to achieve a widespread footprint and higher spectral frequencies then there will need to be many more RRUs than existing 4G RRHs today – perhaps four or five times as many. Hardware cost will be an important factor in each business case. The industry needs to provide a much higher specification oscillator component at significantly lower cost.
Rakon has been working hard to achieve this, and has just introduced a new mini-OCXO as part of its Mercury+ range, which integrates many previously discrete components into a single 9x7mm module. High stability, low phase noise and long lifetime are what you’d expect from an OCXO, but the integration allows it to achieve price points which are 70% less than traditional discrete devices. This brings a major disruption of price versus performance and it will be an important factor that makes 5G RRUs cost-effective.
Rakon has also evolved the VCTCXO range that has been used in many millions of femtocells, tightening the spec from the original 100ppb to achieve 50ppb (-40 to +85 degC) in its latest Neptune series TCXOs. This improves system performance and remains a very cost-effective choice for 5G RRU solutions, where a Rakon TCXO is enough to achieve these stringent synchronisation requirements.
Rakon is a sponsor of ThinkSmallCell
You can find out more at rakon.com
The latest eCPRI specification and an overview presentation can be downloaded from the CPRI organisation website: cpri.info.