All TD-LTE and LTE-Advanced base stations and small cells don't just need accurate frequency delivery, they need phase and time-of-day synchronization as well. Three possible options include:
- Using GPS/GNSS, probably the most common today, but for small cells at street level, GPS may not provide satellite line of sight, and is susceptible to jamming and spoofing. So what happens if that fails?
- Listening into and slaving off outdoor macrocells that have reliable Time-of-Day, but how do you distribute phase timing from the macro base station to the small cell, and what happens deep indoors where small cells are outside the range of outdoor signals?
- Using IP timing packets sent over the backhaul connection using the IEEE 1588 protocol, but what happens if these packets suffer variable delay and delay asymmetry common in backhaul networks?
In practice, operators want all base stations including small cells to be equipped with at least two options to handle conditions where the primary source fails. Phase timing accuracy drops off pretty quickly if the timing source fails, even with expensive OCXO oscillators. When synchronisation is lost, small cells have to be taken offline to prevent them from interfering and causing havoc with the network, and it can take some time to re-acquire synch after the source is restored.
We are hearing that the industry is adopting a combination of GPS and IEEE 1588 for macrocells. The approach varies greatly between different regions. , GPS has become the primary scheme in the US, while certain regions like China have a strategy of migrating towards IEEE 1588 as primary timing, using GPS as the backup.
For small cells, especially those outdoors at street level, network architects need to take account of potential GPS issues including poor line-of-sight view of the satellites and the potential for jamming To avoid expensive rip-and-replace at a later date, all new small cell backhaul networks should be capable of delivering reliable frequency and phase timing from the outset. In this way, both GPS and 1588 can be used in most locations to achieve high resilience and uptime.
Variations aren't wanted
One of the issues with using IEEE 1588 is that the protocol is sensitive to packet delay variation (PDV), especially asymmetries in PDV. The protocol sends and receives IP packets between a master server and the small cell, measuring the end-to-end delay and homing in to acquire synchronisation. Clever algorithms are used to adapt and cope with dropped packets and variations when only frequency delivery is required, but as you might expect, more variable conditions increase the acquisition time significantly. For phase and time-of-day, asymmetric PDV cannot be filtered out, which is why the ITU-T is finalizing the standardization of boundary clocks and transparent clocks that eliminate any PDV and asymmetries in hardware on a hop-by-hop basis.
Synchronous Ethernet (SyncE) can be used to provide frequency synchronisation, provided that all links in the backhaul chain support it and pass it on transparently. This can be used to provide a more stable platform and reference point for the 1588 protocol to work from, but also adds to the cost of the implementation.
Another approach is to locate additional master clocks closer to the small cells, such as at backhaul aggregation hubs. These clean up the timing source and act as a local primary timing source. This reduces the number of hops that the 1588 packets have to pass across between the clock source and small cell. It's a good solution when working with older backhaul equipment with larger and/or unpredictable PDV.
When running small cell backhaul over wireless links (e.g., Microwave or Millimeter-wave), as will be common for metrocells and other outdoor small cells, more attention needs to be paid to this issue. The PDV is the critical factor here, and is essential to engineer within tolerance. This is especially important for TD-LTE and LTE-Advanced.
Dealing with a changing wireless environment
A common feature with microwave wireless links is that they automatically change their modulation scheme to adapt to weather and environmental conditions. For example, a rain shower might cause the 64QAM modulation to drop back to 16QAM or QPSK, sometimes with the transmit and receive links using different rates at different times. These transitions affect the link delay budget and are a well-known issue.
Vendors have solved this problem by automatically compensating for the change in modulation, time-stamping the 1588 packets with an adjusted delay immediately prior to transmission to account for it, or implementing distributed transparent clocks over the links. You could compare it to pre-distortion in RF amplifiers, because it predicts and offsets a known characteristic.
Solving tomorrow's technical issues
Martin Nuss, CTO of Vitesse, which designs and manufactures Ethernet Switch and PHY ICs used throughout the backhaul equipment chain, highlights three technical issues his company is working towards today:
1) The sum of the parts adds up to an (unwanted) bigger number
Currently the 1588 specification defines the total end-to-end packet delay variation (PDV) across up to 10 hops (even up to 20 hops in China), with operators such as EE specifying an aggressive overall 500ns figure required for LTE-Advanced MIMO and CoMP (Coordinated MultiPath). Individual wireless backhaul vendors can claim to be IEEE 1588 compliant without being constrained by the end-to-end picture. At long last, the ITU is dealing with the issue and defining timing classes of equipment types.
There will be two classes defined: one allowing a maximum time error of 50ns per hop and a second one of 20ns. This will allow an operator to add these up and quickly calculate the total time error of their backhaul network. Some recent products can achieve less than 10ns time error per hop.
2) Ethernet switching needs to become service aware
Where direct peer-to-peer links have traditionally been used to connect base stations, Layer 2 VPNs (Virtual Private Networks) are sometimes used to connect small cells. A series of daisy chained small cells along a street will require each to be fully synchronised, and additionally carry multiple service connections to each small cell along the daisy chain or partial mesh. The number of service connection could easily increase when multiple wireless service providers share a common small cell network, for example from a third-party backhaul operators, as is quite common in the US and Europe.
3) Using Layer 2 rather than Layer 3 Encryption
Instead of using IPSec, which encrypts each IP packet at Layer 3, it's possible to run Layer 2 encryption over any unprotected network including an Enterprise LAN (e.g., for Enterprise Small Cells for indoor coverage). The MACsec feature of the 802.1AE standard provides similar functionality to IPSec, encrypting the Ethernet packets, and works alongside KeySec protocols, which securely distribute multiple encryption keys to individual endpoints. It's already embedded in Cisco and other leading Ethernet switches, allowing dedicated secure Ethernet RJ45 ports to be provided almost anywhere. The data stream is de/encrypted directly by the Ethernet Switch or PHY chip at each endpoint, effectively simulating a "Copper PHY" or dedicated physical cable, yet with a multitude of individually encrypted connections over a single port.
Phase and time-of-day synchronization becomes more important as we move to LTE, and specifically LTE-Advanced or TD-LTE. For resilience, small cell products will require multiple timing sources likely to include both GPS/GNSS and IEEE 1588.
Network architects will need to ensure that the backhaul is fit for purpose, design it from an end-to-end perspective to achieve tight timing tolerances. This leads to clear time error specifications for each link in the backhaul network including wireless links, with the option to place additional boundary clocks or even edge grandmaster clocks where appropriate for scaling throughout the backhaul network.
Given that small cell backhaul will be deployed from scratch, it makes sense to ensure that it uses newer backhaul equipment with IEEE 1588 hardware support that meets these new ITU-T timing classes. This avoids the need for an expensive rip-and-replace of backhaul equipment when future LTE features are deployed.
Our thanks to Martin Nuss, CTO of Vitesse, for his assistance in compiling this article.