Network timing and synchronization used to be fairly straightforward. Basestations were connected using 2Mbps links synchronized to a central master clock. Expensive OCXO crystals costing several hundred dollars or more provided holdover times of many hours, allowing almost instant recovery after outages of hours or more. In many networks, a single technical designer might be responsible for the synchronization across the network on a part-time basis.
The enormous growth in mobile broadband data is leading to growing numbers of small cells, especially Metrocells, connected using IP/Ethernet backhaul. This, combined with the introduction of stricter timing requirements for LTE and LTE-Advanced, has some implications for the choice of wireless backhaul technologies used. Here, we explain the technology and discuss the alternatives.
Data traffic has led to IP/Ethernet predominating
Today's mobile networks are now predominantly handling data traffic rather than circuit switched voice – as much as 99% - which has led to a technology shift in backhaul links. IP over Ethernet is used for both wired and wireless links. Initially, operators were cautious about this change and many continued to run both technologies in parallel to many of their larger cellsites. Confidence in the synchronization/timing solutions available for Ethernet has grown, resulting in the withdrawal and replacement of the older and more expensive 2Mbps connections.
Analyst firm Mobile Experts predicted the Metrocell wireless backhaul market size will grow to $1.5 Billion and 1.8 million links by 2016, with over 60% of metrocells connected wirelessly.
Femtocells paved the way
From their earliest days, Femtocells have used only IP for backhaul. Considerable innovation and expansion of NTP (Network Time Protocol - designed originally to sync time of day on computers) was used to achieve adequate frequency synchronization using inexpensive TCXO crystals at remarkably low cost. This can lead to quite a high signaling overhead of up to 10 small packets per second. Residential Femtocells can require quite some time to recover and restore synchronization after a power or internet outage – it largely depends on the quality of broadband connection used.
High capacity public access metrocells are more demanding
Now that small cells have evolved to include higher capacity public outdoor metrocells, 4G/LTE technology and tightly integrated HetNets, the requirements for timing and sync have changed. The impact of intermittent power and backhaul outages lead to demands for much shorter recovery times. The use of LTE Advanced demands not just frequency sync (transmitting at the correct frequency to within 100 parts per billion) but phase sync (transmitting frames aligned to within 1 microsecond of every other basestation). While GPS receivers have been used as one solution for large outdoor macrocells, street level metrocells may not have sufficiently clear view of the sky to rely on it.
These requirements have also a knock-on effect on the choice of wireless backhaul used. What we are hearing is that most of the Metrocell wireless backhaul RFPs (Requests For commercial Proposals) mandate that both Synchronous Ethernet (SyncE) and IEEE 1588v2 technology are supported across the wireless links. The SyncE provides the accurate frequency clock and the 1588v2 supplements that with accurate phase alignment.
I'm also hearing about higher specifications for Metrocell timing oscillators, where the increased cost is justified by a longer holdover time that helps cope with short term outages. The wider temperature range found outdoors also drives a higher specification.
The two main approaches for carrying SyncE synchronization data
While 1588v2 operates at Layer 2 using standard IP packets, the SyncE signal is embedded in the Ethernet Physical Layer (Layer 1).
Radio systems come in two flavours; analogue and digital.
- Analogue radio systems directly modulate the radio carrier from the incoming Ethernet signal in the same way traditional PDH microware radios do. These systems can faithfully reproduce the SyncE signal at the far end of the radio link but have the disadvantage that there is no error correction added to the signal. This means that to ensure that the radio link isn't interrupted by atmospheric fading, the links have to be kept short to allow a large fade margin. These links typically have >6dB less system gain than a digital system. This means that the link distance for an analogue system is half that of a digital system for the same throughput and availability, making them economically less attractive. Another disadvantage of these types of systems is there is no way to implement adaptive modulation schemes due to the fixed nature of the airside implementation.
- Digital radio systems chop the incoming data stream into fixed_size chunks and add data correction bits to form an airside frame. These airside frames are then transmitted over the radio link. These airside frames may contain part of a Layer 2 Ethernet frame or a number of Layer 2 Ethernet frames depending on their size. At the far end of the radio link any errors in these chunks are corrected and the Layer 2 Ethernet frames re-assembled for onward transmission. These types of systems have the advantage that error correction can be added to the data stream to improve system gain, and it is also possible to implement adaptive modulation schemes along with QoS mechanisms that add further economic value. The downside of these schemes is that the frame_based nature of these systems make it more difficult to transmit synchronization data from one end of the link to the other. Processing at both ends includes packing into frames, adding/removing error correction bits and queuing the frames in between data packets of varying sizes. This can lead to jitter (short term variation in the delay when receiving the signals). Techniques to address this require the far end of the radio link to converge the recovered timing to the required level of accuracy over a period of time. This can take anything from a minute to hours depending on the implementation.
Wireless links may operate in either FDD (using separate frequencies to send/receive in parallel) or TDD (sharing the same frequency to ping-pong data alternately in time). Although it may seem that FDD would be easier to use to pass the timing data across, in practice TDD systems have an advantage.
In order for TDD to work in the first place, the system will already have had to determine and align exactly when each side will be transmitting and receiving. The framing scheme for TDD is not usually accurate enough on its own to pass the synchronization/timing data, requiring extra processing to achieve the required accuracy. The jitter buffer in the TDD system is adjusted until it settles at the mid-point. Clever algorithms have reduced the acquisition time for this to less than a minute.
An innovative technique for rapid sync recovery
Vendors have been working hard to reduce the synchronization acquisition time after outages For example, Sub10 have come up with two innovations to deal with the problem (now branded as their SnapBack feature). Firstly, they transmit the sync information separately from the data. Even if the data connection is lost, the sync will continue to operate down to a further 10dB lower signal level. Secondly, sync can be restored in less than a second after a complete outage.
They have also come up with a novel algorithm to measure delay very precisely and rewrite the timestamp in IEEE 1588v2 frames with a correction immediately prior to transmission. This makes it very deterministic rather than best-effort.
Are we testing conformance appropriately?
The test specifications used to check conformance were originally developed many years ago for wireline circuit switched networks. Assumptions were made about overall network design, apportioning the end-to-end jitter/wander across anything up to 10 hops. This has led to lower tolerances for individual links, potentially allowing higher jitter levels to remain within the specifications. This could lead to a much longer acquisition time and higher timing/sync packet overhead. Work is ongoing throughout the industry, including both wireless equipment and test vendors, to address this.
We are seeing growing investment in dedicated silicon chips, targeted for use in wireless backhaul links, which could lead to improved timing/sync performance and substantially lower costs. Many examples include Broadcom's recent announcements for Microwave RF silicon and Bridgewave's 60GHz silicon.
Mass-market/volume deployment will help drive towards that goal. There could be a spin-off from WiGig (unlicenced gigabit Wi-Fi service at 60GHz aimed at consumer devices, but closely related to 60GHz V-Band equipment).
According to Mark Stevens, CTO of sub10 Systems, many of today's WiGig silicon vendors are recognizing that the wireless backhaul market may be buying their chips before the consumer market takes off, but nobody is yet offering a solution which includes the timing/sync capability. In his view, a mass-market chipset will be the only way to achieve true low prices (of around $1000 a link).
DISCLAIMER: Sub10 were acquired by Fastback Networks in March 2015.