Andy Sutton, EE, revealed some of the detailed strategy being used in their initial rollout of small cells as part of their wider 3G and LTE network deployment. This article is based on conference presentations and a webinar given over recent months. Andy reinforces how quickly data traffic can increase once LTE service is available, shares some details about end-to-end latency through their network and explains how today's "SuperCells" are paving the way for small cells in the future.
EE is now the largest UK mobile network and also a fixed broadband network supplier. It shares it's 3G sites and equipment with 3, another UK operator. This means that sites commonly have a 2G+LTE SingleRAN basestation plus a separate 3G MORAN (Multi-Operator Radio Access Network) basestation.
They were the first to launch LTE service using existing 1800MHz spectrum and have now acquired substantial assets in the 800 and 2600MHz bands.
LTE really is a game changer
We've heard this from other leading LTE operators, such as Gordon Mansfield of AT&T. Levels of data traffic rise considerably once LTE is widely available. We saw usage graphs (sadly no scale) which showed a pattern of huge increase in network throughput that aligned with start and end times of specific football matches. You could even tell when half-time was.
Another (non-LTE) operator in the room mentioned that as few as 4,000 users watching video could make a significant difference in total network traffic figures. While Andy wouldn't be drawn on statistics, it's clear that heavy demand from something like the 2014 World Cup could have a major impact.
In common with some other operators, EE also position their use of Wi-Fi as a useful technology as long as used in an intelligent way. LTE is considered a much better technology and would be used preferentially.
Managing the latency budget
Andy is also responsible for the end-to-end latency across the network and walked through each stage of the network. The bottom line is that there are a lot of different systems and aspects which affect it. Delays can arise from each of:
- UE radio layer: There is a need for App designers to be aware and understand their impact on the network. Fortunately, there has been a lot of communications between the industry and app developers, leading to improved behaviour of many apps.
- Managing/optimising TCP
- eNodeB processing
- CSG: Cellsite gateway, which aggregates backhaul from multiple col-located basestations
- Microwave radio or fibre backhaul: It's a little know fact that signals travel faster over wireless microwave than fibre optic.
- MASG: Mobile Aggregation Service Gateway, which splits out the transmission and routes to relevant signaling and bearer gateways.
- Security infrastructure: In the early days, there have been horror stories of lengthy delays through IPsec gateways but these are now less than 1ms
- S1 flex: Each eNodeB can be connected to anything up to 16 MME's for redundancy and load sharing. This could travel through some MPLS P-routers which must be considered
- EPC: Evolved Packet Core
- FGI service LAN environment. This component is critical for multi-RAT and will be for CA (Carrier Aggregation) and HetNet.
- Internet Peering: You can either directly transmit everything into the Internet centrally to actively select which peering point to use. In the main, the Internet is remarkably well engineered and we haven't seen major issues there.
And of course, data is sent both ways, so you must also consider the return path.
Of this long list, the most critical issue remains the air interface, where if there is packet loss, then retransmissions can easily result in delays in 8ms increments.
Typical end-to-end latency figures are 100ms for the 3G network and 30-50ms for LTE. This compares very favourably with the 750ms measured when GPRS was first launched many years ago. It's also more than acceptable for VoLTE (Voice over LTE) where the delay budget beats that of circuit switched voice by some 20-30 ms.
EE's Small Cell categorisation
Today EE have Macrocells and Micros (which they call mini-macros)
EE don't consider residential femtocells as a class of small cell. Instead they categorise them based on use case:
- Small cells as an underlay
- Cell edge
3G wasn't really designed for small cells, whereas LTE is. EE is still planning to deploy some 3G small cells, but could see they would work best for an LTE environment. In dense urban areas, they have had success with "SuperCells" which are 6 sector 2x2 MIMO SingleRAN basestations that can be equipped to use a wide range of different frequency bands. These have given better results than 3 sector 4 MIMO configurations, which use the same total numbers of antenna. These sites can be positioned quite close to each other, certainly less than 1 km apart, and can handle substantial amounts of data traffic. As traffic continues to increase, these will need to be augmented by a layer of targeted public access small cells.
Connecting small cells into the network
These small cells could be connected via the macrocell site as a concentrator or be separately aggregated to an independent fibre hub. It seems likely that a hybrid of both techniques will be used. These would need to be frequency and phase aligned to ensure efficient and effective operation. Small Cell backhaul involves a wide range of situations and so we need flexibility in the choices of backhaul technology used. Typical backhaul bandwidths for 3G small cells are measured in 10's of Mbps, whereas for LTE this increases to 100Mbps or more.
The small cell layer must be cost optimised and consider the evolution to a fully integrated HetNet. Further cost savings arise by powering down small cells when not needed (e.g. overnight).
Carrier Aggregation (bonding two different carrier frequencies together to deliver higher total datarates) would initially be configured on the same cell, which could be either a macrocell or small cell. It is much more complex to do this between a macro and small cell, something that moves into the CoMP (Co-ordinated Multi-Point) feature area.