The engineering challenges involved in moving to small cells

4 mins read

More people will be able to make mobile calls and post social media content because mini base stations will boost capacity. But sprinkling small cells within a LTE macro cell is causing an engineering rethink.

Radio engineers are having to balance the benefits of capacity with the placement of radio elements – small cells, radio heads and baseband units – to best cope with burgeoning networking requirements. Even the make up of a small cell is under scrutiny. Already, a new small cell interface standard is likely to appear in 2015, with silicon to follow. Small cells range from indoor family units to outdoor ones serving hundreds of users (see box). Small cells improve radio performance whilst providing a cost effective way to expand LTE networks. "The best way to deliver data rates is to focus on small cells," said Nick Karter, vice president of product management at Qualcomm Atheros. "The small cell brings a significant cost reduction factor compared to a macrocell." But small cells also introduce complexities. Interference is one; small cells use the same spectrum as the macro cell. Moreover, techniques that boost radio performance and counter interference require cooperation between cells, placing demands on the network. One such scheme – enhanced intercell interference coordination (eICIC) – uses time multiplexing to reduce interference. "When the macro cell is quiet, all the small cells can use that spectrum as theirs," said Ulrich Kohn, director of technical marketing at ADVA Optical Networking. eICIC requires the sharing of frequency and timing information between cells. Frequency information has always been shared, says Kohn, but timing is new. Engineers thus find themselves having to reconcile two contradictory trends. One trend is the distribution of radio intelligence to the network edge with the deployment of small cells. A small cell is self contained, comprising all the required radio functionality: RF and baseband hardware; the LTE software stack and layer-2 (Ethernet) and layer-3 (IP) networking. Having intelligence at the network edge benefits mobile backhaul, with relatively low traffic rates being passed to the core. The second trend, running counter to the first, is centralising radio resources to the network core. Pooling base stations promises higher equipment usage and operational gains. By pooling hardware, capacity can be moved between cells as required, unlike the present, where the full capacity of each cell is rarely used. Pooling baseband functionality also promises the IT benefits of virtualisation and cloud computing. Operators can replace specialist baseband equipment with software running on virtualised servers, an approach dubbed Cloud-RAN. "Virtualisation allows you to build extremely dense nodes that are highly collaborative," said Mike Schabel, vice president of small cells at Alcatel-Lucent. A huge 'virtual' base station also suits LTE developments such as Coordinated Multi-Point (CoMP), a form of MIMO (multiple-input, multiple output) in which more than one antenna is used to send signals to improve cell capacity and coverage. Instead of sending the signal from multiple antennas at one cell site, CoMP extends MIMO's use of radio diversity by using neighbouring cells' antennas, kilometres apart. The LTE macrocell, unlike a small cell, already partitions radio functions, albeit locally. The mast mounted radio head is linked to the baseband unit tens of metres away using the Common Public Radio Interface (CPRI). Employing centralised intelligence, the separation between the radio heads and baseband units grows to tens of kilometres. The downside is that each CPRI stream from the radio head can be up to 10Gbit/s; CPRI carries digitised radio signals, rather than user traffic, which is 50 times smaller. Mobile operators want the greater efficiencies, but may baulk at the networking cost. Hence the need for an engineering rethink to benefit from pooling without the networking expense. "Maybe there is some variation in the middle," said Rupert Baines, a consultant with Cavium Networks' wireless group. "That is essentially what front-hauling is about." The Small Cell Forum announced in June its intention to address virtualisation. This initiative will explore ways of meeting the various small cell deployment scenarios whilst reducing the front-haul networking requirements. Existing LTE base stations and small cells have an S1 interface at one end, used for mobile backhaul to the Evolved Packet Core, and the radio interface at the other end (see fig 2). Small cell vendors and organisations such as the Small Cell Forum are investigating how small cell functionality can be split to reduce front haul traffic. "What does the architecture look like if you shift functionality towards the edge or towards the core?" Baines asked. The main LTE cell functional blocks include the PDCP/RLC/MAC functions that translate between the protocol layers, from time slot data to packets ready for backhauling. Such tasks are performed on a host processor. The FEC/ QAM antenna mapping and resource mapping require DSP, while the FFT/ cyclic prefix and CPRI encode is what Baines calls 'grunt work'. There are four potential interface points where small cell partitioning could be performed: from the MAC-PHY front-haul interface through to the Split III subframe symbol front-haul. Baines says Split III is one possible interface. Here, the IFFT function and cheap digital filtering could be moved to the radio head, reducing the data rate by 5 or 10 times. But he points out that assessing the merits of each potential interface point is far from trivial. "It is an engineering tradeoff. The further up the stack you go, the cheaper the backhaul gets and the easier latency gets," said Baines. "But there is more expense at the edge and we are losing out on eICIC and massive MIMO efficiencies." Cavium's small cell silicon includes the Octeon Fusion, while its Octeon chip performs the PDCP/RLC/MAC functions for 70% of LTE macro cells; equipment vendors provide their own silicon for the DSP and lower layer tasks. Qualcomm offers two small cell chip families: the FSM90xx for residential use and the FSM99xx for enterprise and metro. Both are Small Cell Forum members. Alcatel-Lucent and Cisco Systems are also investigating ways to split the small cell's functions and do away with CPRI. "It allows you to get the best of both worlds: use an Ethernet interface, something very simple, and get the benefits of virtualisation," said Alcatel-Lucent's Schabel. Cisco's goal is using IP for front-hauling. "What we are seeing with virtualisation and IP-based front-haul is new ways of deploying radios – how they integrate into the network," said Adam MacHale, CTO of mobility in EMEAR at Cisco. IP and Ethernet are more prevalent and simplify small cell connection compared to fibre or microwave. "Running traffic over IP removes the constraint as to where you put the controller," said MacHale. The Small Cell Forum's virtualisation work is seen as significant by the industry. The organisation's recommendations regarding the interface of choice will be fast tracked into the 3GPP LTE standards body where it will become formalised. But that won't be before 2015.