Airspan Networks are one of the few independent vendors to have focussed on LTE rural and urban small cells for some years. Has their time finally come? We spoke with Paul Senior, CTO, who has come up with some novel techniques to squeeze the highest spectral performance out of standard urban small cells using only standard Ethernet backhaul.
A quick recap of Airspan
This once public, now privately owned company is split between US, Israel, UK and most recently India, with about 330 employees across all of these countries. They had been very successful in WiMAX but saw the demise before it came and flipped their focus onto LTE Small Cells. They've concentrated on making all-in-one-box, low cost basestations with integrated backhaul. About 40% their R&D effort goes into backhaul, which is tightly woven into their solution.
All of the eNB products are based on the same Qualcomm platform using a class Small Cell architecture, single chip SoC, scaled with different radios from 400mW to 20W. They've mostly been visible in urban and rural applications, but recently introduced an indoor product for the enterprise. The backhaul/fronthaul radios are on TI SoCs.
What do you mean by Virtualised Small Cells?
We wanted to research what could be achieved by centralising the Layer 2/3 functions while leaving the small cells to run just the physical Layer 1. We wanted to eliminate the need for super-low latency and dark fibre associated with traditional Cloud RAN architectures. We've been working closely with Softbank Japan and others to evaluate what could be achieved through using outdoor small cells as part of the mix. This has now been commercially released as our AirRAN solution, which combines are AirVelocity, AirSynergy, and AirHarmony eNodeBs with a Blade Server called AirSymphony.
We used an adapted standard femtocell API (FAPI) as the delineation point. Many vendors already support the FAPI and we think many FAPI compliant small cells should be capable of supporting this mode. The Layer 2/3 functions are concentrated in a standard ACTA compute platform located in a datacentre. We can tolerate up to 3ms of latency between the small cells and data centre, meaning this would work on most Carrier Ethernet backhaul circuits and doesn't need dark fibre and CPRI.
The great advantage of using standard small cell hardware provides the choice to remotely configure any cell to be standalone or operate as part of the virtualised group. Operators can choose to deploy initially as isolated, standalone cells and later upgrade to integrate them into a more closely co-ordinated mode.
Which LTE features deliver most benefit when centrally co-ordinated?
The advantage of centralised co-ordination using CoMP (Co-ordinated Multi-Path) is that each UE (Smartphone) is served by a set of different radios using the same Cell ID and same PCI. This gives a huge performance benefit and avoids the majority of handovers. The handset receives the same data packets from multiple sources, increasing resilience, quality and throughput.
Others have achieved this with macrocells or for dedicated Enterprise systems only. We're the only small cell vendor to date to do this for Outdoor Urban, and it's especially effective for the 2 GHz and 3 GHz bands.
In the future more demanding levels of virtualisation are possible but these would require much higher backhaul speeds (say 1Gbps) and microsecond latency. This means fibre or millimetre wave wireless. This would allow more advanced versions of CoMP (especially on the uplink), and support higher order MIMO (say 4x4 or even 8x8) and 8x8 beamforming.
To complement our existing iBridge NLoS backhaul, we are now developing a 60 GHz and 70/80 GHz mmWave backhaul capability that can provide the higher bandwidth and low latency required for that more demanding level.
What was the purpose of the trial with Softbank in Japan?
Japan is one of the toughest and most demanding wireless environments to satisfy. Huge levels of data traffic need to be served to very demanding customers within very dense urban metropolis.
To help alleviate the issue, the Japanese regulator has allocated 40MHz of 3.5GHz TD-LTE spectrum to each of the three national networks, with commercial service due in 2016. But while there may not be the billions of dollars of licence fees found elsewhere, there are extremely demanding service requirements. The high path loss (i.e. short range) at this frequency means this isn't straightforward, and existing macrocell architectures won't be sufficient.
We've been conducting trials with Softbank, concentrating on a 400m x 250m zone in a very busy central area. We used beam shaping antennas which can be remotely controlled. There was a ratio of about 3:1 small cells to macrocell in the test, but I could see 4:1 or 5:1 in the commercial phase. Each small cell was 2x1W (single sector with 2x2 MIMO).
We ran a variety of A/B tests: standalone vs cluster, shared vs dedicated spectrum, omni vs directional antennas, CoMP enabled/disable etc. In all cases, performances improved with additional small cells. Directional antennas provided a more optimised throughput.
CoMP is most useful at the cell edge, achieving 260Mbps throughput compared to 110Mbps – a gain of 140%. Indoors, CoMP achieved 600% gain compared to macro only and up to 100% compared to small cells alone. eICIC separately delivered 45% gain at the cell edge, and both these features can be combined to greater effect.
Overall, we recorded a three times benefit in QoE with small cells and have been very pleased with the performance of our Virtualised Small Cell approach. Our architecture delivers similar performance benefits to using Remote Radio Heads but avoids the need for expensive bandwidth hungry CPRI fronthaul.
What's your view on LTE Relay cells?
The 3GPP standard includes a feature to support remote relays at the cell edge, which only needs power to rebroadcast the signal into poor coverage areas. However, this requires a separate protocol stack in the macrocell – something which not all vendors have implemented.
Instead, we've built a simple relay using a directional antenna to the macro which operates at a different frequency band, say 2.6GHz TD-LTE, and rebroadcasts at 1800MHz FDD-LTE. The antenna form factor and design enables much better utilisation of the link that when serving smartphones directly, using 64QAM rather than QPSK to achieve much higher throughput within the same spectrum and macrocell resources. The short range radio link to the end users also provides the potential for higher speeds and better service quality. It's a quick and effective solution for enterprise buildings at the edge of coverage.
The potential capacity of an LTE Relay isn't insignificant. If we used LTE with 256QAM, 8x8 MIMO we could see a consistent throughputs of 450Mbps.
I could also see this being useful in transport applications, such as for Connected Cars. We'll be releasing products later this year for vehicle based solutions at various frequency bands.
How will this complexity be managed and optimised?
SON (Self Organising Networks) will be key. Network operators often have their own proprietary schemes – they see this as one of the few "secret sauce" areas of network differentiation. Some are more biased towards C-SON (Centralised SON) while others prefer D-SON (Decentralised SON). Unless the small cells have their own dedicated frequency, both are needed.
The mobile industry is moving into its next phase of development, where small cells will be essential to deliver the full potential that LTE can offer. The combination of remote relays, standalone and co-ordinated small cells is more than capable of meeting that challenge.
The network operators who will win during this latest network evolution will be those who embrace the choice that small cells can bring, rather than view them as an optional extra for special cases.