The capacity crunch is coming; in fact it is already here. Demand for mobile data services is booming, driven by the growing popularity of media-rich applications such as Facebook and YouTube, as well as by new devices such as Apple's iPhone. Worldwide data and internet traffic in mobile networks was estimated to be around 100 petabits per month (1 petabit = 10¹⁵ bits or a million gigabits) in 2008 and it is doubling every 18 months or so, Figure 1. In fact, some report packet data usage to have increased by as much as eight times in a year.

Figure 1: Regional breakdown of traffic and anticipated growth

The demand for higher capacity is relentless, and the data rate of next generation LTE and WiMAX networks (being deployed in the coming years) can rise to a peak of 150 Mbps per connection.

Mobile operators are already facing a serious engineering challenge as high data demands of users are aggregated together and carried from the base station sites into the core fibre optic network. The 2.5G and 3G base stations often required only a single E1/T1, 2 Mbps backhaul connection for voice traffic from several simultaneous users. The peak data rate for LTE downlink is 150 Mbps. An LTE cell site with three sectors could need up to 450 Mbps backhaul capacity.

In addition, operators and equipment vendors face a serious commercial challenge, as many of the new applications are free to the users, who are also used to fixed monthly pricing for their broadband. Contracts for "unlimited" mobile broadband data, with fair usage policies, are now available in several regions. One operator in the UK even offers free calls to and from a VoIP client installed on their handsets.

Although the demand for capacity is exploding, operator revenues are not growing at a corresponding rate, and network upgrade equipment must be at the lowest possible price per bit per second. The "capacity crunch" in backhaul is here, and is only going to get worse.

Backhaul solutions to connect base stations to core networks can be wireline, such as copper and fibre, and wireless at various microwave, and now millimetre wave, frequencies:

1. Copper: Gigabit Ethernet demonstrates that copper wire supports high-capacity transmission but, as anyone who lives too far from their local exchange to receive DSL broadband knows, copper wires (especially the old ones) struggle to maintain high data rates over a long range. If the network node is installed in an urban or suburban area, copper wire might be available from a telephone or cable TV provider. However, the cost of installing new cables, including digging up roads and traffic disruption, is very high.
2. Fibre:By comparison, fibre optic connections offer enormous range, and new laser and modulation techniques are driving capacities well beyond 40 Gbps. However, fibre is not available everywhere. Today, fewer than 10% of US cellular towers have a fibre connection within easy reach, and the installation cost is as high for fibre as it is for copper.
3. Microwave: One can easily see, from the large number of dish antennas mounted on cellular towers worldwide, that microwave links are a very popular solution for backhaul, Figure 2. Perhaps as much as 70% of all mobile backhaul is over microwave. As they can be quickly and easily installed without the costs and disruption of laying fibre, microwave links are an obvious choice when building new networks. A third of operators building out new WiMAX networks will only use microwave for backhaul.
   
   
   Figure 2: A typical microwave dish tower
   
   

   Microwave links for these applications operate in regulated frequency bands from 5 or 6 GHz up to around 38 GHz, depending on the governing authority. To support frequency re-use in urban areas, the bands are planned, licensed, and quite narrow. Achieving the high data rates needed for future LTE systems within narrow 28-MHz or 56-MHz channels is a challenge for microwave equipment makers. It requires high-order modulation schemes, very linear microwave components, and complex digital processing. This increasing complexity drives up cost.
   

   Substantial "truck rolling" costs are involved in sending staff to maintain and repair links, but microwave equipment has been proven over many years to be as reliable as equipment for fibre, and certainly to have a higher Mean Time Before Failure (MTBF) than copper connections. Unlike copper and fibre, the range and performance of microwave links is dependent on the local weather, and can be subject to "rain fade". This is increased absorption and Bit Error Rate (BER) when rain passes through the link, and can result in a complete loss of signal. "Carrier grade" links for mobile network operators must be guaranteed operational 99.999% of the time at their particular location. This is often referred to as "five nines" availability.
   

4. Millimetre wave: Another solution is now available that retains the ease of installation of microwave, but takes advantage of 10 GHz of bandwidth available at higher millimetre-wave (MMW) frequencies. The US FCC has allocated the "E-bands" from 71 to 76 GHz and 81 to 86 GHz for use by Point-to-Point communications links, and has implemented a "light licensing" regime in which users register the location of their link on a central web-based database on a "first-come-first-served" basis. This avoids the often protracted and expensive frequency planning process, but still provides protection against a later installation in close proximity. CEPT & ETSI are also well down the road of allocating these bands, also under this "light licensing" regime, in Europe.
   

   These MMW technologies are already proven in deployments to extend private Gigabit Ethernet networks across campuses, business districts, and to backhaul CCTV video from city centres and traffic intersections to central control rooms. The MMW links currently deployed in the US carry data in both directions simultaneously by Frequency Division Duplex. Although some MMW radios transmit and receive within a single 5-GHz band, most transmit in one 5-GHz band and receive in the other, allowing a less complex and lower-cost diplexer.
   

   In these wide bandwidths, even simple modulation such as ASK or BPSK offers more than 1 Gbps. In the new European allocations, each of the 5-GHz bands is divided into nineteen 250-MHz channels to support frequency re-use as well as the trend towards network infrastructure shared by more than one mobile operator. However, channels can be paired and consolidated to provide wider bandwidth and higher capacity on a link-by-link basis.

In mid-European rain conditions, the working range of a 13-GHz microwave link can be up to 40 km, a 23-GHz link is more than 10 km, and a 38-GHz link up to 5 km. The range of an "E-band" link with five nines availability is 2 to 3 km at the maximum regulated output power. Even so, this isn't as much of a disadvantage as it might seem, because most of the high data-rate users, and much of the new high-capacity infrastructure, are in urban and suburban areas where the length of the backhaul "hops" to the core fibre network are much smaller.

Comparing the cost of millimetre wave
Low-cost radios are paramount in deploying wireless links to tackle the backhaul-capacity crunch. Microwave links are a proven solution and can now cost less than copper or fibre. The cost of a 2-Mbps leased line (E1) in Europe is 600 euros ($800) per month, and often more than $1000 per month. This is one reason that microwave links, with a typical cost under $500 per month, are very popular in European mobile networks.

Historically, the cost of millimetre-wave equipment has been high, partly due to the fact that semiconductors for 70 and 80 GHz have been more expensive than monolithic microwave integrated circuits (MMICs) at 23 and 38 GHz. However, this is changing as the quantities of MMICs increase. New, low-cost processes manufactured on large wafers, such as silicon germanium (SiGe), can operate at these high frequencies.

Conventional millimetre wave components have also demanded expensive, precision machined metal with tight tolerances, hand-tuned to avoid a number of effects that arise when wavelengths approach one millimetre--the same size as MMICs, cavities, and bond wires. These effects can completely destroy performance.

MMIC Solutions in the UK is using technology to negate these issues and so enable the use of lower cost materials and construction, as well as allowing automated module manufacturing, Figure 3. With these technologies, the cost-per-bit-per-second of high-capacity millimetre-wave radios is competitive with their low-capacity microwave counterparts.

Figure 3: MSx600 MMW radio PCB from MMIC Solutions Ltd.
(Click on image to enlarge)

A true comparison should also include the complex digital-modem processing needed for the high-order modulation schemes necessary to achieve high throughput in a narrow microwave channel. This is often implemented in a separate indoor rack-mounted processing unit (the IDU) in addition to the outdoor enclosure connected to the dish antenna (the ODU). In contrast, today's millimetre wave radios are already delivering more than 1 Gbps full-duplex capacity in a single ODU, with simple binary phase-shift keying (BPSK) modulation interfacing directly to optical fibre.

Due to the wavelengths used, high-gain microwave antennas are large and costly. Microwave radios often use smaller, lower-gain antennas to address the concerns of planning officials and local residents. These radios compensate for their smaller size through higher output power. They are almost always tower-mounted by specialist riggers.

In the case of millimetre-wave technology, even high-gain antennas can be small and low cost, Figure 4. With a little training they can be mounted unobtrusively on buildings and even behind windows in some cases.

Figure 4: Example of a MMW antenna and radio

The challenge is that legacy networks are time-synchronized and have Quality of Service (QoS) metrics based on voice. As everyone who has made a call using VoIP over broadband knows, voice calls are very sensitive to delays in the network. GSM networks are time-synchronized by the Time Division Multiplex (TDM) stream and base stations in CDMA networks by the GPS time signal. To provide QoS for voice over IP networks, new standards are being developed, such as "synchronous Ethernet" and "pseudo-wires" to transport packets of voice data in a virtual wrapper of time information. Millimetre-wave links have demonstrated very low latency at 1 Gbps by connecting directly to optical fibre, but achieving low latency though the complex modem processing for high-capacity microwave is non-trivial and often costly.

Data traffic is also often very asymmetric, with much higher capacity required for the downlink to the mobile device than the uplink to the network. Optimizing the network for this can conflict with a design which provides switched circuits for voice traffic. Network architecture, often a tree-and-branch structure, may also change for Ethernet-based networks to reduce costs. For example, mesh architectures can allow multiple, redundant paths at lower cost than conventional "protected" links: two microwave links side by side. Millimetre-wave Ethernet radios are a good solution for mesh architectures, each capable of multiple high-capacity channels and with small antennas.

Can millimetre wave solve the backhaul capacity crunch?
Millimetre wave is obviously not a panacea for all of the issues in high-capacity data backhaul, but does offer very wide bandwidths, and is easily installed. Although lower range than low-frequency (and usually lower-capacity) microwave links, millimetre wave will be used in urban and suburban areas where the capacity demand is greatest and small antennas are very desirable, often in new network architectures with multiple paths to guarantee resilience. As new semiconductor and module technologies are widely adopted, millimetre-wave equipment costs will become competitive with lower-capacity microwave radios.

Millimetre wave is already a clear winner on cost-per-bit-per-second, and will be the major factor in solving the backhaul-capacity crunch.

About the author
John McNicol, Director of Business Development & Marketing, joined MMIC Solutions Ltd., Bromesberrow, Herefordshire, U.K. in 2007, having assisted numerous technology start-ups as a consultant and advisor. John's experience in microelectronics spans more than 20 years, including marketing and business development roles focused on wireless and microwave applications. He has a BSc. in Physics from the University of Nottingham, UK, and researched in compound semiconductor physics. He also has a Masters in Business Administration from the UK's Open University.