Most metrocells will be deployed for capacity reasons rather that just to improve coverage. In the medium term, each unit may handle 3G, LTE and Wi-Fi concurrently. For that reason, fairly high capacity backhaul will be preferred – at least 100Mbps. Ideally that would be fibre, but in many cases that is impractical. Wireless backhaul comes with many choices; what are the tradeoffs between capacity, cost and feasibility?
What spectrum is available for backhaul these days?
As an example, the chart below gives a pretty clear indication of what's available in the UK. As you can see, there's much more available as you go higher up the frequency scale, and little left at lower frequencies. Furthermore, the higher frequencies above 60GHz are often free of spectrum fees completely. Being fairly new, there isn't much to interfere with
Surprisingly, there's still a fair bit of spectrum around in some of the more traditional microwave bands today, especially between about 28-42GHz. The UK regulator Ofcom, issued a detailed report on the use of fixed wireless links in 2011 which makes for interesting reading. For example, in the UK, there is a whole 1GHz of bandwidth at 42GHz lying unused today, having been auctioned off recently, it's owner is targeting use for LTE metrocell backhaul.
The report concluded that there was more than enough spectrum above 20GHz available to meet forecast demand for fixed links, with the opening up of higher frequencies above 60GHz reducing the pressure on existing 20-40GHz allocations.
Of course, spectrum availability and regulatory approach varies considerably in different countries and regions.
The report listed all available terrestrial fixed spectrum bands in a table, which CBNL has presented in the tidy graphic format below.
Non-Line-of-Sight is a useful option, but ultimately capacity limited
NLoS uses spectrum below about 6GHz, which is like gold dust, being both rare and expensive. It's a worthwhile option for sites deployed in more difficult to reach locations, especially not-spots or coverage holes. Costly spectrum and limited throughput of such solutions mean these are less suitable for the majority of use cases.
Reusing mass market technologies including LTE itself will make for a very low cost backhaul option, perhaps cheap enough to integrate into units as a fallback and/or initial connection option. But ultimately, using the same spectrum already allocated for LTE could be self-defeating if the purpose is to increase total system capacity.
Traditional Microwave bands – proven technology but with spectrum costs
In most countries, there remains spectrum available in the traditional microwave bands between 6-42GHz. While this requires line of sight, it offers fairly mature technology and high capacity of up to 1Gbps or more. Most microwave equipment has been designed for longer point-to-point links (e.g. from large cell towers) and the form factor wouldn't suit metrocell deployments.
A few vendors, such as Cambridge Broadband and BluWan, offer compact point-to-multipoint equipment that more closely matches size, weight and power requirements for metrocells. Point-to-multipoint also saves on physical equipment, consolidating one end of each link into a shared hub rather than separate physical units. Depending on the country, some spectrum licence fees and regulations apply, which affects the total solution cost.
Plenty of spectrum at 60GHz and above, but initially equipment may be more costly
This has led to growing interest in extremely high frequency, high capacity links in the 60-90GHz range, the so-called V-Band and E-Bands. These frequencies are also known as millimetre wave, referring to the very short wavelength involved. There is plentiful availability of spectrum at zero or minimal cost, allowing equipment to use radio channels up to 1GHz wide and easily achieve data rates in excess of 1Gbps. The short wavelength also favours more compact dish antennas, helping to reduce the visual impact of the equipment.
By contrast, the cost of equipment at these higher frequencies can be significantly higher. Dedicated RF chipsets are only recently becoming available which will lower the price per link in the medium term.
One reason for this is that it is only recently become commercially attractive to use millimetre wave spectrum (60GHz and above). Particularly above about 50GHz, the range of wireless links drops significantly due to physical absorption characteristics – the signal doesn't travel far through water or oxygen. The green line in the chart above also shows that the signal is heavily attenuated around 60GHz by oxygen, limiting link range to 1km or so. Until recently, this had reduced interest in these frequencies compared with traditional microwave which offers much longer range.
Growing demand for short range high speed wireless links, such as between office buildings, campus areas and other fixed broadband applications has increased interest. The limited 500m to 1km range combined with high frequency reuse matches the requirement for many Metrocell links.
Unlike some microwave solutions, these are all point-to-point links and would require a matched pair of equipment at end rather than a single consolidated central hub.
The cost of the RF front end
The next slide digs a bit deeper into the reasons why RF parts cost more for higher frequencies. We're stretching the boundaries of technology and pushing the limits of silicon switching speeds to open up this new spectrum.
60GHz equipment vendors may disagree with these figures, and I'd caution that this is quite a rapidly changing area of technology so the data may become outdated quite quickly, especially as volume of equipment shipped grows.
There is substantial interest and active development of microwave devices at all frequencies today, so one can expect components prices to come down across the board.
Today's tradeoff between spectrum cost and RF component cost
The chart below, courtesy of Cambridge Broadband, graphically illustrates the tradeoff to be made between cheaper physical equipment and cheaper/free spectrum. They argue that the sweet spot today is in the traditional microwave zone of 10GHz to 42GHz.
Others, such as the upcoming 60/70/80GHz equipment vendors may argue that the cost of their equipment is likely to drop fairly dramatically in the next couple of years. As with many types of electronic equipment, it's volume that drives the innovation and investment which will lead to low unit cost. We can already see this coming for Wi-Gig – the consumer 60GHz Wi-Fi variant – and there should be some spin-off from that. Meanwhile, there have been several announcements of RF chipsets for millimetre wave equipment reflecting growing industry uptake.
I'd argue that this chart will look somewhat different a few years from now, with the crossover point drifting slowly to the right.
The question on operator's minds today might be whether they should help accelerate that trend by actively selecting 60GHz equipment today (driving volumes and investment up, shortening the timescale for reduced costs). Alternatively, they may choose to stick to tried and tested technologies which may be more cost effective in the short term.
There is also a low risk that regulators may choose to step in and tax these frequencies, but in my view that would be counter productive because it would deter investment and innovation due to lower volumes.