There is an ongoing debate in the wireless industry with respect to where the capacity is going to come from. Historically, capacity was added in three dimensions:
- Improvements in spectrum efficiency
- Increasing available spectrum
- Increasing the number of cell sites
Martin Cooper has often been credited as one of the pioneer’s of cellular communication. He was the leader of the team that built the first Motorola cellular phone back in 1984. He is also credited with postulating what’s called the Cooper’s Law that specifies that every 2.5 years or so total wireless capacity in a given geographical area doubles. Based on Cooper’s calculations worldwide wireless capacity had increased 1 Trillion (10¹²) since 1901 when Marconi had his first Trans-Atlantic transmission. Cooper thinks in the last 45 years, capacity increase was approximately 1 Million times. He attributes the majority of the increase to the increase in spatial re-use of the available spectrum, i.e., reducing the cell-sizes and increasing the re-use factor. Following diagram shows the distribution of the factor of increase.
Spectrum efficiency increases are relatively costly. As we argued on an earlier post, ultimately such increase is bounded by Shannon’s law. Multiple techniques including more efficient modulation, coding, error correction, compression, multiplexing techniques all translate into increasing the Signal to Noise Ratio (SNR) and bringing it closer to the Shannon limit. Following figure is from 3GPP TR 36.942 that shows the relationship between various modulation and coding combinations with respect to the Shannon limit.
16 QAM and 64 QAM are modulation schemes that are used in HSPA+, LTE and WiMax to increase the channel capacity. Furthermore, all of these systems use adapt the modulation and coding scheme based on the channel characteristics. The result is the following graph from the same 3GPP document that shows the possible spectrum efficiency of a Single Input Single Output (SISO) system which can be approximated as 3/4 of the Shannon limit.
In other words, the highest modulation/coding scheme (in the arsenal of current 3.5/4G systems) is limited to about 4.8 bit/s/Hz when the SNR exceeds 19 dB. In order to achieve the peak speeds exceeding 1 Gb/s for 4G, systems designers must rely on the concept of Multiple Input Multiple Output (MIMO) along with bonding multiple already-wide channels together to use channels as wide as 100 MHz.
A recent report commissioned by Ofcom (UK regulator for the communications industry) to Real Wireless paints a very detailed picture of the impact of technology (so called spectrum efficiency) advances in the overall capacity. It is a fascinating read of 120 pages of diagrams, charts as well as 95 pages of appendices detailing the survey and research methodology behind it. I would recommend everyone who is interested in the future of the wireless technology to read it or at least go over their major findings. The report was written for the UK market and its timelines reflect its market realities: LTE spectrum auctions haven’t been conducted yet, likely LTE deployments will be in 2012 and only in 2013 some volume of LTE terminals will be in the UK market. Nevertheless, the report includes a valuable compilation of observations that are valid across the world:
- Spectrum efficiency improvements due to LTE and LTE-Advanced will not be able to keep up with the projected traffic demand. By the end of the decade spectrum efficiency is expected to grow roughly 5.5 times which translates into a yearly average growth of 18.5%.
- Using a more pessimistic assumption based on the traffic profile mix and the needed head-room for real network deployments as opposed to full loading the actual gain may be even smaller (as low as 3 times the capacity of 2010).
- Dense urban environment deployments cannot be served by the combination of spectrum efficiency and available spectrum increases. Increasing the number of cell sites in the form of small cells is the only option.
- Cost of deploying these small cells is not well known. Since the demand is directly impacted by the availability of the capacity, it is not known how the unit cost of providing service due to increases in cell sites would translate into impacting the demand itself.
- Primary factor for the increase in spectrum efficiency is the use of MIMO which happens to be more effective in in-door environments and where the effective cell size happens to be small. On the other hand deploying multi-antenna base stations is a significant challenge in in-door, small cell deployments. For example, report predicts that even in 2020 only 5% of the UK sites will have 8 antennas whereas 50% will have 4 antennas and 45% will be with dual antennas.
Coming back to our title, the fundamental decision a mobile network executive will be making is how to balance the investment between the additional spectrum and the new cell sites. We believe when it comes to adding new spectrum there are two complementary paths:
- Invest money in licensed spectrum: Obvious strategies are purchases of new available spectrum, mergers or acquisitions of assets of competitors.
- Invest money in off-loading using unlicensed spectrum: Increase the level of traffic off-load strategies using WiFi.
Licensed spectrum prices varies significantly from country to country. Nevertheless in every country underlying primary metric is Currency/MHz-POP where MHz-POP is the population-weighted spectrum. Some recent transactions to establish pricing for licensed spectrum are:
- AT&T’s purchase of T-Mobile for $39B. If we assume AT&T is purchasing T-Mobile for its spectrum only, then the price per MHz-POP comes to $2.70. Compared to 700 MHz auctions where the per MHz-POP price was closer to $1.40 (for medium and large size markets), AT&T paid almost double. However, considering the infrastructure and people who will join AT&T as part of the transaction as well as 33M subscribers who will be joining the customer base of AT&T, it doesn’t look like AT&T deviated much from the 700 MHz auction pricing.
- Recent estimate from French Ministry is 1.8B Euro ($2.55B) for 30+30 MHz in 800 MHz band. Considering French population is roughly 63M, spectrum price comes down to around $0.68/MHz-POP. This is significantly lower than US prices and it may be reflecting the more cautious attitude of mainland Europe towards growth in mobile data services, especially after the exuberance of 3G spectrum auctions a decade ago.
For unlicensed spectrum cost model is entirely dependent on the network implementation cost which is comparable to small cells of licensed spectrum, especially in very dense in-door environments where providing the necessary capacity is not possible with traditional dense-urban macro-cell or micro-cell deployments.
For example, Real Wireless/Ofcom report identifies King’s Cross train station in London as a case study. Primarily due to the customer density (over 1 Million people per square-km) the expected traffic load in King’s Cross is expected to be over 27 times of typical dense-urban deployment. Considering the station is roughly 8000 square-meters, it can be covered by deploying 6-7 pico/femto cells (single-sector) with Inter-Site Distance of 60 meters. Similar area can be covered with a very dense 802.11n (for the sake of apples-to-apples comparison with 4G systems relying on MIMO and very large bands) deployment using 5 GHz spectrum. Following experiences of companies such as Xirrus, Meru, Ruckus, it seems to be possible to reach comparable traffic densities using WiFi if AP density is at least one order of magnitude higher.
As noted in the Real Wireless/Ofcom report, the first order of business is to calculate the true cost of dense deployments of licensed spectrum technology along with the projected cost of spectrum to calculate the unit cost for wireless data services. Only this way, we will be able to understand if the industry is projected to have significant bottlenecks in addressing user demand in hot spot areas. Certainly such analysis is independent of the natural progress of 4G deployments where a large portion of the coverage area will continue to be successfully covered by macro or micro sites.
Next natural step would be to compare the unit cost for hotspot locations against the cost for the use of unlicensed technology (WiFi). Last 8-10 years have shown that WiFi is the superior choice even with the very limited 2.4 GHz band. We estimate 30-40 % of wireless traffic is already carried over WiFi in developed world where the wired infrastructure is more readily available. Majority of this offload is at home and this traffic grows to be a larger segment of overall traffic. Real Wireless/Ofcom report predicts at-home use will reach 58% of total traffic by 2013. When the office use is added, the total opportunity goes up to 85% of overall traffic.
In summary, choice for the wireless network executive is not a simple bifurcation between spectrum and additional cell-sites. Instead a multi-pronged approach is the advisable path:
- Deploy the technology advances (spectrum efficiency)
- Make spectrum purchases to plan for traditional macro, micro cell deployments for dense urban, urban and rural coverage
- Identify hotspots (those 3-4% of sites that will carry 30-40% of total network traffic) and find ways to use dense WiFi deployments to off-load traffic
- Work with device manufacturers to promote the adoption of higher orders of MIMO for 802.11n and the use of 5 GHz band
- Deploy core network technologies such as IP Flow Mobility (IFOM) to promote the use of WiFi at home and office to off-load traffic from the radio access network while keeping a close tab on the service not to be by-passed by over-the-top service providers