Wireless operators have quietly embarked on acquiring dark fiber – a trend that was picked up in 2013 and accelerated in 2014. Verizon is leading this activity and rolling out dark fiber to their cell site. But why specifically have carriers decided on dark fiber rather than continue lease capacity?
I think the answer has to be looked at from the perspective of long-term developments which matches the time-frame for leasing dark fiber which ranges between 10 to 20 years, and is typically closer to 20 years. Looking at that time-frame, we note that:
1- There has been significant allocation of spectrum to mobile services and it’s not uncommon for carriers to hold over 100 MHz of spectrum (example: Sprint has more than 120 MHz of 2.5 GHz spectrum alone, and EE has 210 MHz in the UK).
2- New spectrum allocation in the future will be in bands that hold wide channel allocations. For example, WRC15 bands include 2.7-2.9 GHz and 3.4 – 4.2 GHz (C-Band).
3- Operators are refarming 2G and 3G spectrum in favor of LTE and actively moving subscribers to the more efficient network.
4- LTE-Advanced aggregates up to three downlink channels of 20 MHz today and can theoretically support up to 5 x 20 MHz channels for a total of 100 MHz. SKT in Korea has already deployed 3-carrier aggregation with a combined 40 MHz bandwidth to achieve 300 Mbps in peak throughput. Moreover, many carriers moving to deploy 4×2 systems (example, T-Mobile; in China TD-LTE is used China Mobile deploys 8-antenna systems).
Based on peak downlink capacity of 150 Mbps for a single 2×2 20 MHz LTE channel, it is easy to see that for a single three-sectored macro cell backhaul capacity requirements exceed 1 Gbps+. When capacity becomes on the order of Gbps, alternative options dwindle:
1- Leased capacity: it becomes harder to procure higher capacity on lease basis and when available, it would be expensive.
2- Wireless backhaul: wireless is traditionally a good alternative or is complementary to fiber. The widest bandwidth for microwave is 2 x 56 MHz. Capacity is now achieved through higher modulation scheme such as 2048 QAM which limits range, or XPIC (cross-polarization interference canceler) which doubles the transceivers, or MIMO which requires multiple antennas and transceivers. This is almost a linear cost-performance curve: to double capacity, you almost double the cost. On the other hand, millimeter wave systems have limited range.
Dark fiber is not cheap. Operators would typically lease dark fiber in long-term indefeasible right of use (IRU) contracts (e.g. over 20 years). The IRU can be treated as a capital expenditure and depreciated on an advantageous schedule. It is not uncommon to pay $150 – $300 per month per mile per strand of fiber in urban areas. But on the other hand, operators can use any optical technology and to upgrade capacity at any time of their choosing per their need. This is powerful not only for the freedom it provides but also because it allows operators more control over the quality of service. But there are also other reasons.
Dark fiber allows operators to change the deployment architecture and centralize base stations in data or fiber centers. Operators in Korea and Japan already did that. While the economics there work better than in North America and Europe, there are clear opex savings to realize in network operation and maintenance expense reduction. In the future, capex savings can follow when NFV is implemented in the radio access network to pool baseband resources. Research shows that as much as 75% in processing power can be saved through baseband pooling because it would be no longer necessary to provision for the peak capacity on a per site basis. Centralized RAN architectures cannot work without dark fiber. With CPRI rate per a single 2×2 20 MHz LTE channel standing at about 2.5 Gbps, fiber will be required to run the 10’s if not 100’s of Gbps expected from such deployment scenario.
Dark fiber provides additional advantages centered on performance improvement related to capacity enhancement and quality of service. This particularly applies to HetNets. Small cells have proved too prone to interference and their performance in capacity hotspot applications range from mediocre to outright damaging to the overall network performance. Interference coordination techniques require low-latency, tight jitter and high-capacity transport links which dark fiber provides. In fact, I think it will be immensely difficult to implement HetNets without a fiber backbone.
Dark fiber is a strategic asset: it is the spine of wireless networks. Operators who own fiber have a high competitive advantage in keeping up with the performance development in access networks.
+ with combined uplink and downlink, peak capacity is 200 Mbps per channel. With 6 channels per site, that’s 1.2 Gbps. There is additional overhead that increases the total. On the other hand, statistical multiplexing may be factored in to reduce backhaul capacity requirements.