Multi-point wireless systems offer much promise for backhaul.  An operator can deploy a single central unit (CU) and the connect end sites as they are required.  The installation of additional sites is very simple, as an alignment is not required, and only one side of the connection needs to be visited.  This is especially attractive in high-density small cell deployments, where installation time and cost are critical, and the location of sites is very dynamic.

However, the potential gains and benefits of multi-point backhaul come about primarily in ideal conditions that are very difficult to achieve in actual deployments.  The two distinct types of multi-point backhaul that need to be considered are unlicensed (sub-6 GHz) and licensed (24-42 GHz).

In the unlicensed bands, primarily 5.8 GHz is used for multi-point systems.  The benefit of this band for multi-point is that non line-of-sight connections are achievable, so that, in the right conditions, a large number of sites can be backhauled from a single hub site. However, there is limited spectrum in this band, typically <50 MHz, limiting the total capacity that would be available.  This limitation means the maximum available capacity would be about 400Mbps, and that would need to be split across multiple sites.  This capacity would decrease if there is interference in the band, or if line of sight blockage drives lower modulations to be utilized.  If 200Mbps is needed per site, then a maximum of two sites could be aggregated per sector, and in many cases it may only be a single site. 

The unlicensed nature of the 5.8GHz bands results in unpredictable link availability for each of the end sites, which may often not be suitable for mobile backhaul.  Lastly, the 5.8 GHz band is a TDD band. When this is combined with multi-point, and interference, delays can be 5-10ms, making it unsuitable for LTE backhaul.  These capacity, availability, and latency limitations are not evident in lab trials and can only be fully understood in a street level urban deployment, where interference, spectrum congestion and line-of-sight blockage are present.  Moreover, the limitations vary between deployments and cannot be anticipated.  In fact, they will change significantly throughout the life of a deployment, making the delivered service very unpredictable and unreliable.

In the licensed bands, typically from 24-42 GHz, many countries have area licenses available with a large amount of spectrum.  Due to the licensed nature, services deployed in these bands are much more predictable and have high availability.  Also, with the large amount of spectrum, higher capacity services are also feasible.  Although, as with all multi-point systems, when more end point sites are deployed, lower capacity per site is available.  If 100MHz were deployed on each site, 800Mbps-1 Gbps could be available for the sector, thus enabling up to 200 Mbps per site if five sites are on that sector. 

The major hurdle with licensed multi-point for small cells is that line-of-sight is required.  This is very difficult to achieve in an urban environment.  Although it may be possible to find high mounting locations that have line-of-sight to one or two street level locations, there are almost no locations that will consistently have line-of-site to more than two desirable street-level small cells.  When a multi-point system is limited to only one to two end points, the economics of the central unit become very detrimental to the business case.  In addition, a high number of central units will be required, and that inefficient use of spectrum will start to reduce the capacity per site that is available.  Licensed multi-point systems also introduce higher delay, typically one ms or above, which may often make them unsuitable for LTE advanced services, and very likely unsuitable for future 5G services.

Although multi-point systems have some very attractive characteristics for small cell deployments, especially in network studies and controlled trials, many of those benefits erode when being used in an actual urban environment.  It is important to thoroughly test these systems, and understand the performance limitations and resulting economics in true street-level environments before building a network based on them.