Books I have co-authored

5G in mmWave bands – good enough to render fiber obsolete?

 

There is an abundance of spectrum in the millimeter
Wave bands, but offering “unlimited bandwidth” in 5G still
requires expensive backhaul (and fiber)

The mobile industry has set very aggressive targets for 5G such as: 1-10Gbps delivered connection speed (not theoretical maximum), and 1000x bandwidth per unit area (see my previous article). What will make this possible is the ability to use spectrum in the millimeter wave bands (30-300 GHz), and the EHF band (30-100 GHz) in particular.

Several factors have contributed to making mmWave bands available for mobile services. The RF technology for components that can operate in these extremely high frequency bands has matured and is ready for mainstream commercial use.

Millimeter waves behave differently from radio signals and their propagation resembles visible light more than traditional radio. For example, it is almost impossible for mmWaves to penetrate solid objects such as walls. Even the human body, foliage, or glass windows with curtains pose significant problems. Another difference is the signal attenuation by the atmosphere and rainfall, which is higher for mmWaves compared to traditional radio signals. It was previously thought that millimeter waves could only be used with direct line of sight (LoS), and this limitation made the frequency band rather useless for wireless communication. However, recent field trials have shown that mmWave signals bounce (are reflected) on hard surfaces in an effective way. This makes it possible to achieve good NLoS outdoor coverage of at least around 200 meters in an urban environment.

The bouncing creates a lot of multipath propagation but this problem can be overcome with high performance signal processors. The limited reach and air attenuation is actually an advantage if mmWaves are deployed in ultra-dense networks. It will make it easier to reuse the same spectrum if nearby access nodes do not interfere with each other.

In addition, there is an abundance of spectrum in the mmWave band. In the traditional radio spectrum (0-3 GHz), the mobile industry has until now managed to deliver their services to 4.5 billion people on less than 0.6 GHz of allocated spectrum. If only 5% of the mmWave band (30-300 GHz) was allocated to mobile services, the available spectrum would increase by a factor 25. This abundance could be used for extremely wide carriers that would enable multi Gbit speed bandwidth.


In the mobile industry’s vision, 5G will provide up to 10 Gbps bandwidth in urban areas with ultra-low latency, almost 100% availability, and seamless fallback to 4G when coverage falters. If and when these targets are achieved, it is claimed that 5G mobile could be a serious competitor to landline fiber access. More outlandish claims are that “everything will be mobile and the fiber network will be abandoned”. I am sceptical for several reasons.

As far as I know, it is still unclear whether mmWave signals can penetrate buildings from an outdoor access point. Can the signal go through a window? Will it be blocked if curtains are drawn or blinds are pulled down? Or will the consumer be expected to mount a 5G repeater unit in his or her window, or outside the window? What if a cat walks in front of the mmWave transceiver on the window sill? These are potential inconveniences and could be a barrier for mainstream market adoption.

If mmWave access nodes are deployed indoors, the wireless backhaul from the indoor 5G nodes must have the capability to penetrate walls. This would require a narrow beamforming antenna (array or horn antenna) using lower frequencies than the mmWave bands. This antenna can be aimed toward a nearby target node with a connection to the core network. If the access nodes are self-deployed by the users, an array antenna could possibly self-configure and form a beam in the right direction. However, I don’t know if an array antenna can form a beam in any 3D direction. The user might have to manually point the antenna in the right direction. Alternatively, the access node could be equipped with a servo that aligns an internal antenna. Health concerns regarding radiation could be another barrier to adoption as the narrow beam from the access point will generate a strong RF signal that passes through the user’s home. (It is of course possible to connect an indoor 5G access node to the existing in-building fiber directly, but that can hardly be called, “replacing fiber with 5G”.)

Indoor use of 5G based on mmWaves face additional challenges. The signal can propagate between rooms by bouncing off the walls but a closed door will most likely cut the connection. Seamless fallback to Wi-Fi and/or LTE can manage this situation but these legacy technologies will not be able to deliver the same performance as mmWaves.

The demands for fast processing in 5G will be extreme. Higher data rates and lower latency require faster processors. The critical building block for the 5G mmWave technology is narrowbeam MIMO antennas, and they rely heavily on signal processing. In a scenario with extreme data rates and extreme user density, the requirements for processor performance will exceed what today’s processors can deliver. Moore’s Law has already slowed down significantly, and if it comes to a halt, processor capacity could be a barrier for the 5G vision. I have spoken with experts whose assessment is that processor technology will not be able to fully meet the requirements for 5G until 2022 or 2023.

processor technology will not be able to fully meet the requirements for 5G until 2022 or 2023


The costs of deploying ultra-dense 5G networks will be substantial. If the range of an access node is limited to a few hundred meters, thousands of access nodes will have to be deployed in urban areas. The major cost driver is not the electronics but installation, cabling, power and maintenance. The traffic from each access node will have to be backhauled into the core network. Probably with Point to Point narrowbeam links which have to be aimed towards a nearby access node that is connected to the core network (via fiber or further mmWave links).

Even though 5G has a fantastic best case performance, the first deployments will be underprovisioned. The number of access nodes will initially be insufficient and when peak traffic exceeds capacity, 5G networks will suffer from the same type of service degradation as 3G and 4G networks. The operators have a limited investment budget, and they will most likely settle for a slower and less costly rollout of 5G. Users who expect their wireless 5G to replace fiber will be disappointed.

In addition, new civil engineering technologies are reducing the costs of deploying fiber in street ducts. 5G is still at least five years away from a broad market launch and, at that time, the fiber networks will have a much larger user base than they do currently. When fiber service providers begin to face competition from 5G, they will of course lower their prices and offer higher bandwidth to stay competitive. The bandwidth that 5G will deliver in a decade should be compared to what fiber can deliver at that time, not with fiber capacity today.


In a competitive market, I don’t see 5G as a viable direct alternative to fiber for at least a decade. But in markets with slow-moving monopolistic landline incumbents, 5G could offer an attractive alternative. The same goes for markets where the landline network is underdeveloped or non-existent.

Update: A rewritten version of this article has been published on Broadband London.

Comments are closed.