The future of microwave planning
Networks are getting denser and, at the same time, higher capacity and longer hop lengths are continuously pursued. Modernized microwave planning and more efficient spectrum use will be significant in achieving this.
Microwave technology is evolving continuously to meet new requirements set by the latest generations of radio access technologies and new use cases. As capacity boundaries are pushed further, it is important to maintain sustainable requirements which involves more balanced dimensioning. Additionally, access to more spectrum, wider channels and high spectrum efficiency are key for achieving higher capacity. More aggressive frequency reuse combined with interference management is a way to improve spectrum efficiency.
Benefits of more balanced dimensioning
Traditional microwave planning methods date back to the introduction of circuit-switched 2G (GSM) networks, which were mainly deployed to provide voice services to mobile users. The first generation of microwave links used a single, fixed modulation, and it was therefore natural to plan the links based on strict availability targets. This meant that a link should have a high likelihood of supporting the voice services it carried – some 99.999 percent of the time – since if the link was down, then voice services were also down. With the adoption of packet-switched networks for data in 3G, together with adaptive coding and modulation (ACM) in high-capacity microwave links, the need for such extreme availability for all ACM levels became unnecessary. Instead, differentiated availability with different availability targets for different ACM levels became more significant.
In practice, differentiated availability means that lower modulation levels have higher availability targets than the higher modulation levels. A high availability target (for example, 99.99x percent) on a committed information rate (CIR) ensures that services like voice and other high-priority service and control operations are guaranteed, while a lower availability target on peak information rate (PIR) allows for traffic peaks that occur more rarely. It is also not uncommon for it to be the radio access network (RAN) that limits the bitrate and the user experience due to phenomena such as radio channel fading, shadowing and interference.
A good strategy, therefore, is to dimension the microwave backhaul in balance with the RAN traffic it carries to avoid unnecessary overprovisioning of the backhaul. Excessively strict availability targets limit the possible use cases of microwave backhaul links. By contrast, a well-balanced dimensioning opens several possibilities, including longer hop lengths, higher capacity, energy savings and use of higher sub-THz frequencies.
As a step toward a more balanced backhaul dimensioning, ETSI ISG mWT has defined a new KPI called backhaul traffic availability (BTA). [1] Put simply, BTA is defined as the probability that the backhaul link capacity supersedes the RAN traffic it carries, which is the same as the probability that the link is uncongested. The BTA thus depends on the probability distributions of the backhaul link capacity and the RAN traffic. BTA applies to all frequency bands, but it is especially interesting for E-band and future sub-THz links since it leads to higher capacity over longer hop lengths, both in single- and multi-band configurations. The need for higher capacity over longer hop lengths is demonstrated by the growing deployments of multi-band links.
On the topic of KPIs, standardized ITU-T definitions of user QoE also exist for different services. QoE is defined by a service-specific function of user rates and delays in the user downlink and uplink, and are typically represented by a QoE rating ranging from 1 to 5, where, for example, 4.5–5.0 is excellent, 3.5–4.5 is good, 2.5–3.5 is fair, and so on.
Balanced dimensioning in action
To illustrate the benefits of more balanced backhaul dimensioning, a simulation was conducted of a 5G RAN in the 3.5 GHz band serving mixed traffic types like video-on-demand, web-browsing, live streaming, cloud gaming and augmented reality (AR). The RAN comprised three sites with three sectors each, with the aggregated traffic from all sectors being transported over a multi-band backhaul link combining E-band and 18 GHz band in the Gothenburg, Sweden, rain zone. The traffic load in the RAN varied between low (20 percent), medium (50 percent) and high (70 percent) utilization, where utilization is set by the number of users within each user type, with the AR users being the most resource demanding.
Figure 8 shows the result from the simulation investigating the effect of BTA on QoE of AR users and the maximum possible hop length of a multi-band backhaul link. The left y-axis shows the percentage of ideal QoE for AR users. Ideal QoE means the maximum possible QoE attained by using an ideal backhaul that does not have any impact on user QoE. Ideal backhaul can be interpreted as a backhaul with infinite capacity and 100 percent availability. The ideal QoE, therefore, only depends on the performance of the RAN and not the backhaul. The blue curve represents how a large fraction of the ideal QoE is attained by using a multi-band backhaul link instead of an ideal backhaul.
Figure 8: Applying the new backhaul traffic availability (BTA) KPI in microwave planning
Doubling of hop distances is achieved by the introduction of new KPI.
For example, 99 percent of the ideal QoE for AR users implies that the QoE for AR users is 99 percent of the maximum possible QoE attained by using an ideal backhaul. It can be argued that this is an insignificant reduction from the maximum QoE and that the AR users will not experience any negative impact during their sessions. The right y-axis shows the maximum hop length of the multi-band backhaul link as a function of BTA for the different RAN traffic load levels. The x-axis is common for all curves and represents the BTA as defined by ETSI ISG mWT. The lower the BTA, the higher the likelihood of congestion in the backhaul, and user QoE reduces correspondingly. However, if a lower BTA and a correspondingly lower QoE is accepted, then the maximum hop length can be increased. This is illustrated by some of the operating points in Figure 8, where there is a clear connection between QoE, BTA and hop length. Take, for example 99.9 percent of ideal QoE, which in the simulated case, corresponds to a BTA of around 99.98 percent at high load and is attained by a maximum hop length of 12.5 km (which is illustrated by the purple arrows). If 99 percent of the ideal QoE (corresponding to 99.77 percent in BTA) is accepted instead, the maximum hop length can be more than doubled, to 27.5 km (which is illustrated by the green arrows). In the ETSI Group Report, it is indicated that the optimum range for BTA values is between 99.5 percent and 99.9 percent, but it is also recognized that the final choice is up to the preferences of individual service providers and the needs of specific services.
It is important to emphasize that more balanced dimensioning does not imply an increased risk of outage of critical services, since the CIR is still associated with a high availability target like 99.99 percent to 99.999 percent. It is rather that the PIR is associated with a more balanced availability requirement that matches the BTA.
When it comes to multi-band backhaul links like the one used in the simulation example, CIR is provided by the low-band link while the PIR is provided by the E-band link. It means that relaxing the PIR availability requirement of the E-band link only has a very minor, if any, negative effect on user QoE while effectively resulting in more than two times longer hop length.
Opportunity with interference management
Modern microwave links are very spectrally efficient, meaning that they provide many bits per second per Hertz (bps/Hz). High spectral efficiency is ensured by successfully employing techniques like high-order modulation, high-performance dual-polarized antennas (XPIC) and multiple-input, multiple-output (MIMO).
Microwave links also operate in regimes with high signal-to-noise ratio (SNR) thanks to efficient power amplifiers, high-gain antennas and high receiver sensitivity. This means microwave links are what is known as bandwidth-limited, meaning that their capacity is more limited by the spectrum bandwidth than by their SNR. Capacity grows linearly with bandwidth but only logarithmically with SNR, which implies that it is more spectral-efficient to try and increase the bandwidth instead of SNR when the link is in the high-SNR regime. This is exactly what universal frequency reuse – or frequency reuse one (FR1) – sets out to achieve. In FR1, all (or at least a majority of) links use the same frequency channel to allow wider channels to all links. For example, instead of allocating four neighboring links a separate 28 MHz channel each, they can all use the same 112 MHz channel which increases their possible peak rate by a factor of four. The downside of using the same frequency channel is increased interference between the links. However, since the links are in the bandwidth-limited regime, the upside of more bandwidth is much larger than the downside of increased interference. As with balanced dimensioning and BTA, a more relaxed availability requirement on peak rate has little, if any, impact on user QoE.
Local traffic-aware transmit power control is an efficient way to limit interference between links in an FR1 network. Traffic varies over both space and time across the network, and local traffic-aware power control continuously adapts the transmission power of each link to the minimum power needed to serve its immediate traffic needs. This way, unnecessary interference is avoided in the network compared to using a fixed output power or some other traffic-unaware power control. Previous Microwave Outlook reports have shown the large benefits of this type of local traffic-aware power control in FR1 networks.
However, if the interference between neighboring links is large, then the links may start competing for capacity by raising their transmission powers in an uncontrolled manner, and such situations need to be avoided. One simple way of avoiding power rushes is to consider this issue in the initial network planning phase, for example, by setting caps on the maximum permitted transmission power of each link in the network based on interference models. Another, more sophisticated way is to employ interference management by using centralized power control combined with traffic prioritization. For example, if multiple interfering links are competing for capacity but one of them carries higher priority traffic than the others, then a centralized controller can allocate or schedule more capacity to the high-priority link. Many links also have a natural isolation between them, which limits the interference and the need for centralized interference management. This isolation is mainly provided by the narrow-beam antennas with low side-lobe levels.
To illustrate isolation between links and how many links may need centralized interference management, the use of FR1 was simulated in a real backhaul network with a very dense deployment of microwave links. The simulation was used to investigate how to divide the complete network into subnetworks based on interference levels. The principle was simple: if two or more links interfered with each other over a predefined interference-to-noise (I/N) threshold, then the links were allocated to the same multi-link subnetwork. If a link was not transmitting/receiving too much interference to/from other links, then it could operate independently of all the other links in a single-link subnetwork. Centralized power control and interference management is only required when links in the multi-link subnetworks start to compete for capacity.
Figure 9: Example of a real network with 15 GHz radio links divided into subnetworks to enable wider channels
Interference management enables four times wider channels (from 28 to 112 MHz) and up to four times higher peak rate.
Figure 9 shows how a large share of the total links in a dense network, with close to 1,000 links operating at 15 GHz, is allocated to single- or multi-link subnetworks. In this example, the antenna size was 0.6 meters, and the availability target was 99.99 percent at 1024 QAM. Two I/N thresholds were assumed, -6 dB and 0 dB, respectively. The higher the I/N threshold, the higher the number of links that can operate independently in a single-link subnetwork. Meanwhile, a stricter (or lower) I/N threshold means that more links need to be allocated to multi-link subnetworks. To illustrate the effect of using more narrow-beam antennas to provide more isolation between links, the same network deployment was also simulated with E-band links. Figure 10 shows the share of total links allocated to single-link and multi-link subnetworks, respectively. The narrow beam of the E-band antenna provides very good isolation between links, which significantly reduces the need for centralized interference management since only 14 percent and 8 percent (at the two different I/N thresholds) of the total number of links belong to a multi-link network. This can be compared to the 15 GHz network in Figure 9, where the equivalent shares were 83 percent and 65 percent, respectively.
Figure 10: Example of a real network with E-band radio links divided into subnetworks to enable wider channels
The narrow beam of the E-band antenna reduces the need for centralized interference management.