Skip navigation
Reducing power consumption in microwave networks

Reducing power consumption in microwave networks

How to reduce power consumption in microwave networks

Energy efficiency in mobile networks is key to reducing operational costs and reaching net zero sustainability goals.

Energy per transported bit has steadily decreased during the last two decades. Both existing and new energy efficiency functions, together with the introduction of AI, provides new and efficient tools to continue this evolution.

The energy efficiency of microwave networks, measured as power consumption over transmitted capacity (W/Mbps), has improved significantly over time. This has largely been due to increased modulation schemes, wider frequency channels, and the introduction of carrier aggregation leading to higher transmitted capacity per single radio. In this way, new microwave equipment has been able to support the transitions from 3G to 4G and to 5G mobile telecom standards. Over the last two decades, the energy efficiency of microwave radios in traditional bands (6–42 GHz), has improved by a factor of 10 (Figure 10).

The introduction of 70/80 GHz E-band radios and up to 2 GHz channels accelerated this evolution with up to a factor of 5. This has resulted in an energy efficiency increase of more than a factor of 50 over the last 2 decades.

50x

The energy efficiency of microwave radios has increased by a factor of 50 over the last 2 decades.

Power consumption evolution of 15 and 70/80 GHz nodes

Figure 10: Power consumption evolution of 15 and 70/80 GHz nodes

Reducing power consumption through traffic-aware power functionality

The traffic in mobile networks varies over time and peak rates are, in practice, often only reached during shorter time periods and not during the full 24-hour day. A microwave link could thus be fully utilized during busy hours while being almost unused during low traffic periods. By being aware of traffic demands, the radio unit can reduce power consumption by dynamically adjusting the modulation scheme and output power to meet current capacity needs. This can momentarily reduce power consumption up to 30 percent. For a complete wireless backhaul system including modem, power supply and traffic processing units this would correspond to energy savings of approximately 10 percent. Figure 11 shows an example of how the power consumption of a radio unit could vary over time when adapting the link capacity to traffic demand.

Traffic-aware output power function

Figure 11: Traffic-aware output power function

To cope with increasing traffic in mobile networks, more and more microwave radio links evolve from single carrier 1+0 configurations into bonded systems of 2+0 and 4+0 links. However, as before, the extra capacity will frequently not be needed during the full 24-hour cycle. Energy can be saved by placing unused carriers into deep sleep mode during low-traffic hours. With AI and machine learning techniques, the microwave nodes can learn the traffic patterns of each link and optimize the deep sleep periods to match the unique traffic patterns of the site. This makes sure that unused carriers are placed in deep sleep when not needed, but only when there is a negligible risk of causing congestion events.

Introducing AI to take the next step in power saving

Figure 12 shows an example of supported capacity (blue and purple) in a 2+0 link configuration with traffic patterns (green) over a week. In between the busy hours there are “windows” where the capacity of only one radio link is enough to fulfill the traffic demand. The second radio link can be placed in deep sleep in these periods, thereby saving energy.

But how do we decide the optimum deep sleep time periods in a way that is as easy, manageable, and non-intrusive as possible? Manually optimizing the sleep period for each link in a medium-to-large backhaul network (>5000 links) is not feasible in practice.

However, as a starting point, one could specify fixed-time windows for the entire network or parts of the network. In Figure 12 we use a fixed-time window of 8-hour deep sleep during night-time (purple). When comparing this scheme with the AI-optimized time window (blue), we can see that the AI-powered scheme is able to put the microwave link into deep sleep for longer periods than the fixed 8-hour time window. Fixed-time windows could be the starting point for introducing deep sleep functionality in microwave networks and AI technology would be the natural evolution to optimize this.

Radio deep sleep in dual-carrier backhaul links

Figure 12: Radio deep sleep in dual-carrier backhaul links

Energy efficiency gains in live backhaul networks

To understand the real potential of traffic-aware output power and deep sleep functionality we studied a live backhaul network operated by a major European service provider in 2022. The network was recently upgraded from a 1+0 to an XPIC 2+0 (bonded dual carrier using cross-polar interference cancellation) configuration to support the xpected traffic growth with the introduction of 5G. By measuring the utilization of each link, we calculated the time the second carrier was needed. This was defined as the time the utilization of the full 2+0 link was higher than 40 percent. The result is shown in Figure 13.

As network traffic grows over time, we can see that the second carrier is used more and more often. In Figure 13 we used the Ericsson Mobility Report 2022 forecast for year-on-year (YoY) network traffic growth of 27 percent. As shown, the network is well prepared to handle the traffic growth in the coming five years.

Utilization of a second carrier in a dual-carrier system over five years

Figure 13: Utilization of a second carrier in a dual-carrier system over five years

In year 1 around 10 percent of the 2+0 links use the second carrier more than 20 percent of the time. In year 3, approximately 25 percent of the links use the second carrier over 20 percent of the time and in year 5 almost 80 percent of the links use the second carrier for 20 percent of the time. In Figure 14 we calculated the expected energy saving for three different use cases: (I) traffic-aware output power; (II) fixed 8-hour deep sleep combined with traffic-aware output power; and (III) AI-powered deep sleep combined with traffic-aware output power.

In the first case, energy saving from traffic-aware output power save is relatively constant, remaining at around 6 percent even as traffic grows over time.

In the second case traffic-aware output power is combined with fixed 8-hour deep sleep, so that the second carrier is put into deep sleep during night hours, for an 8-hour period. Here, the energy saving was around 12 percent. In fixed-hour deep sleep we have accepted that we may congest the link at night, and therefore energy saving is constant as traffic grows, but in fact the congestion events will be frequent.

In the third case, AI is used to learn the traffic patterns at each unique site, and continuously optimize the deep sleep schemes. We will thereby avoid congestion, but as shown we still outperform the deep sleep case with 8-hour fixed time slots even as traffic increases yearly. In year 5, we still see a 15 percent reduction in energy consumption compared to the case without traffic-aware power functionality.

Energy savings over five years

Figure 14: Energy savings over five years

20%

Up to 20 percent of network energy can be saved over 5 years in our studied network with AI-powered deep-sleep and traffic-aware output power.

To estimate the energy savings for a medium-sized network with similar traffic patterns, we extrapolated the energy savings in our case study to a network with 5,000 2+0 links. The accumulated energy saving over 5 years would be 3 GWh for the traffic-aware output power save case, 6 GWh for fixed 8-hour deep sleep case and 10 GWh for AI-powered deep sleep comparable to a 20 percent reduction in network energy consumption over 5 years. In the first and third cases these savings would be achieved without any impact on the network performance or user experience.

Summary

The energy per transported bit for traditional wireless backhaul has been reduced by a factor of 10 over the last two decades. The introduction of E-band in 2015 added another factor of 5 resulting in an improvement in energy efficiency of more than 50 times when comparing traditional frequency bands in the beginning of the century with high-performance E-band links in the 2020s.

With the introduction of traffic-aware power functionality, for example microwave networks that adapt energy consumption to meet the momentary traffic demand, we can push the energy efficiency curve even further. In a case study together with a service provider, with a medium-sized microwave network, the combination of traffic-aware output power and AI-powered deep sleep functionality for multi-carrier links resulted in an energy saving of 20 percent over five years, with negligible impact on user experience or service availability.

Ericsson Microwave Outlook report

The continued strong growth of 5G puts a focus on E-band, power efficiency and maximizing capacity in spectrum limited networks. To find out more about this and other interesting topic like spectrum and 5G in rural areas, download Ericsson Microwave Outlook 2022.

Explore now

Explore other articles

Sign up for the Ericsson Microwave Outlook report