Power usage has always been a key consideration for mobile networks. Supplying and securing power (often in remote locations) for base stations, and cooling the heat generated by the equipment has always represented a significant cost. Energy price rises over recent years, as well as wider climate concerns, have only made addressing the challenge of power efficiency an even greater priority to the telecommunications industry.
Base stations are always going to consume power as they have had to radiate the RF signals fundamental to the network. However, the ongoing challenge is to increase efficiency and reduce the power associated with processing the signals. Underscoring this connection the standard metric is to measure this in terms of Watts per megahertz, so how much power is consumed for each MHz of cellular spectrum served.
Moving beyond terrestrial networks, satellite systems operate in a highly constrained environment and as such continuously strive to reduce the size, weight and power (SWaP) consumption. The combination of mobile networks and satellite technology have made the challenge of delivering the maximum possible levels of energy efficiency even more critical. This requirement has come into stark focus with the next generation of regenerative 5G NTN networks, where rather than the satellite just acting as a repeater, the base station itself is in space.
Typically, a satellite is powered by solar panels when it is in sunlight and uses batteries to save some of that power for when it is in darkness. Every Watt that is consumed by the payload requires more solar panels and more batteries which add cost but also add weight, which in turn increases the launch costs. In the vacuum of space cooling is also a significant challenge, as cooling fans don’t work, with heat having to be dissipated through radiation, using heat sinks or powered heat transfer systems. Any increase in energy efficiency can help reduce this cooling requirement and so have knock on benefits in terms of satellite size and weight.
Optimising the 5G physical layer
Within a 5G base station, the physical layer is the largest consumer of processing power, and at AccelerComm we have been working hard to make use of the latest technology to ensure our solutions are as efficient as possible. The use of the latest arm CPU technology brings a significant (up to 2X) improvement over processors optimised for datacentres, and offloading processing for tasks such as LDPC processing to dedicated hardware accelerators increases efficiency further (up to a further 3X).
In our newest 5G physical layer solutions we have been able to make use of a fascinating piece of technology to make some significant new efficiency improvements. Over the last few years many companies have been investing heavily in processor innovations that reduce the power needed by AI, in particular model training. One area of innovation centres around the efficiency of vector multiplication, something that GPUs are very efficient at, hence their use in AI. It turns out that some of the most processor intensive signal processing tasks in a 5G base station use similar vector multiplication operations and so can also benefit from these developments.
The Way Forward
At the recent MWC 2025 and Satellite 2025 Shows, we were demonstrating how we are making use of this technology in our latest products. We use the vector processing engines in the AMD Versal devices, which AMD refer to as AI engines, for some key 5G uplink signal processing tasks. Our studies show that they are 20X more efficient in terms of Watts per megahertz than a traditional CPU carrying out the same task. The Versal devices are available in a space qualified variant so this solution is already deployable in NTN applications.
Of course, it is not as simple as just being able to port software on these devices. Integrating the processing functions together and managing the flow of data at speeds in excess of 1Gbps needed for a commercial 5G base station in space is another layer of challenge, and many of the signal processing algorithms need adapting to the unique demands of NTN. We have already overcome almost all of these challenges.
As we further iterate our physical layer solution it has been interesting to see how technology being developed to solve the challenges of one application, in this case AI, can also address the challenges of another area. For our purposes that has been unlocking the promise of universal connectivity that is 5G NTN by addressing power consumption challenge.
Get in touch to speak to our experts directly