Designing energy-efficient network products

In this article, we will take a closer look at two factors that impact energy efficiency of telecom networks, continuing our conversation about energy savings in network operations.

Idle power consumption

The first aspect to keep in mind is idle power consumption – the energy used when the base station is in operation but doesn’t transmit any user data. There are several reasons why this specific consumption can get substantial.

Mobile networks are built on the concept of cells, which are limited geographical areas that serve the users currently in the cell. The radio transmission in a cell is supported through base stations that maintain several cells. Cell areas are typically quite small, which leads to fluctuation in the number of active users, as well as the amount of traffic. To account for these changes, an individual base station needs to be dimensioned for peak traffic to enable adequate capacity. However, it results in increased idle energy consumption.

Mobile networks also have another key capability – coverage. Vendors are required to cover sub-urban and rural areas with fewer inhabitants. Due to these factors, there is only a subset of cells that run on very high user data volume capacity, while the majority require modest or small capacity. Although the areas with smaller capacity demands will be covered by base stations with less resources, there is only a limited number of base station configurations possible as capacity is extended by adding transceivers stepwise. With increased hardware integration, those transceivers support more capacity, thus increasing it altogether. This leads to additional over-provisioning, contributing to increased idle energy consumption.

Furthermore, a base station that does not forward any user data still needs to keep a lot of functions in operation, including broadcast channels that serve as a beacon for user equipment. Such functionality depends on the 3GPP standards and cannot be switched off when no user data is transmitted, which causes another uptick in energy used when idle.

Proportional power consumption

Proportional power consumption means that a base station that does not forward user traffic would not consume any energy, and as traffic increases, energy consumption will follow up until max throughput is reached.

There are many challenges in achieving idle consumption proportionality, that span from improved standards for broadcast channels, power-saving features on the radio side, and digital processing.

Arguing about reducing idle energy consumption, we assume that the proportional part (relation) stays the same. It is of course the total energy consumption over the utilization variation that is important.

Most of the base station's power consumption comes from the radio, which implies that it is important to minimize the use of the radiofrequency (RF) transmitter. There are two ways to achieve it:

• Implement power-saving features between times of activity. It means that the RF transmitter uses micro-sleep capabilities as often as possible.
• Implement advanced algorithms that reduce the amount of raw information that needs to be transmitted. This is important as what is sent is subject to a range of algorithms used to secure and optimize the actual transmission, so the raw data to be transmitted can be larger than user data by adding error correction data.

Such radio resource management algorithms are crucial for reducing the raw amount of data. The algorithms typically deal with interference, power control, beamforming, error correction, etc. and are very advanced and require substantial signal processing. These algorithms are key to achieving the performance wanted by the RAN.

This means there is a trade-off between Radio Access Network efficiency and Base station energy consumption for the digital parts as these algorithms themselves require energy, as part of the digital domain of the base station. The portion of the total energy consumed by the digital part of the base station is therefore growing as these radio resource algorithms get increasingly sophisticated.

The computational demand for algorithms scales with different factors. For example, channel estimation & equalization scales with number of antennas and bandwidth, MIMO processing scales with number of antennas and bandwidth, and FFT/IFFT scales with bandwidth. It means that the computation required depends on factors other than user data volumes. Some of these features (like number of antennas) are there to support a higher peak throughput, so there is an indirect relationship to user data volume.

The implication of this is that not only are the algorithms themselves important, but also how efficient they are implemented, how many clock cycles are consumed to implement a certain algorithmic function. That property determines not only the energy needed to execute the algorithms, but also the time it takes to execute them, which can be crucial to be able to iterate algorithms several times within a given time interval.

The algorithms rely on a computational platform that is more or less energy-efficient. With a growing part of the energy going into the digital domain, that computational platform gets more attention. As for the analogue radio parts, we need to focus on energy proportionality also for the digital part. For example, how much energy is used when the base station is not forwarding user traffic, and how the energy scales with growing traffic.

There is now increasing attention to the energy proportionality attributes of the base station baseband platforms, and COTS alternatives to x86 emerge, such as Arm or Nvidia besides the purpose-built platforms.

Besides focusing on idle energy consumption, those computing solutions address energy saving features, such as putting cores to sleep, reducing clock frequency, etc., to achieve better energy proportionality.

If clock cycles could be saved, this would only have an impact on digital energy consumption, and the computing architecture could harvest those savings by utilizing energy-saving features such as sleep-modes.

Areas where Tietoevry Create can help

Tietoevry has a long history in designing signal processing software for the digital parts of a base station and has one of the largest independent pools of research & development resources globally.

Our skills span algorithm definition, recently powered by AI, as well as implementation of Layer 1 and Layer 2 software for a multitude of clients. The key differentiator for all these assignments is the capability to optimize the implementation of such algorithms, thereby contributing to leading energy properties as well as increased performance. There are examples where we have achieved more than 20% performance increase compared to the client baseline.

Tietoevry Create has also long experience in working with energy-saving features in the radio, such as the various sleep-features.

Reach out to Tietoevry to learn how we could assist you in your next product development venture.