When 5G was initially introduced, users, CSPs, and vendors alike were excited about the potential new use cases and advantages of this revolutionary technology. Discussions quickly turned to gigabit speeds, smart cities, massive machine-type communications, ultra-reliable low latency, and network slicing—innovations expected to transform mobile network capabilities.

Major telecom companies began preparing for the 5G journey as RAN vendors introduced new, revolutionary hardware and software features. While 4G deployments were mainly based on Frequency Division Duplexing (FDD) spectrum bands, 5G specifications required the allocation of more Time Division Duplexed (TDD) spectrum bands. Many of these new TDD spectrum bands, such as the 3.5 GHz band, had previously been used for purposes other than mobile networks in many countries.

While the technology’s fundamentals, advantages, and use cases were widely discussed, the challenges associated with the new spectrum intended for 5G received far less attention.

Mobile spectrum is the most valuable, limited, and costly resource in the telecom industry, and operators are constantly seeking optimal spectrum allocation to increase network capacity and coverage. When we discuss frequency bands, it’s important to understand the differences between TDD and FDD. The primary difference is that with FDD bands, uplink and downlink transmissions occur on separate channels, whereas with TDD, uplink and downlink share the same channel, separated only by time. There are benefits to TDD’s uplink downlink multiplexing. TDD allows the use of various antenna technologies such as massive MIMO and beam forming due to channel reciprocity. It also allows operators to use spectrum more efficiently by providing them with the flexibility to allocate more capacity for downlink (DL) or uplink (UL) based on changing demands.

One major challenge that the incorporation of TDD bands created is its vulnerability to certain types of RF interference, notably interference caused by a phenomenon known as tropospheric ducting.

Tropospheric ducting occurs when a layer of warm air sits above a layer of cooler air. This weather phenomenon creates variations in air temperature and moisture content, resulting in a layered atmospheric structure that behaves like a waveguide. As a result, electromagnetic waves can be refracted, allowing them to travel back toward the Earth and propagate over long distances, particularly affecting all the radio signals with frequencies above 50 MHz.

5G and LTE TDD systems are especially susceptible to ducting interference when the propagation time between distant cells exceeds the guard period separating downlink and uplink transmissions. Downlink signals from distant cells, including those from neighboring markets, can travel more than 200 km and interfere with uplink signals, causing TDD self-interference due to ducting.

Ducting interference is a global issue affecting telecommunications networks today. In regions like North America, Asia-Pacific, and the Middle East, it accounts for more than half of the interference impacting operators’ networks, often affecting thousands of cells simultaneously. Ducting interference can lead to a substantial decrease in data throughput, ranging from ~50% to ~90%, and can also degrade audio quality, resulting in customer dissatisfaction and complaints.

Since tropospheric ducting is heavily influenced by weather conditions, it tends to vary seasonally and can be very dynamic. This adds complexity for operators since ducting interference can appear or dissipate quickly, even within the same day. A near real-time interference mitigation platform has become more critical than ever for operators to identify when ducting interference occurs or subsides. More importantly, it helps quantify the impact on network quality and service assurance, while providing actionable solutions to efficiently resolve the issue.

Today, operators often resort to manual interventions or temporary changes in network configuration to mitigate TDD interference, particularly when dealing with tropospheric ducting. These reactive measures are far from ideal, as they are resource-intensive and often fail to keep pace with the dynamic nature of ducting interference. Manual adjustments can also be inconsistent, leading to delays in detecting and resolving problems. Furthermore, operators often depend on RAN vendor software functionalities, which are limited in capability and may not support multi-vendor environments. This results in reduced capacity and a lack of visibility into interference issues.

What operators need is an automated, proactive approach that continuously monitors the network environment, evaluates interference levels, and quickly optimizes performance. This is where Spectrum-NET comes into play. It utilizes advanced AIOps models to continuously enhance the network through a ‘Detect, Analyze, Mitigate’ framework, offering real-time visibility and monitoring across multi-vendor networks.

Mitigation of ducting interference is achieved through parameter reconfiguration. Commonly used techniques include:

• PDSCH blanking at dominant aggressor cells.
• Adjusting the special slot format parameter.
• Reducing victim cells’ footprint and increasing UE transmit power.
• Increasing antenna tilt for both victim and aggressor cells.

When we talk about ducting interference, there are two types of cells involved: aggressor and victim cells. Aggressor cells are those causing the most ducting interference due to factors such as transmitted power, location, and traffic volume. Victim cells are the ones that experience heightened levels of interference and KPI degradation. It’s important to note that any aggressor cell can also be a victim cell.

Mitigating ducting interference is not a simple task, and it often requires a trade-off between reducing interference and maintaining network capacity and coverage. For example, blanking the required PDSCH slots at aggressor cells to eliminate interference at victim cells generally reduces downlink aggressor cell capacity by about 6% to 19%, depending on the number of blanked PDSCH slots. Applying this mitigation approach uniformly across potentially thousands of affected cells can lead to significant reductions in overall network capacity, resulting in issues like decreased network speed and possible congestion.

Therefore, mitigation efforts should target the dominant aggressor cells, as they can provide the most substantial interference reduction for victim cells, ultimately maximizing overall network capacity. One of the key features of Spectrum-NET is its ability to identify these dominant aggressor cells, based on average interference levels and Downlink Physical Resource Block (DL PRB) utilization observed when ducting was detected. By focusing mitigation actions on these dominant aggressor cells, we achieve the greatest interference reduction for victim cells while maximizing network capacity. For example, dominant aggressor cells may account for just over 20% of all aggressor cells, yet increasing the guard time in these cells alone can achieve an 80% overall reduction in interference for all victim cells within the affected market.

Ducting interference can occur not only within an operator’s network but also between different operators. This interference may arise from varying frequency allocations between counties, states, or between operators in neighboring countries—even when their cells are aligned in frequency and synchronized. The latter example, known as cross-border ducting interference, can impact both TDD and FDD cells. To effectively mitigate cross-border ducting interference, cooperation between operators and regulator entities in the involved countries is essential. At Spectrum Effect, we have developed the “QoE Focused Auto-Mitigation feature,” which uses multiple patented techniques applied to the victim cells (considering the aggressor cells are outside of the network) to reduce the impact of persistent cross-border interference on key performance indicators (KPIs) and enhance subscribers’ quality of experience (QoE) through parameter adjustments.

While TDD offers significant advantages for 5G, it also introduces new challenges, particularly ducting interference that can degrade network performance. To address this, a centralized, automated, multi-vendor solution like Spectrum-NET is essential for effective mitigation. It transitions operators from reactive, manual interventions to a proactive approach that continuously monitors and optimizes network performance through AIOps models. By focusing on dominant aggressor cells for self-interference and adjusting victim cells for external ducting interference, Spectrum-NET ensures timely and effective measures are taken, maximizing spectrum usage and enhancing service quality—and doing so in multi-vendor environments. By leveraging Spectrum-NET’s capabilities, operators can increase network capacity, reduce churn, and improve overall QoE, making it an indispensable tool for overcoming ducting interference in modern 5G and LTE TDD networks.

In my previous blog, I discussed the current state of the mobile industry, the economic pressures associated with it, and how Spectrum Effect is helping operators leverage the power of AI to maximize capacity and improve end-user quality of experience with their existing Spectrum and Radio Access Network (RAN) investments.

Now that the Biden administration has released the National Spectrum Research and Development Plan (https://www.whitehouse.gov/wp-content/uploads/2024/10/National-Spectrum-RD-Plan-2024.pdf), it seems like the right time to share some details about Spectrum Effect’s work in the area of Spectrum Coexistence. That is, helping incumbent users and future entrants utilize shared bandwidths in a harmonized, secure, and trusted manner. Let’s start with the problem we are facing here in the U.S., which may help contextualize why our government is aggressively investing in related R&D efforts.

Spectrum is the backbone of today’s economy. Many of the most successful and influential U.S. companies are built around the devices we carry in our pockets, the applications they host, and the services and experiences enabled by mobile connectivity. In the public sector, secure spectrum access underpins a broad range of critical infrastructure and public safety capabilities that similarly impact the lives of all Americans. As a global technology leader, the United States must ensure a robust spectrum pipeline to sustain its leadership in the digital age and ensure access to critical services.

Spectrum is a finite resource, and there are no clean and available spectrum bands in today’s pipeline. Many U.S. mobile industry and government organizations have expressed concerns about this shortage of spectrum, warning that we are lagging behind global competitors. The next likely tranche of spectrum to be made available for U.S. mobile operators falls within the lower 3GHz range, which is currently occupied by the Department of Defense (DoD). One option is for the DoD to vacate this spectrum, but studies indicate that this approach could cost up to $120 billion and take 20 years to implement. That cost, by the way, is one that taxpayers would assume if the DoD were forced to relocate.

Clearly, a more practical solution is needed—one that involves sharing the lower 3GHz spectrum between mobile operators and DoD through a Dynamic Spectrum Sharing (DSS) framework. For DSS to be viable, the framework must be both efficient and effective. Operators have made it clear that current spectrum-sharing methods fall short of meeting their requirements.

We have deepened our understanding of this challenge through our collaborative work with partners like MITRE and AT&T and believe there are two fundamental keys to solving this problem:

1. First, using the Radio Access Network as a Sensor (RaaS) to detect usage by the incumbents.
2. Second, using a wide area, multi-site, intelligent analysis software platform that allows the RAN to dynamically maneuver around incumbent users.

Based on our work with the U.S. Tier-1 operators and MITRE Corporation, a leader in federal research and development, Spectrum Effect has formed a unique perspective in this space, and we believe we are uniquely positioned to deliver a win-win solution for both MNOs and federal users.

Let me give a quick recap of what Spectrum Effect does in U.S. Tier-1 mobile operator networks to illustrate how we can extend our current solution to deliver a DSS framework using the RAN as a Sensor. The Spectrum Effect team utilized RAN performance management data from 50 mobile operator networks across the globe to train AI models that can detect various signatures created by interference present in a mobile network. We employed over 7 million hours — that’s around 800 years of data — to train our AI models. We did this because interference is a major factor impacting throughput and end-user QoE in mobile networks. If an operator can detect and remove or mitigate the interference in their network, they recover significant CAPEX investments and reduce churn, thereby increasing revenues.

When a mobile operator deploys the Spectrum-NET platform, it pulls data from the RAN every 15 minutes. This data is then analyzed using AI models to identify and map out interference in the network, identifying the interference type, location, and even the root cause. While helpful, improving the network requires one more step: removing the interference or configuring the network to avoid it. The Spectrum-NET platform does just that. We use closed-loop AI-driven automation to push configuration changes back into the network, optimizing throughput for the operator and improving QoE. This process is ongoing, as the system continuously detects and optimizes the network in response to changes in interference types and levels.

So, what does this have to do with dynamic spectrum sharing? The goal of the DSS framework is to permit operators to fully utilize the lower 3GHz spectrum, but only when incumbent systems are not active. If they are active, mobile operators are required to cease use of the spectrum until the incumbent system is no longer active.

The various radar systems the DoD uses have unique signatures that will show up as interference in RAN networks. Spectrum Effect is currently partnered with MITRE to train its ML models to recognize representative types of incumbent radar signatures. Spectrum Effect is researching the exact type and reporting frequency of the data from the network needed to align both the detection and mitigation strategies with incumbent protection requirements. This will solve the detection issue.

Next, we must ensure that mobile operators aren’t using the spectrum when the incumbent is present. Spectrum-NET can already push configuration changes in the RAN network to mitigate RF interference and improve network performance. Spectrum-NET will apply the same concept to dynamically push parameters on the grouping of cells to optimally maneuver around incumbent systems.

We are investing in this work because we believe this concept of using the RAN as a Sensor is the key to maximizing the U.S. mobile operators’ access to shared spectrum bands and to helping the U.S. maintain its global competitiveness while safeguarding U.S. economic and national security.

A lot has been said about the mobile telecom sector over the past few years, but not many of those comments have been positive. We all watched the story play out: billions of dollars were invested in 5G spectrum and infrastructure, but the coinciding revenues didn’t follow suit. The plan was always to monetize 5G in the enterprise, which is currently taking shape in the form of private 5G networks and operational technology transformation—a trend I’m optimistic about in the long term. But what about the large-scale consumer networks into which the industry invested billions?

Return on Invested Capital has long been the key metric for our industry. In the 2G days, we invested a ton of capital and delivered revenues in the form of voice services. In the 3G era, that capital outlay was returned in the form of SMS charges at 10 cents a pop. 4G was monetized via mobile broadband consumption. 5G hit a wall. Consumers are glued to their smartphones from the moment their built-in alarm app wakes them up until they finish up with their meditation app at bedtime, yet they don’t want to pay a penny more for the incredible connectivity and coverage operators provide.

Not surprisingly, because of this, there is tremendous financial pressure on mobile telecom operators. When those returns on the invested capital fail to materialize, there is pressure to reduce that CAPEX number. The new normal for the big 3 Tier 1 operators in the US is more than 10 billion dollars less than the level it was a couple of years back. Pressure on OPEX for operators isn’t a new 5G phenomenon, but the need for more automation and increased efficiency is under the spotlight more than ever. If the industry norm is to spend $5 in OPEX for every $1 in CAPEX, then constant pressure is inevitable.

Let’s look more closely at the CAPEX problem. How can the mobile telecom industry meet the insatiable demand for more coverage and faster speeds without continuing to invest capital at the previous rates? The answer is obviously that we need to do more with less, but how? In most cases, it turns out that mobile operators aren’t getting the full capacity they paid for. Billions were spent on Spectrum and Radio Access Networking (RAN) equipment, but the throughputs are often not what they should be because of interference. Interference can stem from various sources, and in cell sites with high levels of interference, throughput can be degraded by as much as 40%. Often, these are the most critical and dense sites in competitive markets. If we can remove or mitigate that interference, we can reclaim that 40% of capacity and avoid the additional CAPEX outlays.

When it comes to the OPEX side of the house, luckily AI has arrived just in time to save the day! I say that in jest because of all the hype, but I do believe AI and ChatGTP are part of the solution. As an industry, we were working on machine learning for 10 years before the ChatGPT craze. The technology has matured tremendously in that timeframe, and ChatGPT has mainstreamed the concept of daily AI usage. The promise of closed-loop, AI-driven automation and AIOps is now a reality. I’d argue that the technology for many use cases has been ready for some time, but it’s only recently that we’ve crossed the comfort threshold, thanks in part to ChatGPT. Not long –ago, telecom executives might have been fired for implementing an AI-based, closed-loop platform; now they might get fired if they don’t. The logic is straightforward in this area: more automation creates opportunities for greater efficiency and allows the reallocation of OPEX to higher-order activities and workstreams that can’t be automated.

What role does Spectrum Effect play in all of this? The team at Spectrum Effect foresaw this trend and has been working on AI and Machine Learning (specifically convolutional neural networks) before it became mainstream and cool. The team gathered data from more than 50 mobile operator networks across the globe and trained the Spectrum-NET AI models to learn the unique signatures left behind by each type of interference. When an operator deploys the Spectrum-NET platform, it maps out the exact levels, types, and locations of interference in the network. Where it is suitable, the AI engine then closes the loop by pushing changes into the network configuration to optimize the throughput and end-user Quality of Experience (QoE). The system continuously optimizes the network in a closed-loop manner, adjusting based on changes in interference levels (such as tropospheric ducting, which varies with weather patterns).

The state of the mobile telecom industry is clear, and the business requirements are well understood. Operators need to deliver more coverage and higher speeds at lower CAPEX levels, they need to show revenue growth, and they need to get more efficient and effective with their OPEX. Spectrum Effect is uniquely positioned to tick all three of these boxes:

• The Spectrum-NET platform optimizes existing spectrum and RAN investments, enabling operators to reclaim up to 40% of the CAPEX already invested in sites with high levels of interference.
• By mitigating interference conditions, Spectrum-NET improves QoE for end users. As coverage and QoE are the largest factors in churn, this QoE improvement reduces churn, giving top-line revenues a boost.
• Interference is one of the most challenging and impactful issues in the RAN, and mobile operators are already throwing a ton of OPEX at the problem, often using manual, inefficient methods and costly third-party services. Spectrum-NET significantly simplifies and automates RAN optimization with AI-based, closed-loop methods.

We have not encountered a network on the planet that hasn’t benefited from the Spectrum-NET platform, and this is especially true in 5G, where interference is typically more significant and impactful. Some of the top operators around the world are benefiting from Spectrum-NET today, and many more are planning their deployments. If your team isn’t already working with Spectrum Effect, now is the time to reach out and learn how we can help.

This is what we are doing today, and it’s quite compelling. Next time, I’ll share what we are working on, it’s even more exciting.