1. What is your role and specific area of expertise at MathWorks?

I am the Product Manager for the RF and mixed-signal area and my background is in RF modeling. At MathWorks, we work closely with engineers worldwide to identify and address the needs of modern wireless systems, mixed-signal ICs and high-speed digital interconnects with customer challenges ranging from RF/microwave software and design to satellite communications to 5G/6G.

2. You’ve been with MathWorks for nearly 15 years, now. What attracted you to MathWorks and how have you seen the company evolve over that time?

I joined MathWorks in early 2009 because of the long-term view, how it keeps investing in emerging technologies and serving cutting-edge requirements. For example, since then, MathWorks has launched Antenna Toolbox for rapid electromagnetic analysis of antenna arrays, and 5G Toolbox for modeling the physical layer of 3GPP wireless communications systems, enabling companies such as MMRFIC to develop and productize radios with innovative hybrid beamforming architectures that would have been subject for advanced academic research only a decade ago.

3. MathWorks is known for MATLAB® and Simulink®, and the company touts a whole host of applications, industries, disciplines and capabilities. Can you describe a bit about the impact MATLAB and Simulink have in advancing engineering and science?

The MathWorks mantra is indeed “accelerating the pace of engineering and science”. In the RF and mixed-signal space, MathWorks has developed solutions for RF simulation, antenna design and signal integrity analysis for our engineering audience. I work closely with users developing systems pushing the boundaries of frequency and bandwidth: moving into the mmWave for wireless applications and above 100 Gbps for high-speed I/O.

Researchers in companies and academia use MATLAB and Simulink to develop new algorithms and architectures for beamforming, equalization and linearization. For example, Otava Inc. has developed a model for their 24 to 40 GHz 8-channel mmWave beamformer integrated circuit that can be used in combination with AMD-Xilinx ZynQ 7000 FPGA for targeting the control logic of the whole phased array system.

4. How do MATLAB and Simulink work together to help designers? Can you give an example of how it can offer engineers more choice in design?

MATLAB and Simulink together enable Model-Based Design: a design methodology centered around modeling and simulation. By focusing on algorithmic modeling early in the design process, engineers can more easily explore new ideas, architectures, algorithms and be even more creative. Model-Based Design is a key enabler of moving verification earlier in the development process where more effort is spent to evaluate specifications and time is saved in finding and fixing errors, leading to overall faster time-to-market and higher quality products.

Supporting multiple abstraction levels and enabling model transformation, Model-Based Design allows designers to target hardware (processors or FPGA) and share models (IBIS-AMI, SystemVerilog, FMI/FMU) faster than using manual coding. Designers can rapidly iterate through different ideas, build prototypes and demonstrators with less effort and find errors earlier in the development process. For example, with Model-Based Design, DigitalGlobe simulated a complete satellite-to-ground communications system and proved that it could support a 50 percent increase in data rate before committing to costly prototypes.

DigitalGlobe communications systems operate in the X-Band, which is adjacent to the frequency band used by the Deep Space Network (DSN). To avoid interfering, DigitalGlobe engineers had to meet stringent out-of-band emissions and power requirements, even as they sought to increase their downlink data rate. The team needed to implement filters with steep edges, but such filters can distort the signal. Further, as they changed the modulation scheme, they needed to accurately model cross-polarization interference to ensure that the system could operate with a bit-error rate of less than 10-6. By using Model-Based Design, the team proved that they could achieve their target by running simulations in Simulink in which 10 million bits were transmitted and by analyzing effects due to distortion, the environment and the satellite’s elevation.

5. As we find ourselves in the age of AI, what role does AI play in helping the simulation and analysis process for engineers?

As time progresses, new design and analysis tools are made available to engineers. AI is delivering its promise and is seeing practical applications every day.

In the areas of RF and mixed-signal, AI is being used for speeding up the optimization of complex designs using surrogate techniques, for enabling more insightful signal classification in cognitive radios, as well as for correcting errors and making communication links more robust with auto-encoding, equalization and linearization algorithms. Engineers are embedding AI techniques in commercial products to solve the most challenging problems, adding to the benefits already provided by Model-Based Design. For example, NanoSemi engineers used MATLAB to develop and tune digital predistortion and machine learning algorithms for linearizing power amplifiers. The engineers needed to develop, debug, optimize and implement their AI algorithms rapidly; iterate with customers to ensure that systems under development met their needs and conduct thorough regression testing across a wide range of operating conditions. With MATLAB, they could share with their customers the fixed-point implementation of a specific module while protecting their IP and achieving early validation, which is a significant advantage because it minimizes errors due to misunderstanding.

Another example is Bharat Electronics Limited (BEL), which applied AI regression learning techniques to estimate the elevation of targets in 3D surveillance radars. With the help of AI, the aerospace and defense company could make its radar more resistant to reflections and noise. BEL engineers used MATLAB to generate synthetic data, saving the team time for complex calculations and gathering field-recorded data. They then used MATLAB to evaluate multiple regression methods to find the best technique for accurate target predictions.

These are all practical examples where AI techniques have been embedded in commercial products, adding to the benefits already provided by Model-Based Design. Using MATLAB, engineers and researchers could focus on innovation, iterate more rapidly with customers and reduce testing and debugging time.

6. From which of the RF and mixed-signal applications are you seeing the most activity? How do you expect this to change in the future?

For wireless communications as well as radar and other localization systems, there is a market move towards mmWave wideband systems development. This same trend is occurring for 5G and 6G, WiLAN and UWB, as well as satellite communications, leading to the exploration of:

· Massive MIMO arrays with hundreds of antenna elements, now growing to thousands of elements for creating intelligent reflective surfaces

· Higher operating frequencies above 10 GHz and approaching the sub-Terahertz range

· Larger signal bandwidths already covering hundreds of MHz and stretching to GHz.

These trends are directly reflected in the algorithmic complexity of communications systems and are also impacting the testing procedures. For example, Qualcomm is developing RF front-end components and envelope tracking technology for 5G mobile devices that support over 30 different RF bands. The number of possible waveform combinations in 5G is 10x greater than in LTE, making device validation much more complex and time-consuming. Using MATLAB, Qualcomm built a complete model of the Tx and Rx paths with fixed-point digital blocks and hardware-accurate power amplifier models. They used simulations to predict key system performance measures, optimize design parameters and automate testing over a range of waveform combinations. Finally, they used Model-Based Design to automatically generate C libraries that could directly be shared with their customers.

Similarly, for high-speed digital interconnects there’s the move towards higher and higher data rates exceeding hundreds of Gbps, moving towards optical and photonics technologies and integrating some algorithmic techniques coming from wireless standards such as higher-order modulations and forward error connection.

In parallel, there’s AI, either machine learning, deep learning or reinforcement learning techniques, being applied to speed up the integrated circuit design process and to develop algorithms for signal recovery and correction.

7. Faced with the transition to 5G, and later 6G, what are the key challenges facing engineers, as well as the most promising RF and mmWave solutions?

As the spectrum becomes increasingly crowded and radios gain more flexibility in accessing it, designers face the challenge of identifying and exploring scenarios that involve interfering signals and adverse channel conditions. Developing strategies for coexistence and mitigation, such as linearization and beam / null-steering, adds further complexity to the task.

Using a virtual platform for algorithmic design and verification is extremely beneficial before developing costly prototypes or committing to a given solution. This drives the need for integrated analysis tools of RF front ends, antenna arrays, channel models and digital signal processing algorithms that can accurately anticipate the interaction between emerging technologies and the environment in which they will operate.

From a modeling and simulation perspective, there’s the need for tight integration of simulation technologies that in the past were used in isolation by specialist teams such as different RF solvers, multiple electromagnetic analysis techniques and ray tracing channel modeling.

8. In looking back at the year, what have been some of the most significant trends you saw within RF and mixed-signal applications?

With the evolution of semiconductor technologies and manufacturing, next-generation wireless communications and localization systems become more integrated and available for mainstream consumer applications. With smaller margins, lower costs and faster design iterations, complex organizations with deep supply chains struggle to assess and iterate on requirements and specifications.

To better support this complex design process, the electronics industry is going through a transformational change, similar to the one experienced by the automotive sector in the past decade. In this context, the most interesting trend emerging is the need to create representative models that can be exchanged and used as executable specifications to document, negotiate and iterate requirements across multiple teams and organizations.

As the models are as complex as the chips that they represent, to speed up the modeling and verification process the industry is starting to make its first attempts to combine traditional fitting and physics-based techniques with AI tools.

9. What emerging opportunities are you and MathWorks excited about?

The increasing need for executable models and digital twins of complex RF systems, as well as advanced simulation tools to connect multiple domains spanning from RF, electromagnetics and digital signal processing, is gaining traction.

Engineers are building the next generation mmWave communications system by modeling complex beamformers operating in different environmental conditions. With tools like MATLAB and Simulink, system designers can develop architectures and algorithms, and before building complex prototypes they can simulate the effects of RF non-linearity, multiple noise sources, antenna coupling in the near and far field, beam-squinting and dispersion effects and ray tracing channels including interfering signals.

10. What else should our readers know about MathWorks and your areas of expertise?

As technology evolves and system requirements become more stringent, using system-level models that integrate digital signal processing algorithms and accurately anticipate complex RF effects is necessary to speed up development, innovate more rapidly and find errors earlier. As components cannot be analyzed and optimized in isolation, executable models for wireless communications, radar and high-speed digital interconnects have become an essential tool to reduce the risk of integration issues and miscommunication.

Tools like MATLAB and Simulink provide a platform where multiple physical domains, data-driven models, abstraction levels and simulation techniques can be integrated together. The benefits of Model-Based Design have been largely proven by the automotive industry, and the electronics sector now has the opportunity to leverage this methodology for the development of RF systems.