IMS 2025: A Look at Industry Technical Trends

What you’ll learn:

• Terahertz and sub-terahertz frequencies come at propagation costs to be made up through direction antenna arrays and beamforming technologies.
• AI and machine learning are coming to the aid of designers of both passive and active elements of RF systems.
• Quantum computing will lean heavily on advanced microwave technologies.

With the 2025 edition of the International Microwave Symposium (IMS 2025) behind us, it’s as good a time as any to look back on the event. In doing so, one can always glean insights into where we’re heading. What did this year’s show say about the state of our industry? And what are some of the trends that will drive it as we embark on the second half of the year?

First, a note on the event itself: IMS 2025 was in some ways a return to shows of old. Being back in San Francisco lent a flavor to IMS that it didn’t have last year in Washington, D.C. For one thing, the weather was nearly perfect versus the suffocating heat and humidity of Washington, which made for a more enjoyable week. For another, the proximity to the nearby tech mecca of Silicon Valley gave IMS more of a feeling of relevance.

The show was reasonably well attended, though as always there were periods of diminished traffic on the exhibit floor. Most of the exhibitors I visited expressed satisfaction with the value of their investment. The larger exhibitors’ spaces (think Analog Devices, Keysight, and MACOM) were almost always very busy with attendees viewing demos or speaking with technical staff.

Riding the Terahertz Waves

Among the top reasons anyone goes to IMS is the technical sessions, from which one can draw conclusions as to trend-setting technologies. One topic that was the subject of several sessions, which brought in large audiences, was the application of sub-terahertz and terahertz frequencies in communications.

With so much unallocated bandwidth available (10X that of the mmWave band), it’s obviously attractive to venture into terahertz frequencies as we look toward the commercialization in the 2030s of sixth-generation (6G) cellular technology. But as usual, physics brings good and bad news. The latter, in this case, is the high atmospheric absorption of terahertz waves, which nominally restricts use of terahertz frequencies to very short links, perhaps largely in indoor networking scenarios.

To make up for the poor propagation characteristics of terahertz waves and enhance their viability as a communications medium, the industry must look to very small directional antennas and/or phased arrays consisting of many antennas. Doing so helps to concentrate the output power of a terahertz-wave transmitter toward a receiver, facilitating communications at longer range.

On the phased-array front, researchers are experimenting with compact antennas-in-package (AiPs) for extended-reality applications in medical operating rooms. These modules, meant for sub-terahertz applications in the range of 92 to 300 GHz, integrate a waveguide antenna array with a divider/combiner within its multilayer substrates.

Within the module are two transmit/receive RFICs designed for a D-band phased array and fabricated on a 65-nm CMOS process. The module can adjust the phase and amplitude of eight transmit/receive signal paths (Fig. 1). The researchers, from a group of companies that includes Panasonic and the Institute of Science Tokyo, have achieved data rates of 40 Gb/s at three meters in the 150-GHz band.

In line with the nascent move to the sub-terahertz and terahertz spectrum, the industry is concurrently thinking through the modulation schemes that would best serve communications at such high frequencies. As we prepare for 6G, we’ll need to develop systems that meet demands for greater bandwidths and much higher data throughput. The sub-terahertz range is ripe for big gains in system data capacity and overall performance.

One avenue explored at IMS 2025 for the sub-terahertz region is ultra-wideband (UWB) vector modulation. Sub-terahertz frequencies of up to 500 GHz would surely serve as a waystation on the climb to the terahertz range. On this front, a group of researchers from the University of Freiburg and Fraunhofer Institute for Applied Solid State Physics is exploring UWB vector modulation for the 200- to 480-GHz range to be integrated with phased-array antenna systems. They’ve been able to demonstrate up to 6-bit phase shifting and achieve a reasonable 16-dB insertion loss at this early stage of the game.

Beamforming technology is another way to achieve efficient sub-terahertz communication. To that end, a team of Intel researchers demonstrated a 28-GHz beamforming element using monolithically integrated gallium-nitride (GaN) and silicon transistors fabricated in a 300-mm process. They built a complete system with transmitter and receiver both containing phase shifting and variable-gain amplification. Within the beamformer are TX and RX paths selected by T/R switches (Fig. 2).

Artificial Intelligence Invading Wireless

There was a noticeable increase at IMS 2025 in conversation surrounding artificial intelligence (AI) compared to last year’s conference. That’s not unexpected, as the industry continues to both embrace and refine how it can apply AI to the task of system design. A full session was devoted to the advances being made in the intersection of AI and RF systems, with more than a passing nod to use of digital-twin methodologies to enhance system-level design flows.

For example, a team of researchers from Epirus presented an investigation of machine-learning (ML) models for RF power amplifiers (PAs). Today’s RF circuit design methodologies are premised on modeling RF components in the frequency domain, which simplifies system analysis and predicts the system’s steady-state behavior. In the time domain, though, one may capture the system’s transient behavior while also gaining insight into memory effects that may arise from nonlinearities.

Using a recurrent-neural-network (RNN) architecture (Fig. 3), the Epirus team has been able to predict PA behavior across several signal amplitudes and bandwidths. Not only have they built software-defined digital-twin versions of hardware PAs, but they’ve also been able to model multiple digital twins to form a digitally cascaded RF amplifier chain. The latter has implications for system-level modeling.

Meanwhile, at the University of Southern California (USC), researchers are applying algorithmic machine-learning approaches to the design of on-chip, multilayered passive networks. In eschewing mature and labor-intensive traditional approaches, USC’s team was able to explore exponentially larger design spaces compared to single-layer designs, with demonstrations centered on RF power combiners and bandpass filters.

Using their inverse electromagnetic design algorithms, they can scale down passive network sizes with minimal loss of performance. Going forward, the team plans to move on to other passive network types and co-design passive elements with active RFIC circuit blocks.

Quantum Computing Meets Microwaves

Microwave technologies are critical in the development and functioning of quantum computers. In addition to the conference’s Quantum Bootcamp devoted to bridging the gap between microwave engineering and quantum physics, one IMS 2025 session explored some of the latest breakthroughs on the microwave side of the equation.

In one instance, a Taiwanese team of academic researchers from the Industrial Technology Research Institute (ITRI) and Academica Sinica’s Institute of Physics hopes to facilitate massive quantum-bit (qubit) arrays with an integrated, broadband front-end module (FEM). The module reads out two transmon qubits using frequency-division multiplexing (FDM) technology.

The cryogenic FEM (Fig. 4) features a customized IQ upconversion mixer that supports frequencies from 4 to 10 GHz with conversion gain of 0.3 (±3 dB) at 4° K. While the FEM is verified for readout of two qubits, the researchers incorporated FDM as a means of scaling up the number of qubits to offer wide RF bandwidths and low spectrum leakage while maintaining sufficient linearity.

At Germany’s Forschungszentrum Jülich, a consortium of multidisciplinary research institutes, it’s hoped that work on a cryogenic photonic link will result in an optical quantum-computing interface. Using 904- and 1310-nm lasers with bandwidth of >1 GHz, the research team seeks to address the growing size, complexity, and thermal concerns posed by scaling of quantum computers with these optical links for qubit control.

The Industry’s Future Parties Hard

Let’s face it: the microwave industry comprises an aging engineering community. IMS has a track record of encouraging up-and-coming young and/or female engineers and entrepreneurs, and this year’s edition was no different. Hopefully, the IEEE’s Women in Microwaves (WiM) reception, sponsored by Microwaves & RF and Samtec, presaged a surge in the industry’s ranks of these energetic and creative engineers (Fig. 5). Judging by the evening’s lively conversation and networking, we’ll be seeing more of them in coming years.