Better transistors won’t solve your RF system issues
To ensure that world-class millimetre-wave transistors deliver their full potential, it’s essential to optimise packaging, thermal design and integration
By Tudor Williams from Filtronic
There’s a tendency to assume that leaps in RF performance come from far better transistors. After all, that’s held true for much of the past few decades. Yet, just like the stock market, past performance is no guarantee of future success.
For those of us that have developed microwave and millimetre-wave systems, it’s common knowledge that system performance is not just determined by the device – beyond a certain point, what surrounds it is just as important. When frequencies rise, thermal management starts playing a critical role, along with packaging, manufacturing tolerances, and the interfaces between components.
In the millimetre-wave domain, small imperfections exert a measurable impact on efficiency, output power and overall system behaviour. As a result, the effectiveness of an RF systemis not simply governed by the performance of the semiconductor device in isolation. What really matters is how successfully that technology is integrated into a robust, manufacturable solution.
The frequency challenge
Success on this front is increasingly important, due to an inevitable shift towards higher frequencies in several significant applications. They include: communications networks, needing more capacity; satellites, demanding more throughput; and defence systems, expected to process and move greater volumes of data than ever before. As the spectrum available at millimetre-wave frequencies provides a way to meet those demands, development activity is continuing to move up, towards frequencies found in the Q, V, E and W-band.
Anyone who has spent time developing hardware at these frequencies knows that design challenges change quite dramatically when moving up the spectrum. At lower frequencies, strong device designs often mask weaknesses elsewhere in the system. But at millimetre-wave frequencies, that helpful margin largely disappears. In this domain once negligible losses increase in significance, packaging features start influencing electrical performance, and even small mechanical variations exert a measurable effect on the behaviour of the finished assembly.
Due to these issues, engineers must shift their focus from the device itself towards the performance of the subsystem as a whole. So, the challenge is not limited to just developing a high-performing transistor. While that’s an essential starting point, the bigger test is ensuring that great performance survives the journey through MMIC design, packaging, interconnects, waveguide structures and every other part of the RF chain. At millimetre-wave frequencies, that’s often where a system succeeds or falls short, and where many of the most interesting engineering challenges now exist.
An engineer at work at Filtronic’s high-frequency engineering laboratory
Packaging matters
Traditionally, packaging is viewed as a relatively trivial task, undertaken after the real RF design work is done. However, while that may have been true at lower frequencies, it is certainly not the case at millimetre-wave. Here, packaging is now very much part of RF design, and separating the two is increasingly difficult.
Those of us that have worked on high-frequency subsystems are painfully aware of this. We can recount times when a design looks excellent in simulation, and performs exactly as expected at the device level, before hopes are dashed when losses and discontinuities elsewhere in the signal path erode performance. There are many potential culprits. Bond wires, transitions, interconnects and package interfaces influence the result, with impacts increasingly pronounced as frequencies and power levels rise.
The real measure of success isn’t what happens in the laboratory. It’s whether that performance is reproduced consistently in a high-yield, long-lifetime manufacturable product, day after day, and unit after unit. Those goals demand as much attention to packaging, assembly and process control as to the device itself.
Increasingly, the best results are coming from close collaboration between RF designers, process engineers, packaging specialists and subsystem engineers, working in unison towards a shared system objective.
The thermal reality
If there’s one area where engineering ambition still has to answer to physics, it’s thermal management. While compound semiconductor devices are continuing to deliver impressive gains in power density, and allowing engineers to squeeze more performance into ever-smaller footprints, there is no escaping the thermal consequences.
Higher power densities generate more heat, and extracting this in an efficient manner holds the key to achieving the performance that the device promises on paper. It’s not uncommon that the largest constraint to output power, efficiency and even long-term reliability is the ability of the subsystem to dissipate heat effectively.
Due to this, it is inappropriate to consider thermal management as a separate engineering discipline. Instead, one must evaluate the direct influence of thermal behaviour on RF performance – it affects everything from gain and efficiency to stability and reliability. As a result, thermal modelling is increasingly intertwined with RF design, particularly at millimetre-wave frequencies, where performance margins may be tight and overall efficiency is typically lower. Invariably, the best solutions build thermal considerations into the architecture from the outset, rather than address this after completing RF design.
High-volume, high-value manufacturing at Filtronic.
Integration: the real differentiator
Today, success is increasingly defined by how well technologies are brought together, rather than the performance of any single component. Customers place their emphasis on whether the complete subsystem fulfils their requirements – it could be a satellite gateway, a defence platform or a communications network.
To excel in this regard, power amplifiers, filters, frequency-conversion stages, waveguide structures, antennas and control electronics must all work together, within a carefully engineered architecture. With system complexity on the rise, alongside operating frequency, the ability to integrate these technologies effectively is an increasingly important differentiator.
The good news is that over many years, there has been a much-needed shift towards system-level thinking. Engineers are now well aware that improving one aspect of performance tends to introduce a challenge elsewhere, whether that’s thermal management, manufacturability, reliability, or size and weight constraints. The challenge is rarely about maximising a single parameter; it’s about achieving the right balance across the subsystem.
Beyond the transistor
As engineers take an increasingly holistic approach, compound semiconductor technology will still remain at the heart of high-frequency communications, with continued device innovation playing a key role in industrial progress. However, many key engineering challenges now sit beyond the semiconductor design itself.
To ensure that a device delivers its full potential, engineers must consider packaging, thermal management, manufacturability and subsystem architecture. These factors are increasingly important as the
millimetre-wave era flourishes. Here, organisations that thrive are those taking a system-level view and recognising that competitive advantage is not simply a matter of building better devices, but integrating them into reliable, repeatable, high-performing RF subsystems.
