The field of power electronics is becoming a strong sector within the UK's technology base, thanks to unconventional techniques such as combining analogue and digital control devices.
Electrification is helping to drive a renaissance in one key part of the UK's electronics business, one that has survived amid economic turmoil that has seen other parts of the industry in the country go to the wall. Work on electric aircraft systems, electric vehicles and the smart grid, as well as power supplies for more conventional electronic systems, is helping to bring the power electronics sector out of the shadows. Furthermore, the trends are helping to drive a wide variety of businesses.
At the same time there is a growing requirement for smarter devices that show themselves operating using energy-efficient technology – by regulation if necessary. This is another key driver for the power electronics sector, and one that will continue to have a defining impact on the direction this field takes in terms of research and development.
Power electronics is about deriving performance gains as well as pure smart energy conservation. One prime example of this is Formula 1 (F1). In the 1990s, F1 racing team McLaren helped revolutionise the sport with the introduction of sophisticated engine controls and radio telemetry to provide real-time data to engineers on the performance of cars around the track.
"Electronics was a differentiator in the early 1990s," says Peter van Manen, managing director of McLaren Electronic Systems, which develops the engine control units (ECUs) for all F1 vehicles, following a decision to standardise them in 2008.
McLaren has stepped into the world of electronic power control with the decision by the electric-car Formula E racing series to offer a standard engine design to any of the teams. In contrast to F1, teams are not yet obliged to use the same powertrain, as there is potential for differentiation at the engine level; but again the emphasis is on using intelligent control to eke more efficiency out of the battery.
Freescale Semiconductor in the UK has been supporting Vodafone McLaren Mercedes Formula 1 team (to give it its full moniker) for several years. Steve Wainwright, general manager of Freescale Semiconductor in Europe, the Middle East and Africa, says: "What you get in motor sport is a consistent level of innovation going on. The other thing we get is system knowledge: do we understand the application as clearly as we can? And in an extremely tough environment in terms of vibration and heat?"
This work translates over to the wider transportation environment, says Wainwright, which should lead to more efficient combustion and electric powertrains: "From 2014, the F1 engine will be much closer to that of a road car, with greater economy, and less fuel carried."
For the electric-power sources for more conventional electronic systems, TDK-Lambda expanded its R&D for intelligent control earlier this year with the creation of a laboratory close to Bristol. TDK-Lambda's main British power-supply manufacturing operation is in the north-Devon town of Ilfracombe, an unlikely location for an electronics company, perhaps, but not unusual for a number of companies working in this field in the UK.
Planning decisions and incentives from the 1960s to the end of the last century pushed organisations out of places along the M4 corridor that now are more familiar homes of electronics companies. "Bristol is fairly close to Ilfracombe, and it is a good hub for attracting people," explains TDK-Lambda marketing director Martin Southam. His colleague Andy Skinner, chief technology officer at TDK-Lambda, says that the work focuses on two areas: making use of new types of device that move away from traditional silicon, and to harness the increasing compute power of standard microcontrollers.
"New devices bring new characteristics and challenges and we want to be able to do more things algorithmically with microcontrollers that are very difficult to do with traditional analogue controllers," Skinner says. "We can have more degrees of freedom."
He adds: "If you look at power-circuit topologies, some are used very widely – because everybody understands them. They are relatively trivial to understand mathematically, although they are not necessarily trivial to implement well. And there are circuits that people have shied away from because they are more complex. They have not been studied in great detail, so we are looking at options that have previously been discounted."
Demands for greater efficiency are leading to designs that use the microcontroller to shift between different states, whereas older designs use the same algorithm to drive the circuitry across its entire operating range. "If you take a power supply from ten years ago, it may have been dissipating 20W at zero load," says Skinner. "That means that the power factor correction circuitry never sees a lower load than 20W, narrowing the range over which is has to operate."
Power-factor correction is vital in the high-efficiency supplies used to feed most electronic equipment to avoid harmful distortions being fed back into the grid. "[These days] that hardware is designed to run at 3W or lower at no load to more than 100W. Having a single controller work over the whole envelope has become more difficult. An alternative is to break it into a number of steps," Skinner points out.
Southam adds: "The microcontroller is easier to set up for multiple operating states that depend on factors such as the load. That way you can optimise efficiency across more of the load curve."
Increased digitalisation is feeding through into integrated circuits (ICs), although analogue technology expertise remains vital. David Baillie, CEO of fabless power chipmaker Cambridge Semiconductor, admits: "We talk about our hybrid power controller; 100 per cent digital is not good. You need to put a lot of thought into what you do in analogue and digital – but the overall trend is towards digital."
The requirement for increased digital content in otherwise analogue ICs provides an opportunity for companies in the UK that have found it increasingly difficult to raise venture financing for the digitally dominant devices operating at the heart of mobile phones and tablets.
Such firms need hundreds of millions of pounds of advance funding to develop their chips for sub-40nm process technologies with no guarantee of success at the end as they come up against more established players such as Qualcomm or Broadcom. Even the masks used to define the on-chip features on those billion-transistor devices cost millions of pounds to manufacture.
"Starting a 'bleeding-edge' start-up, where you are trying to operate at 20nm or below – those days are gone," explains Baillie. "It is just too expensive, too risky, and takes too long to get products like that to market. It's not just because of the cost of hardware development, but because they have significant software content as well. It's better now to try to focus on areas where you can exploit a wide base of foundry capacity and older technology, where mask sets are one-twentieth of the cost of advanced processes. You also have shorter cycle times in fabs and all those benefits."
He adds: "I believe the 'more-than-Moore' model is key to the future of semiconductor-related start-ups. However, we as a country need to consider what happens if we do no system-level stuff. That's not healthy either."
The 'more-than-Moore' model Baillie refers to uses advances in packaging to bring different types of chip technology together rather than trying to squeeze everything on to one die. Processes optimised for power, for example, use much older, larger-geometry processes, and potentially different materials to silicon to provide features such as high breakdown voltages to prevent circuits shorting out. By putting these on different chips to those used for core computing, each process can be operated in its sweet spot. Power is a prime candidate for more-than-Moore, because of its need for high voltages and increasing use of advanced digital control.
Companies such as Cambridge Semiconductor can use older processes at foundries to make their mixed analogue and digital power-control devices. Several years ago at an industry conference, Russell Haggar, director of consultancy Enlightened Technology, coined the term 'Moore's outlaws' for this class of company. To Haggar they were "the guys who don't need to be playing the Moore's Law game".
Baillie says the split between fabless and owning your own fab comes down to whether you need to customise the process to make the devices you want.
"If you are trying to do ultra-high-voltage power switch, which we did for many years, it is difficult to optimise that in a fabless context," Baillie explains. "I would argue that no company has successfully done that. Some fabless companies offer high-voltage, but then they tend to have a very close relationship with the foundry. The winners in that space long-term will be people who own their own fab."
On the controller side, however, "you don't need things like ultra-high-voltage", Baillie adds. "So we can use more standard foundry processes, used to make things such as LCD display drivers. There is a plethora of foundry capabilities in that area, so access to manufacturing is not a major constraint."
Cambridge Semiconductor, meanwhile, is pursuing niches that align with some of the faster-growing sectors in digital systems, as well as new markets such as LED lighting – providing the circuitry for AC-connected power converters. Others, such as Dialog Semiconductor and Nujira, are focusing on the circuitry inside the devices themselves to provide more power-efficient radio control.
Others are pursuing the fab-owning route, focusing on more exotic processes for the ability to handle high-voltages among other things. The longest serving fabs in the UK are more likely than not to be making devices for power circuitry and some unexpected locations.
They have lasted much longer than the logic-oriented fabs operated by companies such as Freescale, Fujitsu and NEC, among others, that were attracted to the British mainland partly by inward-investment deals.
At the end of the 1990s, Zetex Semiconductors took its name from a series of bipolar transistors developed by Ferranti in the 1970s and 1980s. After the commercial collapse of the defence contractor in the early 1990s, subsequent sale to Plessey and reluctant absorption into GEC, management bought out the transistor-manufacturing operation based around two fabs in Oldham.
An older, 4in fab line shut in 2008, a year after the company was acquired by US-based Diodes, with production moving to a more- recently constructed 6in line in Oldham. In general, as long as volumes are high, chips are cheaper to make on larger wafers.
On the other side of the Pennines, AEI Semiconductor set up in the city of Lincoln almost 60 years ago, before becoming another part of the GEC empire as Marconi Electronic Devices (MEDL) in the early 1980s.
Like Zetex, the company bought out of the defence contractor before being acquired by another overseas player – in this case China-based Zhuzhou Times Electric – but its Lincoln-based fab remains, and it also continues to prosper.
Even older is the fab operated by Raytheon UK (also known as Raytheon Systems Limited) in Glenrothes, first opened by the Hughes Aircraft Company (of Howard Hughes fame) in the early 1960s. The company claims that it has provided a cornerstone for the development of the so-called 'Silicon Glen'.
Raytheon UK specialises in defence, national security and other government and commercial markets around the world. Its latest incarnation exemplifies one of the strands in the UK's modest power-led semiconductor renaissance: the introduction of novel materials. The company opened the UK's first fab that is equipped and able to make chips from the ultra-hard material silicon carbide in January 2013.
Semiconductor business lead at Raytheon Glenrothes Paul D'Arcy says that the company is trying to exploit the high-temperature characteristics of carbide. It is doing foundry work for a number of small power and discrete-components companies who need a certified production shop as well as working with university researchers.
He adds: "We also have a power module group. These wide-bandgap materials will find their way into the next generation of power modules. The module guys are actively using silicon carbide to further their aims in things like aviation and other harsh environments."
Diamonds are this fab's best friend
In County Durham, start-up Evince Technology has turned to an even harder material – diamond – for a new generation of power device. In 2009, the company demonstrated a power switch able to handle voltages up to 15kV, more than twice as high as silicon, and is now working to build components such as diodes that can be used in grid-switching systems.
Much further south, diamond miner De Beers' industrial subsidiary Element Six has consolidated R&D at a facility close to Rutherford Appleton Laboratory near Oxford, where the company is exploiting diamond for a number of applications, although it has yet to start making diamond-based switches.
"We have done some research on power semiconductors," says Element Six researcher scientist Dr Ian Friel, "using layers of diamond deposited using chemical vapour deposition, a common semiconductor technique." Admittedly, the work has not yet yielded products. However, diamond is useful for power in other respects. Chris Wort, new technology manager at Element Six, says: "One of diamond's more extreme properties is its ability to conduct heat."
Thin slivers of diamond are now used to conduct heat away from the power-conversion devices in high-end Internet switches and high-power lasers. The slivers of diamond are often used as heat spreaders, taking heat from a small device to a much larger, finned heatsink, which is less efficient at thermal transfer but is cheaper to make in large sizes.
As power comes out of the shadows, it is becoming more actively researched and promoted by a number of the UK universities. The University of Nottingham, for example, is leading the creation of a national research centre for power electronics with £18m of funding from the Engineering and Physical Sciences Research Council. Two months earlier, PowerelectronicsUK was formed, backed by industrial, government, and academic groups with the aim of boosting the number of engineers working in the sector. TDK-Lambda's Andy Skinner says that there is a shortage of UK-based talent in this field, particularly in areas such as digitally-intensive power control.
"Recent activities in power electronics in UK universities is excellent news. Power electronics is hugely important, and it hasn't had the recognition it deserves," Skinner adds. "We hope that more people will study power electronics."
Additional reporting by James Hayes.