How did you get into electronics/ engineering and when did you start?

I trained as a physicist at Bath University between 1983 and 1987 and was awarded a PhD from Warwick University in 1993, regarding the use of coherent optics for full-field analysis of transonic airflows by holographic interferometry. This technology is valuable in the wind tunnel testing of civil and military aircraft.

I subsequently worked on electronic speckle pattern interferometry (ESPI) and shearography for non-destructive evaluation at BAe. This work, and the analysis of airflows, both involve the processing of images formed from coherent laser radiation. Laser illumination and radio waves are superficially very different, but both are electromagnetic radiation and their propagation has a lot in common. Furthermore, the processing of the resulting images also has a lot in common with the functions needed for digital channelisers and receivers.

I gained an in-depth understanding of wireless communications, while working at Toshiba’s European Research Laboratory between 2000 and 2007. I was involved in MIMO research and performed wideband channel sounding to assess the performance of space-time codes that were being considered for use in the IEEE 802.11n WLAN standard. I subsequently led a group of PhD qualified engineers in the research of gigabit/s data rate wireless solutions using ultra-wideband (UWB) transmission.

Since 2007, my theoretical understanding of communication systems has been complemented by practical experience at RFEL. Notably, I have designed and modelled ultra-high data rate, pipelined, switched-length fast Fourier transforms for use in channelisers; and led a team in the design and production of an entire multi-channel demodulator from the ADC to the decoder.

What are your favorite hardware tools that you use?

These days my work is based on modelling and therefore the only hardware I use is my PC. My colleagues that implement FPGA solutions in hardware use the usual laboratory instrumentation, such as spectrum analysers, oscilloscopes and arbitrary waveform generators. In my previous optics research my favourite tool was a 100 MW pulsed ruby laser that was used to produce holograms with exposure times as short as ~10 ns!

What are your favorite software tools that you use?

All of my modelling is performed using Matlab, which is a versatile tool that is fast enough for most tasks and also provides comprehensive signal processing libraries.

What is the hardest/trickiest bug you have ever fixed?

It is often challenging to achieve bit-true agreement between a Matlab model and a VHDL implementation (RTL and gate level). It is particularly difficult for complex systems, such as demodulators, that have diverse processing requirements for a range of different burst types and modulation/coding schemes. A ‘divide and conquer strategy’ is the only truly effective way of finding bugs in complex systems, but this may be made more complicated if any third party IP is involved in the design.

What is on your bookshelf?

Amongst others:

• R.E. Crochiere and L.R. Rabiner, “Multirate digital signal processing,” Prentice-Hall Signal Processing Series, 1983.
• L.R. Rabiner and B. Gold, “Theory and application of digital signal processing,” Prentice-Hall International, 1975.
• P. Horowitz and W. Hill, “The art of electronics,” 2nd Edition, Cambridge University Press, 1989.
• J.G. Proakis and M. Salehi, “Communication systems engineering,” 2nd Edition, 2002.
• J.G. Proakis and D.G. Manolakis, “Digital signal processing: principles, algorithms and applications,” Third Edition, Prentice-Hall Inc, 1996.

Do you have any tricks up your sleeve?

A methodical modular approach is essential in implementing FPGA designs, with as much of the design modelled in Matlab as possible. It is also important to gate-level test the design, as synthesis tools may introduce bugs that were not present in the RTL.

What has been your favorite project?

Each design presents unique challenges that I find interesting. It is always rewarding when you finally get a complex core operating at the target speed and within the FPGA resource estimate. I also still have fond memories of when I produced holograms in wind tunnels. The thrill of capturing a three-dimensional flow image that potentially has never been seen before, and then reconstructing the image in all its glory away from the test facility is hard to beat!

Do you have any note-worthy engineering experiences?

I was awarded the Sweet Smith prize and the Thomas Hawksley Gold Medal from the Institute of Mechanical Engineers for my work on visualising transonic airflows, which was supported by a Fellowship from the Royal Commission to the 1851 Exhibition. One of my airflow holograms was displayed in the science museum. My work on NDT led to the Kenneth Harris James Prize and a system that was used to inspect whole aircraft panels.

I presented my research on the performance of multiple-input multiple-output (MIMO) space-time codes to the High Throughput Study Group of the IEEE that evolved into the IEEE 802.lln Task Group, which formalised the most recent MIMO WLAN standard. At RFEL I have had the pleasure of working with great teams of engineers in the design, modelling and implementation of state-of-the-art signal processing cores that are now in active service.

What are you currently working on?

I am working on a ‘one-design-fits-all channeliser’. At the simplest level, channelisation can be performed by just a simple FFT, although it is often challenging to design such a system with high throughput, low processing latency and within a tightly constrained FPGA hardware budget. However, an FFT channeliser has limitations and an application may demand certain features that require the use of other technologies, such as the weighted overlap and add (WOLA) FFT and the pipelined frequency transform (PFT).

The new flexible channeliser will use a variety of techniques to support channels of different spacings and bandwidths (potentially overlapping). In addition, it will support channels of differing output sample rates, to match the oversampling rates required for modulated communications channels. Importantly, the new channeliser will also provide tight control of the frequency response of the individual channels, so that the passband has a flat response and the overlap with adjacent channels is well defined.

What direction do you see your business heading in the next few years?

RFEL has developed a highly modular design philosophy, with a library of highly optimised rudimentary processing blocks developed in the early days to promote re-use. This approach did not yield dramatically complex designs immediately, but resulted in a sustained improvement, which has ultimately led to the capability of bringing complex and bug-free designs to market very quickly. With each coming year, the portfolio of functional blocks increases and the level of sophistication and complexity of designs increases.

RFEL now has the capability of designing and implementing multi-channel demodulators and potentially entire communication systems. In addition, RFEL has begun branching out into image processing, where there is the potential to increase the processing functionality that can be performed in real-time. Overall, I can therefore envisage the company developing increasingly complex systems and being a partner in exciting product opportunities.

What does RFEL do that is different from other providers of FPGA IP?

RFEL is very different from other companies where I have worked. Initially, a start-up has to be extremely competitive and focussed to survive, especially when exposed to volatile market conditions. The fact that RFEL has survived, won many accolades, attracted numerous diverse customers and is now a subsidiary of Rheinmetall Defence Electronics GmbH, is testament to the quality of the staff and their innovations.

RFEL’s success is partly down to its small size. System engineers, who also have intimate knowledge of the hardware, direct the signal processing design and modelling. Early on in the design, a pragmatic stance is taken on how to implement the design in an efficient manner. In many companies, senior staff are often promoted into positions where they are not using their best skills. In contrast, at RFEL the technical director is engaged at all critical times in most projects and is even a ‘hands-on-coder’ when needed. This would be impossible in a larger organisation and would be deemed to be inappropriate. However, at RFEL it ensures that every design benefits from the considerable experience of the most expert staff. In addition, a small company is naturally more dynamic and versatile, and can cut through bureaucracy and, when needed, can go the extra mile to reduce the critical time-to-market.

Why do people come to RFEL for solutions?

When companies come to RFEL they can offload risky aspects of a large project in the knowledge that RFEL has the pedigree in the area to deliver highly optimised solutions to agreed timescales. RFEL also offers considerable flexibility in their pricing strategy and how projects are resourced, which is appealing to large companies.

What challenges do you foresee in our industry?

Two aspects spring to mind: the need to produce a system that offers wireless data rates capable of empowering the mobile video applications of the future; and the need to unify wireless technologies to reduce the cost, complexity, footprint and power consumption associated with supporting numerous wireless standards. An ideal system would be capable of dynamically configuring itself to allocate channel resources to the required application, with continuous sensing of how the medium is being used by other users.

UWB technology has attempted to achieve the data rates required for non-compressed video transfer and wireless USB, but so far it has not realised its true potential. Possible reasons for this include performance failing to reach expectation, political in-fighting and the associated cost during protracted standardisation, fundamental cost of implementation and lack of immediate killer applications. There is a fundamental limit to how fast component technologies evolve and the concept of UWB is probably ahead of its time. A supporting technology that will increase the demand for gigabit/s wireless connectivity is the widespread deployment of domestic fibre-optics to increase broadband speeds. Recent advances in gaming such as the Wii and the Xbox Kinect, suggest that the consumer wants full immersion in a virtual reality, networked gaming environment; and wireless video transfer is the natural enabler of such systems. Migration from desktops to laptops, through to tablets and smartphones, is further evidence that the consumer wants ubiquitous truly personal computing. For the ultimate in mobility and ergonomics, this will surely mean a decoupling of the display technology and the computing resource, which will require gigabit/s wireless connectivity.

The rapid growth of wireless technology to satisfy diverse applications has meant the proliferation of wireless standards operating over either different or competing radio bands. Ideally, these services should be unified under a single communications standard, which allocates resources on an ‘as-needed’ basis. Spectrum should be allocated intelligently and sympathetically to enable maximum re-use without competition. A requirement for such a software defined radio system is a chain of highly flexible and reconfigurable system elements, governed by a standardised transmission protocol, which can adapt as and when required. This flexibility is required for the antenna and matching network, power amplifier, LNA, ADC, channeliser and transceiver.

In trying to achieve ultra high data rate WPAN connectivity with maximum re-use, the strength of UWB has also become its weakness. Legislation requires the transmission of very weak signals (-41 dBm/MHz) over very large bandwidths (500 MHz). This makes the system highly vulnerable to interference. The frequencies of narrowband interferers are not necessarily known in advance and therefore cannot be filtered out in the analogue domain prior to the ADC. The ADC, which typically has a resolution of only 4 or 5-bits, is therefore overwhelmed by this interference and cannot easily reject it in the digital domain. There is therefore a need for a ‘smart’ ADC/channeliser combination that can analyse the received signals and ‘notch out’ interferers that are above the expected receive level for UWB signals using active interference cancellation. RFEL’s flexible channeliser is potentially a vital component of such a system.

What is it like living and working on a small island like the Isle of Wight?

Happiness is all about achieving the optimum balance between work and home life, while remaining fit and healthy to enjoy it. The IOW is a beautiful environment and it is rare that a company of RFEL’s calibre is situated in such an idyllic location. There is nothing more invigorating than starting the working day by cycling across the island and taking in all of the scenery and wildlife.