HOME :: JOB LISTINGS :: ARCHIVES :: MEDIA KIT :: SUBSCRIBE


IMEC
Research for the Real World

by Dick Selwood, Embedded Technology Journal

You are designing a device and you want to add radio connectivity; what type of interface do you want? Wi-Fi? 3GSM? Bluetooth? WiMAX? Wouldn’t it be great to use just a single chip-set (or even a single chip) to deal with all these different standards? Software Defined Radio (SDR) aims to achieve this, and Cognitive Radio goes even further and automatically selects the standard, so a device will connect to 3 GSM on the move and then switch seamlessly to WiFi for telephony over IP at home.

Just to remind you, and simplifying greatly, wireless standards normally cover two things: the frequency and bandwidth at which the signal is transmitted and the transmission protocol. So equally there are two elements in a receiver: the front end – effectively the tuner - which selects the signal (working in the analog domain) and then the baseband, which carries out digital processing to convert the signal for use, for example as an audio signal for entertainment.

There are several companies who are working in this area: Technoconcepts has created a demo RF front-end of a single implementation for four different flavours of WiMAX, and BitWare and AsicAhead are also working in the front end. Others are working on a baseband device. However, a government founded research lab in Belgium is working on a complete solution, has products for both front end and baseband, and is looking like it will be a leader in this technology.

IMEC in Brief

In the early 1980s, many governments were trying to copy the “Silicon Valley” phenomenon, often missing the point that Silicon Valley had nothing to do with government intervention. Few of these efforts have survived except in the most vestigial form, but IMEC, based in Leuven, Belgium, some twenty miles to the east of Brussels, has been thriving for 25 years. It works as an independent R&D organisation with the mission: “To perform research and development ahead of industrial needs by 3 to 10 years in microelectronics, nanotechnology, and design methods and technologies for ICT systems." Well, that gives them plenty to keep them going.

It has arevenues of 240 million Euros (roughly US$300 million), of which just under 17% is a grant from the government of Flanders (equivalent to a US state government). There are 1500 staff from over 50 nationalities, with 946 actually on payroll and the balance made up from 200 PhD students and visiting scientists and 350 “Industrial Residents” - staff from commercial organisations working within IMEC as part of the teams that are working on joint projects.

These projects cover a great deal of ground, but are centred on the two clean rooms, 200mm and 300mm, where IMEC is tackling what they call (sub) 32nm CMOS and pursuing their More than Moore programme in co-operation with semi-conductor equipment and materials companies. IMEC provides a neutral ground for materials and equipment companies to work together, so that when new equipment arrives at the end user, there is no embarrassing pause while the material suppliers come up with something new.

To give some idea of the breadth of industrial involvement at IMEC, the (sub) 32nm CMOS programme includes the top five memory suppliers (Hynix, Elpida, Micron, Qimonda, and Samsung) as well as logic IDMs and foundries including Infineon, Intel, NXP, Panasonic, STMicroelectronics, Texas Instruments, and TSMC. Most of these companies have staff working on-site in Leuven to bring the technology together. IMEC also has links with many University research teams.

Applications

Building on the clean-room activities are several research groups, including Packaging and Interconnection Technologies, Organic Electronics, Solar Cells, High-Power High-Efficiency Electronics for Wireless Communication, Wireless Autonomous Transducers, and Nomadic Embedded Systems. The groups work with industrial partners so that their work is firmly anchored in real world needs, and they use the manufacturing and device expertise to create real products.

The Nomadic Embedded Systems group has just begun their Apollo programme, working towards the ‘embedded systems of 2011 and beyond’ with eight research programmes, including multi-processor system on chip (looking at mapping sequential C onto multi-processors), processor and compiler (looking at compilers for different kinds of parallelism), technology aware design (using system design to overcome issues of sub-45nm processing technology), multi-media (mapping video and 3D graphics onto multi-processor platforms), 60 GHz radio, and three programmes specifically in the SDR domain. For SDR the programmes are working on the front-end, on the baseband, and on cross-layer management for controlling the energy/performance trade-offs in SDR. While technically separate programmes, they work closely together for future integration.

Working Front End

The front end programme has already produced working silicon with two generations of the SCALDIO chip already unveiled. SCALDIO (SCALeable raDIO) has a frequency range between 147 MHz and 6 GHz with supported bandwidth between 700 kHz and 46 MHz, which IMEC says will cover current and future cellular, wireless local area network (WLAN), wireless personal area networks (WPAN), broadcast, and positioning standards. The device is currently in 130nm CMOS and consumes between 60 and 120 mW, depending on what standard is being received, comparable with the power consumption of a single mode chip.

The design contains a fully reconfigurable direct-conversion receiver, transmitter, and two synthesisers, all of which are adjustable; RF carrier frequency, channel bandwidth, noise figure, linearity, filter characteristics, etc., can each be configured to the specific characteristics of the standard being received, in a manner analogous to turning knobs on an old fashioned radio.

Reconfigurability can also be used for real-time power/performance trade-off; the example that IMEC uses is reducing the filtering level when the interferer level is below the standard’s worst case. Because this trade-off is dynamic, adjusting the power needed as the situation changes, the SCALDIO can produce better performance than domain-specific devices.

To twiddle the knobs on the different blocks within the chip, IMEC has developed an on-chip network, which ensures that all the different settings are coordinated. A circular topology uses a master node and multiple slave nodes linked with serial communication. The number of nodes and the number of bits can both be scaled, and the I/O interface is fixed on the number of bits rather than the number of nodes or analog blocks. The network is also independent of the number of analog “knobs” so that the routing area is fixed, and an EDA tool generates the network from a high level specification.

A universal programmable synthesiser uses a single wideband VCO (voltage controlled oscillator) with a 3-5 GHz range, instead of the more usual three VCOS, saving both power and silicon real-estate. A final radical element of the design is a direct up-conversion mixer with auto-calibration. The mixer, the synthesiser and the network are all protected by patents.

Baseband

To support the front end, IMEC has developed a low-power programmable digital baseband architecture, supported by a design and programming environment. The current design of the flexible air interface (FLAI) addresses 802.11n (Wi-Fi), 3GPP-LTE (3GSM), and 802.16e (WiMAX), and hardware emulation suggests that in 802.11n active mode it will consume only 300 mW. Baseband processing is carried out by a scalable processor architecture that IMEC has developed. ADRES (Architecture for Dynamically Reconfigurable Embedded Eystems) uses a VLIW (Very Long Instruction Word) processor and a coarse-grained two-dimensional array of processors.

Signal detection uses IMEC’s new configurable SyncPro tiles, the first silicon for which is in process. These consume very little power when in listening mode. A designer can specify multiple tiles for different standards, which will power up only when they detect a relevant signal.

Overall, the platform is controlled by an ARM processor, which switches areas on and off as needed, to maintain low power consumption.

ADC Record

An important element in wireless applications (and other areas) is analog-to-digital conversion, and IMEC have been innovating here as well, particularly addressing power consumption. They have created a charge-sharing SAR (successive approximation register) architecture. It uses passive charge sharing so there is no active power, and an asynchronous controller removes the high speed lock normally required. An implementation in 90nm CMOS with a core size of 400 x 200 µm, running 9-bit resolution with a sample rate of 50 Msamples/sec uses only 0.7 mW. Bit resolution and sample rate are scalable to the application, and since it is built out of MOS switches and metal-oxide-metal capacitors, the ADC will scale to deep sub-micron process technologies.

Cross Layer Management

IMEC is also looking at the total system, not just the component parts. While both the front end and baseband architectures are designed to be energy efficient through configurability and re-programmability, IMEC is also exploring how an overall management layer can minimise total energy consumption through co-ordinating these and other scalable elements, such as ADCs and video codecs. And of course IMEC is looking closely at making the reconfigurability automatic, creating Cognitive Radio.

No Ivory Tower

The SDR programme has been developed with industry partners, not as an ivory tower exercise. It was chosen as it is a very challenging area requiring a range of different skills drawn from across IMEC and its industry partners. Just a quick sample includes analog and digital design at demanding process nodes, development of design and programming tools, understanding of the difficulties of RF, and the complexities and subtleties of implementing the different radio standards.

And that is the point of IMEC. If it is going to fulfill its mission to be three to ten years ahead of industry needs, and then it has to attempt the hard things, not just in the clean rooms but in the products that the production clean rooms of 2011 will be producing in volume.

I am not certain what it was that the Flemish government thought it was setting up twenty-five years ago. What they have now is one of the electronics world’s most important centres for bringing together the skills and expertise of academia and industry for products and ideas that will work in the real world.

Click here for printable PDF
(By clicking on this link you agree to Embedded Technology Journal's Terms of Use for PDF files. PDF files are supplied for the private use of our readers. Republication, linking, and any other distribution of this PDF file without written permission from Techfocus Media, Inc. is strictly prohibited.)

Dick Selwood, Embedded Technology Journal

July 17, 2007

[back to top]


RECENT COMMENTS ON WWW.JOURNALFORUMS.COM:

All material on this site copyright © 2003-2007 techfocus media, inc. All rights reserved.
Embedded Technology Journal
Privacy Statement