CORDIS Archive
This page has been archived. It will no longer be updated.
Report of the
Mixed-Signal Design Workshop
Brussels, July 14, 1997
Technologies for Components and Subsystems (TCS)
"Fostering Excellency in Design Skills and Broadening their Use"
Participants
M. Declerq, EPFL
F. Dielacher, Siemens
E. Dijkstra, CSEM
J. Faura, SIDSA
J. Franca, IST/ Chipidea
G. Gorla, Italtel
H. Gschwendtner, Bosch
Q. Huang, ETH-Z
J. L. Huertas, CNM Sevilla
E. Janssens, Alcatel-Mietec
P. Jespers, UCL
R. Kempe, AMS
F. Montecchi, Univ. Pavia
A. Roggero, SGS Thomson
E. Tiittanen, Nokia Mobile Phones
V. Valence, MEAD
R. van de Plassche, Philips
H. Forster, EC
D. Broster, EC
M. Hohenbichler, EC
P. Reynaert, EC
C. Maloney, EC
The aim of this one-day workshop was to discuss trends in mixed-signal design in preparation for the addition of the mixed-signal theme to the ESPRIT Design Clusters action.
Representatives from European industry and academia were asked to identify driver applications for mixed-signal design in 3 years from now, the demands these applications will pose and which of these demands are most likely to be met.
In line with the goal of broad use of design skills, they were also requested to present their ideas on how an effective information flow within the mixed-signal design community can be maintained and what level of information is needed.
Mixed-signal is taken to refer to integration of digital and analogue functions on a single chip.
Mixed-signal ICs offer the possibility to integrate analogue with digital processing and memory functions. The main challenge is to achieve integration on a single IC both effectively and cheaply. Single-chip solutions generally consume less power and take up less space.
The market for mixed-signal ICs is forecast to reach 19,6 B$ by the end of the century as mixed-signal ICs represent an increasing fraction of the overall IC market. The emergence of new communications markets was identified as the major factor driving this growth. Communications devices comprise an interface to the analogue environment – via a wired or wireless link - combined with some form of digital signal processing.
The communications marketplace has seen tremendous growth in wireless and mobile consumer products and applications, among them: wireless transceivers (RF and HF), cellular phones and base stations, cordless phones, pagers, LANs and WLANs, mobile radios, GPS, faxes, modems, ISDN and xDSL, DVD and set-top boxes, cable modems and wireless multimedia.
These are the kind of high volume applications that can support the development of dedicated ICs. The development of essential building blocks is required, such as, data converters, low noise amplifiers, mixers, filters, synthesizers and DSPs, that are suited for on-chip integration. Integration of converter technology with analogue has now become commonplace. Integration of state-of-the-art converter technology and DSP technology is being driven by computer audio and computer multimedia products as well as video and mobile radio.
Other areas expected to present a growth opportunity for mixed-signal ICs include biomedical and automotive applications, where microsystems, sensors, and signal conditioning ICs are finding increasing usage. Smart power ICs, which combine power drivers with control logic and/or feedback signals, will be used in automotive power supplies and motor control applications. Contactless data transmission applications such as smart cards and tags also require small, integrated transceivers and processing functions. Particular requirements for these applications will be microsystems (analogue sensors or actuators with baseband or RF signals), data converters, front- and back-end processing, on-chip voltage generation, multi-level voltage supply, low noise and linearity.
Finally, disk-drives, though low-key, account for a significant number of mixed signal ICs.
The evolution of these applications will depend on the availability of innovative mixed-signal ICs; their competitiveness will depend on the performance of these ICs.
Challenges for mixed-signal design can be categorised into three areas: i) methodology ii) performance and iii) process. Mixed-signal CAD tools are addressed within each area.
i) methodology
Full custom, mixed-signal ICs can provide highly integrated solutions on a single chip. For these large ICs, only a top-down design methodology will enable product development and delivery within the increasingly shorter time-to-market constraints.
Top-down design techniques for full custom, digital ASICs are well established. Subsystems are described using VHDL, then simulated, and finally synthesized against technology and cell library. In analogue design however, there is a lack of hierarchical design tools and the approach still used is bottom-up. This has impacted developments in mixed-signal design, where the need for a system level methodology with systematic design rules supported by top-down CAD tools was emphasised. Mixed-signal, multi-level simulators with acceptable simulation times are needed in order to allow an efficient design flow and high level verification of functionality and connectivity.
At the system level, partitioning (of functional blocks, hardware/software, analogue/digital, onchip/offchip components) is an area where trade-offs between efficiency and performance must be optimised. In particular, optimisation of transceiver architecture will remain a challenge.
In digital design, reuse of building blocks can increase design productivity and offers a solution to decreasing time-to-market. In analogue (and mixed-signal), retargetting of blocks to different processes, i.e. decoupling of cores and libraries from the fabrication process, is an approach which could be given consideration, in particular to meet increasing demand and complexity.
A block driven methodology with dedicated analogue and digital macros in a platform independent format would improve design productivity and shorten design cycles. The need for commercial IP blocks so as to enable a front-end approach was stressed. A challenge here is robustness against parameter tolerances.
For test, it is crucial to be able to identify problems as early as possible in the design flow. Design for test and automatic test generation may help facilitate this as well as ensure improved reliability. It was also pointed out that compatible test methodologies for analogue and digital are needed.
ii) performance
With portability and mobility becoming increasingly important, there is a demand for low power/low voltage while maintaining low noise and high-speed capabilities. Tools which allow analysis of substrate and supply coupling, and parasitic coupling effects, are required for applications with low level analogue RF and digital baseband signals. Optimisers and power estimators are also necessary for low power applications. In transceiver design, it is difficult to isolate the high level RF transmitter signal from the receiver – in particular on a single substrate. Modelling of digital noise and degradation of analogue performance due to digital noise is an area needing attention. For high speed circuits, modelling and simulation of crosstalk and ground noise, power bus noise and delay estimation are necessary.
iii) process technology and device modelling
CMOS offers low power consumption and low cost for both analogue and digital ICs and is seen by many as the process technology of choice for mixed-signal. As CMOS scales to lower dimensions, the transit frequencies it achieves increase (.35µm channel lengths allow 5 GHz operation), making CMOS RF an active research topic. However practical issues needing attention are substrate coupling, parameter variation with temperature and process, and device modelling for RF. Presently there are no commercially available CMOS RF ICs used in communications products. The issues of system-level performance (noise and high output power), design and packaging must be improved before they can compete with current solutions. It is desirable in any case that mixed-signal remains compatible with the evolution of digital CMOS technology, in particular as more than 90% of the total (mixed-signal) chip area is generally occupied by digital circuitry. There remain applications which will require higher frequencies than standard CMOS can provide. SOI CMOS, BiCMOS and SiGe are expected to provide solutions for higher frequency applications. SOI may additionally find use in applications requiring precision analog functions due to its dielectric isolation nature and its potential to offer reduced substrate noise.
Of the driver applications identified previously, wireless transceivers for cellular phones and broadband communication transceivers for digital subscriber lines (xDSL), cable modems and set-top boxes are probably the most significant in volume terms.
It follows that these are the areas where most significant advances are expected, in particular:
Advances in microsystems technologies area also expected to impact mixed-signal design with the emergence of high performance micro-machined sensors and data acquisition systems on the same chip.
Each design experiment will require exchange of technical information between participants to take a design from specification to successful (and often state-of-the-art) performance. Industrial companies may, either take up practices to-date only applied in research, or, utilise resources (methodologies, flows, tools plus expertise) available at universities; universities gain the system/application level of information which will ultimately make their graduates the best candidates to be taken on by industry.
The question arises as to what type of technical information can be captured from design experiments for take up by the broader design community. For the case of mixed-signal design, the view is that improvements to methodologies and design flows rather than tool-based design automation will lead to better designs.
The situation in industry is that the pressure is on designers achieving results rather than on developing methodologies. As designs become increasingly complex, improvements to methodologies, design flows and tools ensue. However this is not happening in a way such that these improvements are exchangeable between companies – or even designers themselves. Herein lies a task for the design cluster.
The clustering of design experiments provides for participants a framework for the exchange of technical information on mixed-signal design practice and methodologies. Exchange of information, while it may serve to educate, is not sufficient to stimulate broad use of better design methodologies and techniques. A more effective approach – afforded by the cluster – will be to use information captured from numbers of design experiments to contribute to the systematisation of design rules and structuring of methodologies.
This will require individual design experiments to provide in a detailed format a suitable description of the methodology, design flow or tool together with advantages, limitations and comparisons with previous approaches. The availability of such information for incorporation into courses and workshops or a cluster web page will contribute, albeit to a limited extent, to the goal of broad use. The cluster coordination should go a step further in that it could take on the task of generating methodological rules from this information.
This should allow the reuse of knowledge generated within design experiments and remove the barriers to its broader take up.
More specific types of design techniques may require a different approach. For such cases, rather than to expect broad take up of the techniques by industry, the maintenance of links to, and use of, specific resources – methodologies, flows, tools plus expertise - should be fostered.
In summary, a more systematic (and hence more reusable) methodological approach is a key requirement in enabling better design practice on a broad scale, allowing not only compatibility of design approaches but also, ultimately the development of IP which can be reused within and between companies.