CORDIS Archive
This page has been archived. It will no longer be updated.
"Improving Design Productivity"
27 June 1997
Participants
J.-M. KARAM TIMA B.D. JONES RAL Jan SUSKI Schlumberger J.J. SIMONNE LAAS Georg PELZ University Duisburg Frédéric RENOUX Dolphin Integration Marc BELLEVILLE CEA Leti Werner RIETHMÜLLER Fraunhofer ISIT Stein Ivar HANSEN SensoNor Henrik JAKOBSEN SensoNor Trond Inge WESTGAARD SINTEF Gerard KELLY NMRC Arno HOOGEWERF CSEM Dirk BEERNAERT CEC Philippe REYNAERT CEC Jorge BLASCO CEC
1. Introduction.
The objective of the "Workshop on productivity in Microsystems Design" was to identify areas of priority for R&D in design tools and methods, addressing the gaps in the design flow that make todays' microsystems design efforts a handcrafting activity just as microelectronics was some years ago.
As a summary, today the design of complex integrated microsystems is only possible by skilled people (comparable to high-end analogue ASIC design), and involves large efforts, costs and risks of several expensive design iterations.
Section 2 describes the typical design situation today and identifies the needs in CAD tools, section 3 summarises the conclusions of the workshop and recommends activities for the future.
2. Description of the needs.
Microsystems are designed today using specialised Finite Element Modelling (FEM) packages like ANSYS or SOLSTIS (same as in the technology development step), the results of these simulations are ported to analogue behavioural models describing the physical equations (HDL-A, Verilog-A, VHDL-AMS,...) that afterwards are used for simulations in EDA environments. The FEM tools used for different physical phenomena vary according to their type (mechanical, optical, gas sensors, microfluidics, etc., etc.).
A complementary approach models the microsystem structures through partial differential equations (PDE), and solves the equations through industrially available software packages like SMASH-MEXEL and Emblem. The same approach is being ported to non-mechanical processes (heat transfer, wave propagation, diffusion,...).
Both approaches claim to be able to model a microsystem structure in a timeframe of some weeks. In both approaches the confidential character of the foundry parameters and models is guaranteed thanks to compiled code that cannot be read by the end-user.
There is a very large portfolio of existing microsystems cells (pressure sensors, accelerometers, gas sensors, micropumps, thermopiles, flow sensors, etc., etc.), each one with a different structure. As a consequence of the lack of maturity of design procedures and tools, microsystems libraries would enable non-specialists to have access to an existing portfolio of cells provided by the manufacturer (similar to standard-cells in microelectronics).
Regarding simulation, the key issue identified has been the need for accurate device models whose parameters might be tailored to different technologies. In some cases such as gas sensors and piezo-electric devices, the models are not accurate enough.
Microsystems' vendors recognise the need to supply their customers with Cell Libraries and design kits but acknowledge that today they sell directly to end-users and redesign themselves parts of the layout for some specific customers. The desirable change from full-custom to ASS (Application Specific Sensors) has to be supported by standard EDA tools. Those design kits should include all the interfaces between the Microsystems' technologies and processes and the CAD environment: design rules, library elements, parameterizable layouts, component models [functional and geometrical], basic HDL-A models, simulation-model parameters [electrical, mechanical,...], etc.,etc.
While the use of parameterisable cells has been quite extensive within single organisations, those cells have not been made widely available to customers as part of an EDA package or as part of a design kit running over standard EDA environments. The models should be generated for existing devices by collaboration of skilled process engineers from the manufacturers and algorithm and software engineers from CAD vendors.
There was wide agreement on the need to make the microsystems tools as compatible as possible with microelectronics EDA tools, it was also acknowledged that behavioural models (HDL-A like) seemed to be a good link between both worlds, and that a complementary approach was to use PDE models (Partial Differential Equations).
In general no strong claims for new tools were expressed, more emphasis was placed on the need for design kits from the microsystems manufacturers running over off-the-shelf EDA platforms. That did not prevent the attendees from listing their wishes for future tools:
Some doubts were expressed about the need for a "light" PC-based design tool that would allow non-experienced designers to participate in some parts of the design that afterwards would be finished with the help of specialised skills. This approach has to be differentiated from the high complexity and price packages now available on high performance PCs and functionally identical to those ones running on UNIX workstations.
Some of the attendees claimed the need for process simulation tools, while others disagreed arguing that the market for those tools is extremely limited (tens of units in Europe). There was not a common opinion on the manpower needs for the development of a process simulation package, while some considered that between 4 or 5 person-years would be enough others thought it would be more of the order of several tens of person-years. The market for 2.5D anisotropic etch simulator (Front-Back) and 2D-3D translators is also very small (only foundries).
3. Conclusions and recommendations.
The URL for this page is /esprit/src/tcsmicro.htm
It was last updated on 15 October 1997, and is maintained by Colette Maloney - Colette.Maloney@dg3.cec.be