Our research group at the InterUniversity MicroElectronics Center (IMEC—more information about IMEC can be found on www.imec.be) in Leuven, Belgium is working on the simulation of microelectronic fabrication processes and microelectronic devices. Somewhat misleadingly, this is usually called Technology Computer Aided Design (TCAD). Simulation is gaining importance in the microelectronics industry as processes and devices are becoming more complex. Real prototype fabrication currently takes too much time and money in the development cycle for new technologies.
However, producing a good fabrication plan for a new semiconductor technology is difficult. Numerous parameters in the plan, such as furnace temperatures, process duration, implant doses of impurities and so on, influence the resulting semiconductor microstructure. Also, the electrical characteristics of the diodes, transistors and capacitors depend on the geometry and impurity of the contents of the microstructure. The engineer is responsible for coming up with a fabrication plan, such that all the electrical characteristics of the devices not only meet their specified values but also exhibit a minimal sensitivity towards the uncontrollable fluctuations on the numerous process parameters. Trial and error methods, combined with physical insight, were once the only tools available for engineers.
In order to help the engineer on this laborious and grave task, we developed a software package called NORMAN/DEBORA. Input to NORMAN/DEBORA is a template design for the fabrication plan, which contains a likely range for some of its parameters. The package then fires off simulations of some well-chosen fabrication plans and constructs mathematical models for the electrical characteristics.
The engineer can then interpret the template fabrication plan using these models. She will find the critical parameters and will see how the electrical characteristics are linked, whether the specified values are feasible and, of course, which parameter values could be used to achieve this. The engineer can judge the sensitivity of the electrical characteristics when parameter variations occur, as in real fabrication. This will determine the robustness of the fabrication plan. Finally, NORMAN/DEBORA can perform an optimization of the fabrication plan, taking into account specified electrical characteristics and parameter sensitivity.
Most engineering software currently runs on the more powerful Unix systems of various breeds, and so does NORMAN/DEBORA. When the product was leaving the research phase, demonstrations were held for potential customers. At first, we envisaged the microelectronics industry as the sole user. However, the general concept of iterative simulations, plan evaluation and optimization could also be applied to other fields of engineering, such as injection molding of plastics and mechanical analysis of dynamics, acoustics and vibrations.
Finding a place in the sun in the market for engineering software proved to be tougher than we expected. We visited customers, gave demonstrations and prepared successful examples on the customers' particular types of engineering projects, using their engineering software. For easy travelling, and to avoid installation procedures, we decided to install the software on a portable machine. Linux was the most attractive operating system, because it installs on a small, inexpensive portable PC. Since Linux itself is free, using it would be extremely cost-effective, with no compromises in the required functionality.
Linux was hooked up to our network of Unix workstations. In order to do this, we needed the PCMCIA driver package in addition to the common distributions. Currently, we are using FTP for transferring the source code, and remote login for working on the machine. To integrate the machine even further, nfs could be used, but it would require reconfiguration when the portable is not connected.
We can run the simulators from our portable machine using the remote execution facility of NORMAN/DEBORA. However, most companies don't allow us to connect our portable to their network.
For porting, Linux has a POSIX-compliant programming interface and a good compiler for the C language readily available. As some older and mathematical modules were written in FORTRAN, we needed f2c to complete the porting since there was no full compiler for FORTRAN available at that time.
Porting our graphical user interface required an X server for our graphical hardware, which needed to be selected with care. As the user interface had been developed with Tcl/Tk, this step also went smoothly. We are using gnuplot to generate graphs and pbmtoxbm from the netpbm graphics format conversion library to display these as X bitmaps. As some companies don't want to have free software installed on their machines, we intend to replace this graphical interface, using widgets from the Motif library, which is available for Linux at a low price.
- Linux Kernel Testing and Debugging
- Tails above the Rest, Part III
- Wanted: Your Embedded Linux Projects
- NSA: Linux Journal is an "extremist forum" and its readers get flagged for extra surveillance
- The 101 Uses of OpenSSH: Part I
- Docker: Lightweight Linux Containers for Consistent Development and Deployment
- RSS Feeds
- Dolphins in the NSA Dragnet
- Are you an extremist?
- Tails above the Rest, Part II