Time-Dependent Variability: A test-proven modelling approach for systems verification and power consumption minimization

Following the Moore's law the semiconductor industry has delivered continuous increase of systems functionality and speed over the last 50 years through the aggressive downscaling of the transistors. In the last 20 years the UK IC-design based industry has grown to a level of national and international importance. While IC designers in the past enjoyed the freedom that all transistors in a chip could be treated identically, this is no longer the case for the nano-meter sized transistors used in the present and future technologies. Statistical device-to-device variation is introduced by the discreteness of charge and granularity of matter and is inversely proportional to gate area, so that its impact on circuits increases with the reduction of transistor dimensions. When the number of logic gates in a system increases and the architecture becomes more complex, the tolerance to variability is greatly reduced. Even if two devices were identical after fabrication, they could suffer from different aging during operation, causing a time-dependent variability (TDV). TDV is becoming a major threat to the correctness of electronic systems, but there are no tools for its verification because of the lack of a complete understanding.

The aim of this project is to carry out an in-depth investigation of the defects and mechanisms responsible for TDV and, based on that, to develop a test-proven TDV simulator, allowing IC designers to assess the impact of TDV on their circuits. The researchers at Glasgow University have pioneered variability simulation and the researchers at Liverpool John Moores University have specialised in experimental characterization of defects. Their highly complementary skills bring them together and make them well positioned to tackle this challenge. By working together with UK companies, the impact of their work on UK industry will be direct. The collaboration with IMEC and its industrial consortium also opens an effective impact pathway on an international scale. The successful control of TDV will deliver reliable electronic products and minimize their power consumption.

To optimize circuits' performance, IC-designers need an in-depth knowledge of the devices used in the state-of-the-art fabrication technologies. This project provides such knowledge in the area of time dependent variability (TDV) of nano-meter transistors. By developing new measurement techniques, test and simulation methodologies, and models, this project aim to provide tools for circuit designers to assess the impact of TDV on their circuits. The pathways to impact can be summarised below.

Direct impact on UK industry: In this proposal, we will work together with two large companies in this industry: Arm and CSR. In the first half of the project, the main research is of fundamental nature and the UK industrial partners will provide steering advice to ensure the relevance of the research to industry in this period. After laying the foundation, the research in the second half will be increasingly weighted towards qualification of the model in a circuit environment and the industrial involvement will intensify. Some work packages in the project were designed after consulting with the partners and the initial test circuits were selected together. By working together with industrial partners and performing tests directly on their samples, the output of this project will have a direct impact on UK industry.

International project partners based at IMEC: The industrial research consortium based at IMEC includes virtually all top semiconductor manufacturers and the project team has collaborated with IMEC over 20 years. In this project, IMEC will supply test samples, not only used in current technology, but also in development for the future technologies. IMEC will steer the project from the angel of future chip fabrication and development. The inclusion of IMEC adds an international dimension and an effective pathway for direct dissemination of the research outputs to international companies.

Test equipment and software suppliers: A major challenge for the time-dependent variability (TDV) tests is its huge number of test samples and data. The team will work together with the test equipment suppliers to find the best solution to this problem. It is expected that the existing equipment cannot provide the solution in a cost-effective way and suggestions will be made to the suppliers how to develop future equipment, in anticipation that TDV will become an essential qualification specification for designers and manufacturers in future. Gold Standard Simulations Ltd will provide pathway to the commercial exploitation of the models and software developed in the project. It will help with the productisation of the modelling techniques developed in the project.

Communications to public user groups: The non-IP research results will be published on Journals and Conferences of high industrial impacts. It is planned to hold workshops for training users to use the TDV simulators to be developed in this project. Research seminars/presentations will be given to the targeted companies and audience. Exhibitions will be arranged to demonstrate the simulators. Communications also will be through various networks and committees the team is involved.

Society: The general beneficiaries are the Electronics and Education sectors. The public will benefit from the successful control of TDV by having more reliable products with more functionality and higher speed. A detailed knowledge of TDV also allows the minimization of operation power of electronic systems, reducing CO2-emission and helping in the battle against climate change.