Electrical characterisation of matched pairs for evaluation of integrated circuit technologies

Abstract

This thesis is about evaluating differences between electrical properties of closely spaced, supposedly identical, transistors (matched pairs) in integrated circuits (IC's). Electrical properties of transistors in IC's generally exhibit relatively large spreads due to variations of production equipment. Such spreads limit the attainable performances of IC's and deteriorate the precision by which certain (analogue) signal processing algorithms can be realised. Closely spaced identical transistors in one IC however, are fabricated simultaneously and under practically identical conditions, which results is substantially smaller differences (mismatch) of electrical performances. This attribute, usually indicated with the term 'matching', is used in many IC's to realise better performances. The remaining mismatches between transistors of matched pairs are generally of a stochastic nature and are hence indicated with the term mismatch fluctuations. By measuring and characterising populations of matched pairs, mismatch fluctuations can be quantified and evaluated. The observed effects are generally attributable to fluctuations of microscopic transistor construction elements such as doping atoms, edge roughness and grain boundaries. Besides these stochastic effects, so-called systematic mismatch effects are frequently encountered in practice. These are caused by the fact that transistors of a matched pair cannot always be realised or used identically. This thesis discusses the required microelectronic test structures and measurement capabilities for correctly analysing and evaluating mismatch effects. The discussed examples demonstrate that mismatch studies can provide valuable insights into the, not always equally well chosen, microscopic architecture of transistors. The discussed techniques and insights are used to optimise IC technologies and lead to better electronic performances and higher product yields.