In current manufacturing processes, certain layout configurations are likely to have reduced yield and/or reliability due to increased susceptibility to stress effects or poor tolerance to certain processes like lithography. These problematic layout configurations need to be efficiently detected and eliminated from a design layout to enable better yield. In this paper, such layout configurations are called processhotspots and an efficient and scalable algorithm is proposed to detect such process-hotspots in a given layout.The concept of a range pattern is introduced and used to accurately and compactly represent these process-hotspots. This representation is flexible and can incorporate information about the deficiencies of available modeling and/or subsequent correction (for instance, mask synthesis) techniques. Each range pattern can also be associated with a scoring mechanism to score the problem regions according to yield impact. A library of range patterns is being developed in collaboration with a fab. The proposed process-hotspot detection system assumes that process-hotspots are specified as a library of range patterns and determines all occurrences of any of these range patterns in a layout. It is fast and accurate and can be applied to large industrial layouts. Unlike previous work, the proposed scheme can identify problems that cannot be efficiently modeled or corrected by subsequent mask synthesis techniques and can thereby complement existing work in that area. Experimental results are quite promising and show that all locations that match a range pattern in a given layout can be found in a matter of minutes.

No abstract

No abstract

No abstract

Ahstract-In this paper, we present a placement method for analog circuit. This method can handle multiple constraints, which include multi-symmetry, proximity, signal-path clustering constraints. All these constraints are common requirement in analog circuit design. However, none of previous works considered the relations between constraints, and how to handle these multiple constraints effectively to get a convergent solution is also a seldom solved problem in previous works. Our contributions are threefold: 1) introduce signal path constraints and handle all constraints with Sequence Pair representation; 2) present a multi-stage simulated annealing strategy to acquire a faster convergence of searching process; 3) first introduce a post refinement stage to further improve the result placement, using linear programming. The effectiveness and efficiency of experimental results are promising.