In this paper we present the architecture of an intelligent test error detection agent that is able to independently supervise the test process. By means of rationally applied bin and cause specific retests it should detect and correct the majority of test errors with minimal additional test effort. To achieve this, the agent utilizes test error models learned from historical example data to rate single wafer runs. The resulting run specific test error hypotheses are sequentially combined with information gained from regular and ordered retests in order to infer and update a global test error hypothesis. Based on this global hypothesis the agent decides if a test error exists, what its most probable cause is and which bins are affected. Consequently, it is able to initiate proper retests to check the inferred hypothesis and if necessary correct the affected test runs. The paper includes a description of the general architecture and discussions about possible test error models, the inference approach to generate the test error hypotheses from the given information and a possible set of rules to act upon the inferred hypothesis.
| Identifer | oai:union.ndltd.org:DRESDEN/oai:qucosa.de:bsz:14-qucosa-83943 |
| Date | 20 February 2012 |
| Creators | Kirmse, Matthias, Petersohn, Uwe |
| Contributors | Technische Universität Dresden, Fakultät Informatik |
| Publisher | Saechsische Landesbibliothek- Staats- und Universitaetsbibliothek Dresden |
| Source Sets | Hochschulschriftenserver (HSSS) der SLUB Dresden |
| Language | English |
| Detected Language | English |
| Type | doc-type:workingPaper |
| Format | application/pdf |
| Relation | dcterms:isPartOf:Technische Berichte / Technische Universität Dresden, Fakultät Informatik ; 2012,02 (TUD-FI12-02-Februar 2012) |
Page generated in 0.0022 seconds