Design for Test: For Digital Integrated Circuits

• Authors: Alfred L. Crouch
• Venue: Design for Test: For Digital Integrated Circuits
• Year: 1999

From the Book:PREFACE: PrefaceThis book is made primarily for design engineers and managers, and for test and design-for-test engineers and managers. It can also be used for students of digital design and test, as well. The purpose of this book is to introduce the basic concepts of test and design-for-test (DFT), and to then address the application of these concepts with an eye toward the trade-offs of the engineering budgets (silicon area, operating frequency target, power consumption, etc.), the business drivers, and the cost factors.Currently, some very good test and DFT texts are available. However, many of them are from an academic focus. In my years of being part of the integrated circuit design community, I have had to train many IC designers and junior DFT engineers in test and design-for-test. I have discovered that corporate education is remarkably different from academic education. A design engineer on a project, who must learn and apply DFT techniques, is learning them while being responsible for 60+ hours of other design tasks per week and while meeting regular design deadlines and milestones. In this environment, learning the DFT tasks and techniques is difficult with a book that focuses on the "mathematical" or "theoretical" point of view. History has taught me that a direct "how to do it" text is more effective.Another interesting aspect of the competitive corporate environment is that the design process may be "factory-ized." The overall design process for a chip or a portion of a chip is no longer the responsibility of the design engineer, but of teams of chip design functions. For example, the logic gate cells may be designed and characterizedbyone group (standard cell and library development), and the design may be modeled and synthesized by a different group (HDL design and synthesis), verified by yet another group (formal and simulation verification), and ultimately, mapped to a physical process by yet another group (floorplanning, place&route, and physical design). In this case, the teaching of DFT techniques must be spread out to the various organizations contributing to the overall design. A teaching description of a DFT technique, such as scan design, is not effective if it is not related to the tasks, scheduling, trade-offs, and the separations into the various organizational elements. Again, history and experience have taught me that an effective text here is one that relates the topic being addressed to the design methodology and design flow.So direct experience in corporate technical training and teaching has led to the style and content of this "practical guide" on the test and Design-for-Test (DFT) topics of scan test, embedded memory test, and embedded core test. This text has been developed more along the lines of a "just what you need to know—and how to do it" guide that explains the topic, explains the trade-offs, and relates the topic to the design flow. My hope is that using this text will reduce the "learning curve" involved in the application of test and design-for-test techniques, and will result in designs that have a higher quality-reliability level and a lower cost-of-test.A practical text on DFT and DFT techniques, based on the industry point of view, is needed right now for several reasons. First, the "cost of test" is beginning to dominate the recurring (per-part) cost involved in the manufacturing of the final silicon product for many of the consumer markets—parts with high volume and a low selling price. Second, shorter product lifetimes and increased time-to-market (TTM) and time-to-volume (TTV) pressures are forcing the need to have some form of structured, repeatable and automatable test features included in the device as part of the overall design methodology. Third, the move to reuse cores, and core-based design, as a reaction to shrinking process geometries and TTM pressures, is also forcing designed-in test features to become portable since design units may be distributed and reused in several different chip designs with completely different configurations. And finally, the shrinking process geometries also enable "system-on-a-chip" and ULSI (Ultra-Large Scale Integrated) designs with massive integration—more integration means more faults and more vectors—which leads to a test data management and cost-of-test problems.Taken all together, these changes in parts of the semiconductor design industry are changing the way test and DFT are viewed, addressed, and implemented. Organizations that once ignored DFT are now being dragged kicking and screaming into modern age because of test cost, TTM, TTV, test data volume, and having to deal with the test features delivered with commercially available cores. Test is one of the three major components of recurring per-part cost involved with the manufacture and selling of digital semiconductor integrated circuits (with the cost of silicon die and the cost of packaging being the other two). As with every product, trade-offs are made to achieve a quality level and to fit within a target cost profile. I hope that this text will eliminate the view that understanding the cost-of-test and applying DFT during the design phase of a product is a "black art" for organizations and individuals that must address managing the cost factors of a chip design.If you have questions or comments, I can be contacted at [e-mail]