STRATASkip to the research
Right, original XLVII plate

Right

In 1976 Taiwan’s Industrial Technology Research Institute sent engineers to learn integrated-circuit design, fabrication, testing, and equipment maintenance under an agreement with RCA. The demonstration factory completed in 1977 established Taiwan’s first integrated-circuit production line.

Jensen Huang and Lisa Su, both born in Taiwan, received the National Medal of Science for distinct contributions to computing. Huang’s work extended the graphics processor to general computation; Su’s advanced processor design and chiplet integration. The transfer of a design and the reproducibility of its manufacturing process involve different evidence, including equipment, process settings, yield, and failure records.

Historical record

  1. A proposed industry begins as a specific development project.

    ITRI’s institutional record describes the meeting that led to a Taiwanese integrated-circuit development plan. The account names participants and a subsequent written proposal, providing a traceable start to the capacity-building effort.

  2. A transfer agreement is paired with training.

    ITRI and RCA sign a technology-transfer and licensing agreement. Teams study design, fabrication, verification and equipment; the knowledge is divided among practical disciplines that must eventually work together.

  3. A demonstration factory turns training into production experience.

    ITRI records completion of the demonstration factory and Taiwan’s first IC production line. The significance is a working process that engineers can diagnose and improve, beyond possession of a design or a signed agreement.

  4. The new process reaches a manufactured application.

    The first batch of locally made watch ICs is completed. The product gives the process a concrete destination and joins technical learning to an economic use.

Yield improvements need a cause that another shift can reproduce.

Consider an illustrative lot with 100 tested dies, 70 passing and 30 failing. After a process change, another 100 yield 80 passes. The ten-percentage-point increase is worth investigating, but two small lots do not establish why it occurred. A useful record includes wafer position, test version, process settings and failure category. This example does not reproduce ITRI’s historical yield data; it asks what evidence would make a modern improvement credible.

  • Map failures by location and category before averaging them into one percentage. A cluster can suggest a different investigation from scattered failures.
  • Repeat the comparison with a declared sampling plan and comparable lots. Record every accompanying change that might explain the result.
  • Give the next shift the settings and diagnostic evidence needed to reproduce the gain. If only one operator can obtain it, the process still contains knowledge that has not been transferred.

Recordings and archives

Research sources

  1. Adopting Technical Knowledge from RCA to Develop Taiwan’s IC Capabilities

    Industrial Technology Research Institute · ITRI’s institutional record of the 1974 discussion, 1976 RCA transfer, 1977 factory and 1978 watch ICs.

  2. Evaluation of measurement data—Guide to the expression of uncertainty in measurement

    Joint Committee for Guides in Metrology / BIPM · JCGM 100:2008, reporting a result with a model and evaluated uncertainty.