Logo image
Built-in Timing Parametric Measurement
Dissertation

Built-in Timing Parametric Measurement

Ming-Jun Hsiao
Doctor of Philosophy (PHD), 國立清華大學, 電機工程學系
2003

Abstract

時間參數量測 雙斜率轉換 雙指數曲線轉換 設定時間 保持時間 時間數位轉換器 安定時間補償 timing measurement dual-slope dual exponential curve setup time hold time time-to-digital converter (TDC) settling time canceling
As the progress in deep sub-micron technology, the integration of more heterogeneous components into a System-On-a-Chip is performed for reducing production cost. However, the testing cost is increased rapidly due to lack of controllability and observability, and thus general test access mechanism and high-performance Automatic Test Equipments (ATEs) are needed to overcome the testing problems. Furthermore, the required general test access mechanism usually introduces extra timing penalties for parametric testing, and hence reduces the accuracy of AC measurement. In this way, embedded measurements are needed to accurately measure the parameters inside the chip with the cooperation of low-cost ATEs for reducing test cost. Based on the pulse-stretching principle, two types of timing measurement circuits are proposed in this thesis for embedded measurement. The first one is founded on traditional dual-slope conversion while considering the non-ideal effects such as settling time to extend the measurable range below 0.5ns. Since a self-timed method is proposed to reduce the requirements of controller, the area is only 72mm X 43mm n a 0.25mm 1P5M CMOS technology, and the capability of setup/hold time measurement is provided in the meanwhile. The second one is a brand-new method based on dual-exponential curve transformation. By utilizing the inherent exponential curves, high-linearity components are not necessarily required and the design efforts are relaxed. Furthermore, the dual-exponential curve transformation needs no reference time intervals for calibration, and hence it immunes to the uncertainties such as jitter of the reference inputs. A design guideline is also proposed for this method based on the error analysis.

Metrics

1 Record Views

Details

Logo image