고속 집적 회로 데이터 채널에 대한 실험적 특성화 및 효율적인 시뮬레이션 방법

Abstract

본 박사학위 논문에서는 고속 동작 집적 회로 데이터 채널에 대한 실험적 특성화와 효율적인 시뮬레이션 방법에 대한 연구를 수행하였다. 먼저, 전송선로를 특성화하는 새로운 방법에 대하여 연구하였다. 공진 문제와 주파수 종속 특성에 기인한 전송선로 파라미터 불안정 문제를 해결할 수 있는 방법을 제시하였다. 테스트 패턴은 패키지 공정을 사용하여 설계하고 제작하였으며 VNA를 사용하여 100 MHz 에서 26.5 GHz 까지 S-파라미터를 측정하였다. 패드 기생 성분은 디임베딩 방법을 사용하여 제거하였다. 주파수 종속 복소 유전상수는 데바이 모델(Debye model)과 전파상수를 사용하여 결정하였다. 공진 문제를 해결한 전송선로 파라미터는 주파수 종속 복소 유전상수와 ‘Eo 와 Eisenstadt 방법’을 사용하여 결정하였다. 다음으로, 두 신호선에서 신호천이와 누화 특성화 방법을 연구하였다. 두 신호선을 20 GHz 까지 4-포트 S-파라미터 측정을 하여 실험적으로 특성화 하였다. 대칭적인 두 신호선은 두개의 고유 모드로 분리하여 전송선로 파라미터와 신호천이 파형을 S-파리미터로부터 직접적으로 결정하였다. 또한 실험 및 분석 결과를 사용하여 전송선로 파라미터는 주파수 종속적이고, 주파수 종속적인 영향과 신호선의 비 이상적인 특성들은 신호천이와 누화에 상당한 영향을 준다는 사실을 검증하였다. 마지막으로, 단일 신호선과 두 신호선의 아이-다이어그램을 효율적이고 정확하게 결정하는 시뮬레이션 방법을 연구하였다. 부호간 간섭(ISI)을 고려한 가장 간단한 입력 신호 모델을 제안하였으며 이를 이용하여 두 신호선에서의 입력 신호 모델은 스위칭 조건에 따른 누화를 고찰하였다. 제안한 입력 신호를 사용하여 계산한 아이-다이어그램은 의사 난수 이진 수열(PRBS) 입력 신호를 사용한 스파이스(SPICE) 시뮬레이션 결과와 일치하고, 계산 시간은 스파이스를 사용하는 경우보다 매우 적게 소요된다는 것을 보였다.| This dissertation investigates the experimental signal integrity verification methods of interconnect lines concerned with integrated electronic system. First, a novel transmission line characterization method is presented. Due to inherent resonance effects and frequency-variant dielectric properties, up to date, stable and accurate circuit model parameters of thin film transmission line structures cannot be experimentally determined over a broad frequency band. A new, simple and straightforward frequency-variant transmission line circuit model parameter determination method is presented. Experimental test patterns for high-frequency transmission line characterizations are designed and fabricated using a package process. The S-parameters for the test patterns are measured using a vector network analyzer (VNA) from 100 MHz to 26.5 GHz. The parasitic effects due to contact pads are de-embedded. The frequency-variant complex permittivity is determined by combining the propagation constant with the Debye model. The resonance-effect-free transmission line parameters (i.e., the propagation constant and characteristic impedance) are then determined in a broad frequency band using the complex permittivity and ‘Eo and Eisenstadt method’. Second, a new signal transient and crosstalk noise characterizations of coupled transmission lines are investigated. Coupled transmission lines are experimentally characterized by using 4-port S-parameter measurements in a broad frequency band (up to 20 GHz). Symmetrically coupled transmission lines are decoupled into two eigen modes that can be readily determined from the measured S-parameters. Then transmission line parameters and signal transient waveforms are directly determined by using the measured S-parameters. It is shown that not only are the transmission line parameters frequency-dependent, but also the frequency-variant effects and non-ideal characteristics of transmission lines have a substantial effect on signal transients and crosstalk noises. Finally, a new efficient and accurate analytical eye-diagram determination technique of single and coupled interconnect lines is presented. The simplest input test signal model for the inter-symbol interference analysis of high-speed data links is mathematically formulated. The crosstalk effects depending on switching condition are taken into account for coupled lines. The proposed technique shows excellent agreement with the SPICE-based simulation. Furthermore, it is very computation-time-efficient in the order of magnitude, compared with the SPICE simulation, which requires numerous pseudo-random bit sequence input signals.; This dissertation investigates the experimental signal integrity verification methods of interconnect lines concerned with integrated electronic system. First, a novel transmission line characterization method is presented. Due to inherent resonance effects and frequency-variant dielectric properties, up to date, stable and accurate circuit model parameters of thin film transmission line structures cannot be experimentally determined over a broad frequency band. A new, simple and straightforward frequency-variant transmission line circuit model parameter determination method is presented. Experimental test patterns for high-frequency transmission line characterizations are designed and fabricated using a package process. The S-parameters for the test patterns are measured using a vector network analyzer (VNA) from 100 MHz to 26.5 GHz. The parasitic effects due to contact pads are de-embedded. The frequency-variant complex permittivity is determined by combining the propagation constant with the Debye model. The resonance-effect-free transmission line parameters (i.e., the propagation constant and characteristic impedance) are then determined in a broad frequency band using the complex permittivity and ‘Eo and Eisenstadt method’. Second, a new signal transient and crosstalk noise characterizations of coupled transmission lines are investigated. Coupled transmission lines are experimentally characterized by using 4-port S-parameter measurements in a broad frequency band (up to 20 GHz). Symmetrically coupled transmission lines are decoupled into two eigen modes that can be readily determined from the measured S-parameters. Then transmission line parameters and signal transient waveforms are directly determined by using the measured S-parameters. It is shown that not only are the transmission line parameters frequency-dependent, but also the frequency-variant effects and non-ideal characteristics of transmission lines have a substantial effect on signal transients and crosstalk noises. Finally, a new efficient and accurate analytical eye-diagram determination technique of single and coupled interconnect lines is presented. The simplest input test signal model for the inter-symbol interference analysis of high-speed data links is mathematically formulated. The crosstalk effects depending on switching condition are taken into account for coupled lines. The proposed technique shows excellent agreement with the SPICE-based simulation. Furthermore, it is very computation-time-efficient in the order of magnitude, compared with the SPICE simulation, which requires numerous pseudo-random bit sequence input signals.