Please use this identifier to cite or link to this item:
http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/41774
Title: | 應用於睡眠電晶體之電流模型的靜態時序分析 Sleep Transistor-Aware Modified Composite Current Source Model |
Authors: | Shang-Chien Pai 白尚謙 |
Advisor: | 陳中平(Chung-Ping Chen) |
Keyword: | 漏電功耗,超大型積體電路,睡眠電晶體,電流源模型,態時序分析, VLSI,Leakage Power,Sleep Transistor,Current Source Models,Static Timing Analysis, |
Publication Year : | 2009 |
Degree: | 碩士 |
Abstract: | 隨著製程技術不斷的進步,特徵尺寸不斷的縮小,目前45nm甚至是32nm的製程已經被大量且廣泛的使用。伴隨著製程的進步,漏電的情形也因為閘氧化層的厚度不斷變薄而日益嚴重。最新的研究指出,漏電功耗已經佔了整體功率的約 42%之多,成為了設計超大型積體電路不得不解決的問題。
睡眠電晶體(Sleep Transistor) 是一個被廣泛用來解決漏電的技術。他的原理是使用睡眠電晶體將不執行所需計算的邏輯閘關閉去改善漏電的問題。然而,由於睡眠電晶體兩端的電壓下降,導致傳統的電流源模型(Current Source Model)無法有效的準確靜態時序分析。 在本篇論文中,我們提出了一個改良的電流源模型(Modified Current Source Model)來同時模擬標準邏輯閘的輸出以及睡眠電晶體兩端的電壓下降的效應。實驗結果顯示我們跟HSPICE 模擬軟體的平均延遲誤差僅僅只有2%以下。 As the semiconductor technology advances, feature sizes continue to shrink. No-wadays 45nm technology and even 32nm technology have been largely employed in the design of modern Integrated Circuit Design. However with the advancement of shrink-ing feature sizes, comes the problem of leakage current. As the gate oxide thickness be-coming thinner with the advancing technology, this problem can no longer be ignored. The newest research results demonstrate that the leakage power accounts for 42% of total power and thus must be reduced for the design of VLSI circuits. Sleep transistor is a vastly used technique for reducing leakage power. The theory behind it is to simply cut off the gates that are not in use to further reduce the leakage power. However the traditional Current Source Models for Static Timing Analysis have no consideration for the effect of the addition of a sleep transistor. This limits the design ability of circuits with sleep transistors and introduces uncertainties in the design. In this thesis, we proposed a Modified Composite Current Source Model to not only model the output of standard gates but to model the effect of a non-stable ground. In addition this model demonstrates the ability of persevering the output waveform from stage to stage. There is no need of recalculating an input transition value after ever stage. Experimental results compared with HSPICE results show an average delay error of less than 2%. |
URI: | http://tdr.lib.ntu.edu.tw/jspui/handle/123456789/41774 |
Fulltext Rights: | 有償授權 |
Appears in Collections: | 電子工程學研究所 |
Files in This Item:
File | Size | Format | |
---|---|---|---|
ntu-98-1.pdf Restricted Access | 2.6 MB | Adobe PDF |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.