隨著積體電路(IC)製程愈做愈小，使得供應電路的電壓也變得愈來愈小，因此電路設計者必須考量到，供應電路的電源線(Power line)是否會因突發的變動或者其他因素而影響電路的運作情況，若能在電路設計的初期就能提早知道電源線的變動對整體電路所造成的影響，則有助於加速電路的開發並節省開發成本。 由於類神經網路(Neural Network)具有良好的預測效果，而離散餘旋轉換(DCT)讓我們將時間域上較不規則的電流波形轉到較有規則可循的頻率域上來做處理，因此本研究提出一個，結合了這兩種理論來建立高階模型的方法，實驗的結果顯示出，它是有一定的準確性存在著，即我們可以在設計電路的初期大約估計電源線電流對整體電路所造成的影響。 As IC process scale down, power supply is becoming small. Therefore, IC designer must consider the variation of power supply line that affects circuit operation. If we could know current information of power supply line in early design stage, then IC design may become fast and effective. Neural Network is good in estimation and Discrete Cosine Transform can translate irregular current waveform in time domain to regular form in frequency domain. Therefore, we propose a high-level model which combines Neural Network and DCT to reconstruct the current waveform of power supply line. The experiment shows it is possible to estimate current information in early design stage.
