積體電路需在合理的溫度下操作以維持在正確的工作狀態。由於超大型積體電路的元件密度持續增加,在20世紀末,一顆晶片的總功率消耗已經超過了100瓦。一些多晶片封裝的模組甚至可以消耗數千瓦的功率。所以,我們需要精密的冷卻方法去降低這些能量所產生的熱能。然而,任何一些週邊環境的變化或干擾使得晶片暫時脫離冷卻系統的控制,都可能造成過熱而導致系統永久的損害。 為了防止上述原因,特別的量測機制分別在產品測試和晶片操作期間建立。最直觀的方法就是在超大型積體電路的晶片中建立溫度感測器,用合適的電路提供方便的輸出以供利用。其中,嵌入式相較於分立式無論在成本或是精準度上的考量都較適合作為實現的方法。類似著名的測試性設計,熱測試性設計也被提出。 在這篇論文中,利用金氧半電晶體操作在弱反轉時漏電流的溫度特性,一個適合用於多感測器架構的溫度感測器被提出。除了具有面積小、低功率消耗的特性外,此架構包括了一個數位化的介面,使得量測結果更容易和數位系統溝通。而在多感測器的架構中,量測系統對於製程偏移所產生的不理想效應將有更大的容忍程度,進而降低了在生產後的校正程序。藉由此量測系統的使用,將有助於熱管理系統的穩健程度,並大大地增加了系統晶片的可靠度。最後,這個電路是以台積電0.18微米的製程模型作模擬。
Power densities in the high-performance VLSI chips are increasing vastly. Transistors dissipate power during operation; the faster the operation’s speed is, the more heat is generated. This leads to the rise of temperature in chips. Thus, many cooling solutions, those were dependent on the temperature sensor for detection of temperature, were presented to manage the heat dissipation. This thesis presents a temperature sensor as a basic building unit in a multi-sensor based temperature sensing system. The dependence of leakage currents and temperature is utilized to develop a temperature sensor as a basic unit of a temperature sensor scheme. Exhibiting an acceptable accuracy of 1.3oC, the proposed sensor is characterized by a pretty low area overhead of 0.001mm2 and the low power consumption of 250 uW. The proposed multi-sensors architecture can be much more robust than single sensor version and the calibration cost can be reduced. Applying this temperature sensing method in dynamic thermal management, the VLSI circuits are more reliable and operated in a safer environment.