▸创新点1:混合温度传感器网络架构(系统创新)。该架构结合少量高精度传感器和大量微型相对温度传感器,分别捕获低频和高频热信息,显著降低传感器总面积的同时保持热图精度。
▸创新点2:热系统空间低通滤波效应利用(方法创新)。通过理论分析证明热信息主要集中于低频区域,从而合理化稀疏高精度传感器布局,减少70%传统传感器数量。
▸创新点3:基于引导上采样的热图合成算法(算法创新)。利用高精度传感器数据作为引导,通过自适应核函数将低分辨率相对传感器数据上采样至5倍分辨率,峰值误差<0.3°C。
▸创新点4:微型相对温度传感器设计(电路创新)。采用差分环形振荡器结构实现0.004mm²超小面积,功耗仅12μW,满足高密度部署需求。
Abstract
Spatial thermal distribution of a chip is an essential
information for dynamic thermal management. To get a rich
thermal map, the sensor area is re quired to be reduced radically.
However, squeezing the sensor size is about to face its physical lim-
itation. In this background, we propose an area-ef ficient thermal
sensing technique: hybrid temperature sensor network. The
proposed sensor architecture full y exploits the spatial low-pass
filtering effect of thermal systems, which implies that most