Increasing the Sensitivity of On-Chip Digital Thermal Sensors with Pre-Filtering

Thermal monitoring has been broadly used to protect high-end integrated circuits from over-heating and to identify hot-spots in complex circuits. In this paper, we present a method to increase the sensitivity of an on-chip digital thermal sensor. In contrast to the existing mechanisms that characterize the overall temperature profile on a die, our solution is able to detect the submerged thermal variation caused by specific predefined events (SPE), under the precondition that the SPEpsilas dominant frequency does not overlap with those of other thermal events. This is made possible by pre-filtering of the temperature value. A demonstrator is implemented in an ordinary FPGA, in which the SPE is a personpsilas finger touching on the FPGA package. We successfully show that our design can do a correct and reliable detection of the finger touching event while ignoring other larger variations caused by other reasons. Because the finger touching event has no other special characteristics except for its unique frequency, we conclude that our solution is also applicable to other SPEs, especially low-frequency ones. In general, our method is sensitive, reliable and also flexible.