On-line real-time temperature and power estimation of an IC using time-domain thermal filters

Post-Silicon Estimation of Spatiotemporal Temperature Variations Using MIMO Thermal Filters

“…The gain at sensor #2 is the largest since it is at the closest location to the heat source. The phase response also has to show low-pass like response [22]. The extracted phase response, however, is somewhat different due to the responsiveness of our delay-based digital sensor.…”

“…Essentially, the TSI is used to extract the thermal filter connecting the power source and each of the sensors; and the extracted filter was used for temporal temperature estimation. A follow-up work on the topic by Kung et al [22] has designed a time-domain thermal filter that can represent the extracted thermal filter. The timedomain filter eliminates the need for FFT or inverse FFT (IFFT) in the TSI process and enables real-time prediction of temperature.…”

“…To achieve faster responsiveness (less slowdown in applying DTM techniques), computation should be done in hardware. To remove the use of a complex FFT/IFFT hardware as in (14), a simple on-chip digital thermal filter has been designed to predict the temperature at a single location [22]. To extend the usage into a MIMO thermal system, thermal filter banks or a pipelined thermal filter can be designed, as shown in Fig.…”

“…To estimate the area, dynamic power and performance of the proposed digital hardware in [22], Nangate 45-nm technology library [31] is used and synthesized the register transfer level (RTL) design using Synopsys Design Compiler [32]. Then, P&R is executed for more accurate area estimation using Cadence Encounter [33].…”