By Topic

Calibration of on-chip thermal sensors using process monitoring circuits

Sign In

Cookies must be enabled to login.After enabling cookies , please use refresh or reload or ctrl+f5 on the browser for the login options.

Formats Non-Member Member
$31 $13
Learn how you can qualify for the best price for this item!
Become an IEEE Member or Subscribe to
IEEE Xplore for exclusive pricing!
close button

puzzle piece

IEEE membership options for an individual and IEEE Xplore subscriptions for an organization offer the most affordable access to essential journal articles, conference papers, standards, eBooks, and eLearning courses.

Learn more about:

IEEE membership

IEEE Xplore subscriptions

2 Author(s)
Datta, B. ; Univ. of Massachusetts, Amherst, MA, USA ; Burleson, W.

Remarkable increase in peak power-density values coupled with the hotspot migration caused by workload variance motivates the need for multiple thermal monitoring circuits distributed across the die. The effect of intra-die process-variations on deep sub-micron circuits is significant enough to undermine their robustness. Accordingly, there is change in the response of thermal sensors occupying different process-corners which causes a shift in their calibration-constants. To save on tester cost, modern microprocessors employ a single, 2-point hard calibration model (slope-intercept form). In a multi-sensor environment, a single calibration equation will be rendered ineffective due to sparse sensor distribution that will be afflicted by varying degrees of process-variation. Thus, our aim is to estimate the process-induced drift in the calibration-constants of the thermal sensors. To this end, we propose a novel, supply and temperature independent, process-sensor which offers a high sensitivity of 3.35%/5 mV variation in Vth and a low power consumption of 4-25 nW. The process-estimates obtained are plugged into an analytical model used to describe the process-dependence of a ring-oscillator based thermal sensor and generate the process-shifted calibration constants. HSPICE simulations in 45 nm indicate that in the presence of process-variations having 3-¿ variability of +/-15% in all process-parameters, the average measurement error of a ring-oscillator-based thermal sensor with process-corrected calibration constants is reduced by >3X for slope and >10X for intercept as compared to one with static constants.

Published in:

Quality Electronic Design (ISQED), 2010 11th International Symposium on

Date of Conference:

22-24 March 2010