Separating temperature effects from ring-oscillator readings to measure true IR-drop on a chip

There have been numerous attempts at adding ring-oscillator counters to measure IR-Drop on chips. The difficulty with this approach has been that any reading of the ring count always combined the effects of Temperature and Voltage versus Voltage only reading. The technique relied primarily on the fact that temperature effects take a long time to affect the ring-oscillator counter value. This paper will validate this premise by using two identical ring-oscillator counters next to each other but powered by two separate power supplies. One ring-oscillator counter will be connected to the core under test, the second ring-oscillator counter will be connected to a clean supply. By comparing the different readings, we can accurately remove temperature from the ring-oscillator counter reading and arrive at an accurate IR-Drop measurement.