A Temperature Estimation Method Using the Ratio of Emitter-to-Base Voltages

A temperature estimation method is proposed which utilizes the ratio of two emitter-to-base voltages of bipolar transistors. This removes the need for a precision reference voltage from the digitization step by implicitly implementing the reference in the digital backend. Using measurement data for p-n-p transistors in a 65-nm CMOS technology, we show that the proposed scheme is tolerant to tens of millivolt of offset and up to ±50% gain errors without the need for any calibration. The designed sensor core, incorporating the new estimation scheme, reports room-temperature trim accuracy of ±1°C (3σ), and an area and power of 0.009 mm
²
and 200 μW, respectively.