Electrothermal Filter Research Articles

Compact Thermal-Diffusivity-Based Temperature Sensors in 40-nm CMOS for SoC Thermal Monitoring

An array of temperature sensors based on the thermal diffusivity (TD) of bulk silicon has been realized in a standard 40-nm CMOS process. In each TD sensor, a highly digital voltage-controlled oscillator-based ΣΔ ADC digitizes the temperature-dependent phase shift of an electrothermal filter (ETF). A phase calibration scheme is used to cancel the ADC's phase offset. Two types of ETF were realized, one optimized for accuracy and one optimized for resolution. Sensors based on the accuracy-optimized ETF achieved a resolution of 0.36 °C (rms) at 1 kSa/s, and inaccuracies of ±1.4 °C (3σ, uncalibrated) and ±0.75 °C (3σ, room-temperature calibrated) from -40 °C to 125 °C. Sensors based on the resolution-optimized ETFs achieved an improved resolution of 0.21 °C (rms), and inaccuracies of ±2.3 °C (3σ, uncalibrated) and ±1.05 °C (3σ, room-temperature calibrated). The sensors draw 2.8 mA from supply voltages as low as 0.9 V, and occupy only 1650 μm 2 , making them some of the smallest smart temperature sensors reported to date, and well suited for thermal monitoring applications in systems-on-chip.

An SOI thermal-diffusivity-based temperature sensor with ±0.6 °C (3σ) untrimmed inaccuracy from −70 °C to 225 °C

This work presents the first thermal-diffusivity-based temperature sensor realized in SOI technology; it has an untrimmed inaccuracy of ±0.6°C (3σ) from −70°C up to 225°C and uses up to 7× less power than prior art. The sensor uses the phase shift of an Electrothermal Filter (ETF) as a proxy for the thermal diffusivity of silicon, D, which has a well-defined 1/T1.8 temperature dependence. The ETF's output is then digitized by a phase-domain sigma-delta modulator. Measured data from several process lots show that that the sensor's inaccuracy is mainly limited by lithographic spread, and not by wafer-to-wafer or batch-to-batch variations.

A Thermal-Diffusivity-Based Frequency Reference in Standard CMOS With an Absolute Inaccuracy of $\pm $0.1% From $-$55 $^{\circ}$C to 125 $^{\circ}$C

An on-chip frequency reference exploiting the well-defined thermal-diffusivity (TD) of IC-grade silicon has been realized in a standard 0.7 <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$\mu{\hbox{m}}$</tex></formula> CMOS process. A frequency-locked loop (FLL) locks the frequency of a digitally controlled oscillator (DCO) to the process-insensitive phase shift of an electrothermal filter (ETF). The ETF's phase shift is determined by its geometry and by the thermal diffusivity of bulk silicon ( <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$D$</tex> </formula> ). The temperature dependence of <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$D$</tex></formula> is compensated for with the help of die-temperature information obtained by an on-chip band-gap temperature sensor. The resulting TD frequency reference has a nominal output frequency of 1.6 MHz and dissipates 7.8 mW from a 5 V supply. Measurements on 16 devices show that it has an absolute inaccuracy of <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$\pm $</tex></formula> 0.1% ( <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$\sigma= \pm $</tex></formula> 0.05%) over the military temperature range ( <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$-$</tex></formula> 55 <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$^{\circ}{\hbox{C}}$</tex></formula> to 125 <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex Notation="TeX">$^{\circ}{\hbox{C}}$</tex></formula> ), with a worst case temperature coefficient of <formula formulatype="inline" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex Notation="TeX">$\pm \hbox{11.2~ppm}/^{\circ}{\hbox{C}}$</tex></formula> .

A Temperature-to-Digital Converter Based on an Optimized Electrothermal Filter

This paper describes the design of a CMOS temperature-to-digital converter (TDC). It operates by measuring the temperature-dependent phase shift of an electrothermal filter (ETF). Compared to previous work, this TDC employs an ETF whose layout has been optimized to minimize the thermal phase spread caused by lithographic inaccuracy. To minimize electrical phase spread, the TDC's front-end consists of a wide bandwidth gain-boosted transconductor. The transconductor's output current is then digitized by a phase-domain SigmaDelta modulator whose phase-summing node is realized by a chopper demodulator. To minimize the residual offset caused by the demodulator's switching action, the demodulator is located at the virtual ground nodes established by the transconductor's gain-boosting amplifiers. Measurements on 16 samples (within one batch) show that the TDC has an untrimmed inaccuracy of less than plusmn0.7degC (3sigma) over the military range (-55degC to 125degC).

Interface Electronics for a CMOS Electrothermal Frequency-Locked-Loop

This paper describes a new implementation of a CMOS electrothermal frequency-locked-loop (FLL), whose output frequency is determined by the temperature-dependent phase shift of an electrothermal filter (ETF). The FLL maintains a constant phase shift in the ETF, and as a result drives it with a signal whose frequency is a well-defined function of temperature. Compared to a previous implementation, the FLL described here has significantly more loop gain, less electrical phase-spread, and is more suitable for full integration. Measurements on 16 samples (from one batch) show that the temperature dependence of the FLL's output frequency agrees very well with the known thermal properties of bulk silicon. The untrimmed spread of this frequency is less than plusmn0.45% (3sigma) from -40degC to 100degC, which corresponds to a temperature-sensing inaccuracy of less than plusmn0.7degC (3sigma).

A CMOS Temperature-to-Frequency Converter With an Inaccuracy of Less Than $\pm \hbox{0.5}\,^{\circ}{\hbox{C}}$ (3$\sigma$) From $-\hbox{40}\,^{\circ}\hbox{C}$ to 105$\,^{\circ}\hbox{C}$

This paper describes a temperature-to-frequency converter (TFC) implemented in a standard CMOS process. Its output frequency is determined by the phase-shift of an electrothermal filter, which consists of a heater and a temperature sensor realized in the substrate of a standard CMOS chip. The filter's phase-shift is determined by the geometry of the thermal path between the heater and the sensor, and by the temperature-dependent rate at which heat diffuses through the substrate. The resulting temperature-dependent phase-shift is quite well-defined, since filter geometry is defined by lithography, while the thermal diffusivity of the high-purity lightly-doped silicon substrate is essentially constant. The filter was used as the frequency-determining component of a frequency-locked loop (FLL), whose output frequency is then a well-defined function of temperature. Using this approach, a TFC with an inaccuracy of plusmn0.5degC (3sigma) over the industrial temperature range (-40degC to 105 degC) has been realized in a standard 0.7-mum CMOS process

Three-layer low frequency hybrid electrothermal filter

The heat transfer function for a three-layer structure with a heat source between layers is given. Measurements on a practically realised lowpass filter built with the described structure are given.