Industrial environmental monitoring often requires temperature and relative humidity (RH) to be measured together. Yet these parameters typically rely on different sensing elements: temperature can be encoded as a resistance change, while humidity-sensitive polymers produce a change in capacitance. A reported CMOS temperature and humidity sensor addresses this mismatch with a unified analog front end (AFE) that reads both transducers and converts their signals into a time-based output.
The design is relevant to engineers considering compact, low-power sensing for process monitoring, storage, transportation, and connected devices. Its central contribution is not simply combining two sensing elements on one chip. It is using a shared measurement architecture to handle resistive and capacitive inputs without a separate AFE for each.
Why combine resistive and capacitive sensing?
Conventional temperature and humidity interfaces are built around different circuit functions. A resistive temperature sensor needs a way to bias and measure resistance; a capacitive RH sensor usually measures changes in charge or capacitance. Implementing independent signal chains can add circuit area, power demand, and design complexity.
In the reported approach, a silicided polysilicon resistor serves as the temperature transducer, while a polyimide-coated capacitor responds to humidity. A multiplexer selects which transducer is connected to the shared front end. Measurements are time-multiplexed, allowing the circuit to obtain temperature and RH readings in succession rather than requiring two continuously active interfaces.
This arrangement also accounts for temperature’s influence on the humidity reading: the temperature measurement can be used to remove temperature dependence from the RH result. The circuit must settle quickly between modes so that ambient temperature can reasonably be treated as constant across the two measurements. The reported mode-switching resettling time was 0.6 ms.
How the unified analog front end works
The AFE converts resistance and capacitance into a timing signal using a frequency-locked loop (FLL). At the heart of the loop is a switched-capacitor Wheatstone bridge. One bridge branch contains the sensor element, while a reference branch provides a comparison. The loop adjusts its clock until the bridge is balanced; the resulting period carries information about the selected transducer.
In simplified terms, the sensor’s electrical property changes the timing required to balance the bridge. The FLL then turns that timing into a frequency-related output that can be digitized. Because the loop operates on the relationship between the sensor and a reference, it helps reduce the effects of oscillator nonlinearity and process, voltage, and temperature variation that can complicate open-loop frequency-based conversion.
A counter-based time-to-digital converter (TDC) digitizes the FLL output. Rather than relying on a conventional voltage-domain data converter, this approach samples transitions in the phase domain. Continuous digitization provides noise shaping, which supports high resolution while keeping the readout largely digital.
Incomplete settling as a power-saving design choice
The bridge uses switched capacitors but does not depend on each switching cycle reaching full electrical settling. Full settling would require actively driving the circuit, potentially adding power and noise. Instead, the design accepts incomplete settling and uses a charge-preserving capacitor, Cf, to reduce the resulting error.
Cf acts as a passive integrator. It retains charge at the bridge node while the switched capacitor transfers charge between clock phases. With an appropriately large Cf-to-sensor-capacitance ratio, the average bridge behavior approaches that of a balanced divider, even though the node does not settle completely during each cycle. The analysis indicates that a ratio above 60 can make settling error negligible for the demonstrated application.
This technique also reduces sensitivity to parasitic capacitance, clock timing imperfections, and switching effects. Those influences matter in practical sensor interfaces because connections to the readout circuit and physical layout can alter the electrical behavior of a small capacitance. The passive approach avoids using a power-hungry active driver to correct the settling problem.
Reported performance and engineering context
The sensor was implemented in a 0.18-µm CMOS process. The AFE consumed 15.6 µW and reported a temperature resolution of 2 mK and an RH resolution of 0.0073% RH. The measurements covered an industrial temperature range of −40 to +85 °C; RH resolution was reported from 10% to 95% RH.
These figures describe the demonstrated circuit and measurement setup, not a guarantee of performance for every deployment. In particular, resolution is distinct from calibrated accuracy. After first-order calibration, the reported three-sigma errors were 0.55 °C across the temperature range and 2.2% RH over 10–95% RH, with the RH error limited by the measurement equipment. Supply sensitivity was measured across 1.5–2 V.
For system designers, the key takeaway is architectural: one front end can serve two transducer types when the conversion method is designed around their shared timing relationships. The FLL and TDC provide a route to digitization, while the switched-capacitor bridge and charge-preserving capacitor address power and settling trade-offs.
Implications for industrial automation
Combining temperature and RH measurement can simplify the sensing node where both values are needed at the same location. A shared AFE may reduce duplicated circuitry, an important consideration for compact or battery-powered monitoring devices. Potential application areas include environmental monitoring associated with process control, storage, and transportation, where both measurements may inform operating decisions.
When evaluating a similar architecture, engineers should consider more than nominal resolution. Transducer characteristics, calibration requirements, environmental response, settling time, supply range, and integration constraints all affect whether a design suits a particular installation. The reported implementation demonstrates a circuit technique; it does not establish suitability for any specific industrial environment or replace application-level qualification.
Conclusion
A unified temperature and humidity sensor can reduce the need for separate resistive and capacitive measurement front ends. In this design, an FLL translates both sensor responses into timing, a counter-based TDC digitizes the result, and a charge-preserving capacitor enables low-power operation despite incomplete settling. For automation engineers and system integrators, the practical value lies in this shared-interface approach—and in weighing its measured resolution, calibration performance, power, and operating conditions against the requirements of the intended monitoring system.