The integration of on-chip temperature sensors within various systems, industrial Internet of Things (IoT), and wireless sensor networks is greatly facilitated by their small size, cost-effectiveness, and capability to provide direct digital output. However, the diverse application scenarios pose challenges in designing these sensors. On one hand, real-time clock calibration demands high-precision temperature sensors, while on-chip heat management emphasizes compactness and low-voltage operation. Additionally, streamlining the calibration cost for mass production holds significant practical value. Addressing these challenges, this study systematically investigates on-chip complementary metal-oxide-semiconductor (CMOS) temperature sensors based on distinct signal domains processed by temperature readout circuits. Specifically, the research commences by analyzing the issues of several degeneracy points in the front-end circuit of a bipolar junction transistor (BJT) temperature sensor with current gain compensation technology. To address the intricate design challenges in advanced technologies and calibration complexities in industrial applications, dynamic component matching, current gain compensation, and chopper stabilization are harnessed. A novel dynamic current gain canceling technique for temperature readout is introduced, enhancing temperature measurement accuracy without incurring additional power consumption or area overhead. Ultimately, an all-digital CMOS temperature sensor is realized using the SMIC 55 nm CMOS process. Occupying a mere 0.29 mm2 of core area, the design operates efficiently across a wide supply voltage range of 1.2 V to 3.6 V. Covering a temperature spectrum from −40 °C to 125 °C, the sensor demonstrates a calibration error of just ±0.7 °C. This achievement is attributed to the incorporation of the proposed dynamic current gain compensation technique.