Chip-based high-precision fluorescent fiber-optic temperature sensor

Abstract

Fluorescent fiber-optic temperature sensors have found widespread applications owing to their high sensitivity and broad temperature-sensing range. However, the noise induced by numerous discrete components remains challenging to address via algorithmic improvements. In this paper, a chip integrating both readout circuit and LED driver is designed and fabricated in 0.18-µm CMOS process. Based on this chip, an integrated fluorescent fiber-optic temperature sensor is built. Compared with its discrete counterparts, the integrated sensor exhibits a 12.18% reduction in noise voltage. With identical fluorescence lifetime extraction algorithm, the integrated sensor achieves a 56.47% improvement in fluorescence lifetime accuracy. Moreover, the integrated sensor exhibits a temperature resolution of 0.056 °C, representing a 39.13% improvement over the discrete sensor. Meanwhile, the temperature measurement range spans from − 20 °C to 120 °C, with an average temperature sensitivity of 0.0064 ms°C^− 1. This performance enhancement renders it well-suited for high-precision temperature measurement.

Introduction

Fiber-optic sensors have remarkable adaptability to various complex environments, leading to their extensive utilization^1–4. Improving the accuracy while expanding the temperature measurement range has emerged as a primary focus in research^5,6. Wang et al. proposed a fiber-optic temperature sensor based on quantum dots (QDs)⁷, which exhibits excellent temperature measurement linearity over the range of −10 °C to 120 °C. By improving the encapsulation method of QDs, the fiber-optic temperature sensor designed by Liu et al.⁸ achieved a measurement deviation of less than 0.5 °C. Furthermore, Zhao et al. proposed a fiber-optic temperature sensor encapsulated within a hollow core fiber (HCF) using graphene quantum dots (GQDs)⁹ This design enhanced the sensor’s sensitivity, but the temperature measurement range was further restricted to 10 °C to 80 °C. Fiber-optic temperature sensors based on fluorescence intensity ratio (FIR) or fluorescence lifetime can achieve higher accuracy, yet broadening the temperature measurement range remains a challenge. Zhao et al. proposed a compact fluorescent fiber temperature sensor doped with oxyfluoride germanate phosphor¹⁰, achieving a measurement error of 0.45 °C. By improving the design of the fluorescent material, Li et al. proposed a ZnS: Cu-based FIR temperature sensor¹¹ that achieved an average temperature sensitivity of −0.00515 °C⁻¹ over the range of 20.1 °C to 101 °C. The ratio-metric fluorescent temperature sensor based on silica-tellurite composite proposed by Liu et al.¹² achieves an average temperature deviation of 0.276 °C, but its range of measurement is unsatisfactory. By designing tapered optical fibers, Li et al. proposed a self-powered fluorescent temperature sensor¹³ that achieves a temperature resolution of 0.11 °C, though its measurement range is only 30 °C to 50 °C. Therefore, improving the performance of fluorescent fiber-optic temperature sensors requires not only hardware enhancements but also algorithm optimization. Many studies have been devoted to improving the accuracy of fluorescence lifetime extraction algorithms and excellent progress has been made. Sun et al. employed the fast Fourier transform (FFT) method to fit the decay curves of fluorescence signals¹⁴, enabling them to obtain more accurate fluorescence lifetimes while accelerating fluorescence data processing. Similarly, He et al. utilized the weighted least squares (WLS) fitting method to calculate the fluorescence lifetime¹⁵. By selecting more appropriate data points, He et al. further enhanced the accuracy of fluorescence lifetime measurements.

For conventional fluorescent fiber-optic temperature sensors, fluorescent signals are susceptible to interference from noise generated by numerous discrete components. This interference ultimately leads to degraded sensor accuracy, which remains a critical challenge demanding attention. However, research aimed at improving accuracy by reducing noise in fluorescent fiber-optic temperature sensors remains scarce. In fact, numerous studies have achieved breakthroughs through on-chip integration technology^16–19. He et al. monolithically integrated photodetectors (PDs), light-emitting diodes (LEDs), and optical waveguides²⁰, resulting in an integrated photonic chip that achieved high-speed communication of 150 Mbps with excellent transient response. In the context of fiber grating sensing systems, Li et al. perfected the monolithic integration architecture of arrayed waveguide gratings and photodetectors²¹, which substantially improved the detection accuracy. Many studies indicate that integrated devices play a significant role in noise reduction^22,23. Yan et al. integrated chopper stabilization and analog bias calibration on-chip to realize a differential temperature sensor with ultra-high sensitivity²⁴, effectively suppressing low-frequency flicker noise and direct current (DC) offsets. To miniaturize high-g sensors, Shi et al. monolithically integrated application-specific integrated circuits (ASICs) and microelectromechanical systems (MEMS)²⁵. This integration resulted in improved resolution and signal-to-noise ratio (SNR) compared to discrete sensor configurations. By integrating analog buffers, multiplexers and other peripheral devices, Ahn et al. developed a low-noise single-chip sub-terahertz imager array that delivers excellent performance²⁶.

To achieve high-precision and wide-range temperature measurement, we achieved single-chip integration of the readout circuit and LED driver while maintaining the PD’s independent external configuration. Leveraging this designed chip, we developed a prototype of the integrated fluorescent fiber-optic temperature sensor. Compared to its discrete counterpart, the integrated sensor suppresses noise voltage by 12.18%. Temperature calibration results over the range of −20 °C to 120 °C demonstrate that the integrated sensor improves fluorescence lifetime measurement accuracy by 56.47%. Additionally, the sensor’s temperature resolution is improved by 39.13%. The constructed sensor exhibits an average temperature sensitivity of 0.0064 ms°C^− 1 and a measurement range of −20 °C to 120 °C. Experimental results confirm that the integrated sensor constructed in this work exhibits great potential for application in high-precision temperature detection scenarios.

Conventional fluorescent fiber-optic temperature sensor

The composition of the fluorescent fiber-optic temperature sensor is shown in Fig. 1(a), while the accuracy of measurement depends on the precision of the fluorescence lifetime. Light absorption from a light source initiates the excitation of fluorescent materials, leading to the production of fluorescence. The fluorescent signal is received by the PD into the readout circuit. Through the transimpedance photoelectric conversion circuit, amplification circuit, and low-pass filtering circuit, the fluorescent signal is converted to an electrical signal and amplified. Then, it is transmitted through the microcontroller into the personal computer, so that the fluorescent signal attenuation curve is obtained. Subsequently, the decay curve of the fluorescence signal undergoes processing by data processing algorithms, and getting the fluorescence lifetime. The fluorescence lifetime τ is defined as the time for the fluorescence to decay from the initial intensity to 1/e, which satisfies the following equation:

1

Fig. 1.

(a) The composition of fluorescent fiber-optic temperature sensor. (b) The PCB of hardware system. (c) The optical path structure.

where is the fluorescence initial intensity, is the fluorescence intensity at time , and is the fluorescence lifetime.

To facilitate subsequent comparisons, we first fabricated a prototype of a conventional fluorescent fiber-optic temperature sensor. Its printed circuit board and optical path structural components are shown in Fig. 1(b) and (c), respectively. Notably, the optical path structure is universal and independent of the specific sensor design.

Design proposal based on single-chip integration

Enhancing circuit integration is an effective way to reduce noise, but it has been a challenge to appropriately integrate the circuits in fluorescent fiber-optic temperature sensors. In this paper, an integrated chip applied to fluorescent fiber-optic temperature sensor is designed, and monolithic integration is achieved through the rational design of the fluorescent signal readout circuit and LED driving circuit, and a corresponding sensor prototype have developed. It should be noted that due to certain technical limitations, the PD is not integrated within the chip but is instead an off-chip component.

The integrated chip with readout circuit and LED driver

The principle of the integrated chip designed in this paper is shown in Fig. 2(a). Fig. 2(b) and Fig. 2(c) show the layout and overview of the designed integrated chip, respectively.

Fig. 2.

(a) Principle of the chip-based fluorescent fiber-optic temperature sensor. (b) Layout of the integrated chip. (c) Overview of the integrated chip.

It should be noted that the internal optical circuit structure and the external quartz fiber of the fluorescent type fiber-optic temperature sensor are omitted. It is imperative to emphasize the integration of an LED driver circuit module with an adjustable function into the chip, which is a pivotal circuit module. On the one hand, an excessively high luminous power of the LED can result in an excessively large fluorescent signal amplitude, causing saturation distortion or cutoff distortion. On the other hand, if the LED has insufficient luminous power, the fluorescent signal may be too weak to ensure a satisfactory signal-to-noise ratio. Therefore, the optimal luminous power of the LED is determined by the maximum and undistorted amplitude of the fluorescent signal. Additionally, the stability of the initial intensity of the fluorescent signal is also a prerequisite for the normal operation of the sensor. To this end, this paper proposes an adjustable LED driving circuit that utilizes a negative feedback control technique to modulate the luminous power of the LED. This circuit enables indirect modulation of the intensity of the fluorescent signal, thereby enhancing the functionality of the sensor. In this paper, the integrated chip has been designed and fabricated in 0.18-µm complementary metal oxide semiconductor (CMOS) technology, and the specification of the package is small out-line package (SOP).

The chip-based integrated fluorescent fiber-optic temperature sensor

Based on the proposed integrated chip, the PCB of corresponding fluorescent fiber-optic temperature sensor is structed as shown in Fig. 3(a) below. Additionally, this paper also accomplishes the design of the accompanying software based on the STM32 microcontroller, and the software flowchart of the sensor is shown in Fig. 3(b). The workflow of the software mainly uses the important ports of the microcontroller such as analog-to-digital converter (ADC), timer and direct memory access (DMA). After initialization, the LED is turned on and the fluorescence decay curve is collected. The fluorescence lifetime is calculated according to the fluorescence decay curve, then the fluorescence life data is transmitted to the computer through the serial port. Through the calibration of temperature, the relationship between fluorescence life and temperature can be obtained. As a comparison, both the sensor integrated with the designed chip and the discrete sensor use the least squares fitting method to calculate the fluorescence lifetime.

Fig. 3.

(a) PCB of fluorescent fiber-optic temperature sensor using the integrated chip. (b) Flowchart of the software.

Experiment and discussion

Noise voltage experiment

For the hardware circuit of the fluorescent fiber-optic temperature sensor, both passive and active components introduce noise interference. However, the weak fluorescent signal requires amplification by the TIA, making the TIA the primary source of internal noise within the sensor’s hardware circuit. In this work, the TIA is present within the photoconversion circuit. Here, we conducted noise voltage experiments on two sensors within a thermostatic chamber. It should be noted that both two sensors have the same model for their core components. The models of MCU, PD, LED, LDO and crystal oscillator are STM32F103RCT6, SFH203, STC395-5M05, TPS79333DBVR and SMD5032 respectively. The readout circuit and driver of the discrete sensor are respectively constructed by LMV796 and TLV2620. The measurement bandwidth was fixed at 1 MHz. The oscilloscope model was RIGOL MSO5104 and the sampling frequency was set to 5 MSa/s. The input coupling mode was DC coupling, and the vertical sensitivity was set to 5 ms/div. For comparison purposes, the noise voltage in this work refers to the root mean square (RMS) value of the output noise voltage from the transimpedance amplifier within the hardware circuit. The experimental temperature range was set from − 20 °C to 120 °C, with temperature points spaced at 10 °C intervals. At each temperature point, 10 sets of output voltage data without fluorescence signals were collected, and the RMS output noise voltage was calculated. Finally, calculate the RMS output noise voltage values for the average, minimum and maximum values within the entire temperature range. Throughout the experiment, the ambient temperature was maintained at approximately 25 °C. Both sensor types were tested under identical electromagnetic, temperature, and illumination conditions to ensure consistency in external noise input intensity. The results of the noise voltage experiment are presented in Table 1 below.

Table 1.

Comparison of RMS output noise voltages.

Noise voltage	Discrete sensor	This work	Reduction
Average	22.99 mV	20.19 mV	12.18%
Maximum	38.92 mV	31.54 mV	18.96%
Minimum	7.99 mV	6.39 mV	20.03%

Table 1 shows that the maximum RMS output noise voltage of the discrete sensor ranges from − 20 °C to 120 °C, with a maximum value of 38.92 mV and an average value of 22.99 mV. In comparison, the integrated sensor designed in this work exhibits a maximum RMS output noise voltage of 31.54 mV and an average RMS output noise voltage of 20.19 mV. The minimum RMS output noise voltages for the discrete and integrated sensors are 7.99 mV and 6.39 mV, respectively. In summary, within the temperature range of −20 °C to 120 °C, the integrated sensor achieved reductions of 12.18%, 18.96%, and 20.03% in the mean, maximum, and minimum RMS output noise voltage, respectively. Fig. 4 more clearly illustrates the significant noise reduction achieved through on-chip integration technology.

Fig. 4.

The comparison of RMS output noise voltage.

Calibration of temperature

Experiments for fluorescence lifetime acquisition were conducted with the integrated and discrete sensor. The required experimental equipment included a thermostatic chamber, two sensors, a serial port receiver, and a computer. The experimental temperature range was set from − 20 °C to 120 °C with 10 °C increments. To evaluate the accuracy of lifetime measurements, the Newtonian iteration method was employed to calculate reference lifetime values. The Newtonian iteration method is commonly used to solve the roots of mathematical equations, with the advantage of achieving extremely high-precision solutions through sufficient iterations^27–29. However, excessive iterations significantly prolong the algorithm’s processing time, which is detrimental to real-time temperature measurement. Therefore, we will use the fluorescence lifetime obtained via the Newtonian iteration method as a reference value for comparing the accuracy of integrated and discrete sensors. It should be noted that the reference value is derived from the fluorescence decay curve of the discrete sensor to ensure it accurately represents the precision level achievable in current work. The deviations of both sensors from the reference values at different temperatures are shown in Table 2 below.

Table 2.

Comparison of fluorescence lifetime accuracy of two sensors.

T	Discrete sensor	Integrated sensor	Reference value	Deviation of discrete sensor	Deviation of integrated sensor	Reduction
−20℃	2.9148 ms	2.8891 ms	2.8944 ms	0.0204 ms	0.0053 ms	74.02%
−10℃	2.8133 ms	2.8292 ms	2.8361 ms	0.0228 ms	0.0069 ms	69.74%
0℃	2.7639 ms	2.7754 ms	2.7846 ms	0.0207 ms	0.0092 ms	55.56%
10℃	2.6805 ms	2.7017 ms	2.7142 ms	0.0337 ms	0.0125 ms	62.91%
20℃	2.6399 ms	2.6317 ms	2.6282 ms	0.0117 ms	0.0035 ms	70.09%
30℃	2.5611 ms	2.5629 ms	2.5674 ms	0.0063 ms	0.0018 ms	71.43%
40℃	2.5056 ms	2.4914 ms	2.4946 ms	0.0110 ms	0.0032 ms	70.91%
50℃	2.4347 ms	2.4339 ms	2.4342 ms	0.0005 ms	0.0003 ms	40.00%
60℃	2.3706 ms	2.3650 ms	2.3567 ms	0.0139 ms	0.0083 ms	40.29%
70℃	2.2937 ms	2.2993 ms	2.3053 ms	0.0116 ms	0.0060 ms	48.28%
80℃	2.2383 ms	2.2355 ms	2.2361 ms	0.0022 ms	0.0006 ms	72.73%
90℃	2.1723 ms	2.1755 ms	2.1813 ms	0.0090 ms	0.0058 ms	35.56%
100℃	2.1156 ms	2.1153 ms	2.1148 ms	0.0008 ms	0.0005 ms	37.50%
110℃	2.0586 ms	2.0553 ms	2.0523 ms	0.0063 ms	0.0030 ms	52.38%
120℃	1.9906 ms	1.9948 ms	1.9998 ms	0.0092 ms	0.0050 ms	45.65%

As illustrated in Fig. 5(a), we characterized the temperature-dependent characteristics of fluorescence lifetime for both integrated and discrete sensors. While both sensors exhibit a decreasing fluorescence lifetime as temperature increases, the integrated sensor maintains better consistency with the standard value over the temperature range of −20 °C to 120 °C. In contrast, the discrete sensor shows a larger deviation at specific temperatures (e.g., −20 °C and 80 °C), thus indicating lower precision. The correlation between fluorescence lifetime and temperature is clearly evident, and the calculated correlation coefficient of integrated sensor and discrete sensor are 0.9985 and 0.9978, respectively. Furthermore, Fig. 5(b) compares the deviation of the fluorescence lifetime obtained by the integrated sensor and the discrete sensor. Compared with the discrete sensor, the fluorescence lifetime accuracy obtained by integrated sensor is improved by an average of 56.47% and up to 74.02%.

Fig. 5.

(a) Temperature calibration results. (b) Deviations of fluorescence lifetime.

Temperature measurement

Next, we evaluated the measurement accuracy, stability, and resolution of both sensors across the temperature range of −20 °C to 120 °C. Both sensors were placed at the same position within the thermostatic chamber, and their respective fluorescence lifetime values and converted temperatures were obtained using the same fluorescence lifetime extraction algorithm. Measurements at each temperature were conducted for 90 min, with 10-minute intervals. By comparing against the incubator’s standard temperature,

we determined the sensors’ temperature measurement accuracy and assessed their stability. Sensor resolution was defined as the ratio of fluorescence lifetime precision to sensor sensitivity. Fluorescence lifetime precision was obtained by calculating the standard deviation of 10 lifetime measurements at each temperature. Sensor sensitivity was determined by calculating the ratio of the change in fluorescence lifetime to the change in temperature across the measurement range. The experimental results are shown in Fig. 6. In the temperature measurement accuracy and stability experiments, the average temperature measurement deviation of the discrete sensor was 0.115 °C. In comparison, the average temperature measurement deviation of the integrated sensor designed in this work was 0.075 °C, representing a reduction of 34.78%. In the temperature sensitivity test, the integrated sensor achieved 0.0064 ms·℃⁻¹, representing a 3.13% improvement over the discrete sensor. Furthermore, the integrated sensor reached an optimal temperature resolution of 0.056 °C at −20 °C, surpassing the discrete sensor’s resolution of 0.092 °C at the same temperature by 39.13%.

Fig. 6.

Comparison of temperature measurement stability from − 20 °C to 120℃.

As shown in Fig. 7, to comprehensively evaluate the performance of the structured fluorescent fiber-optic temperature sensor, it was compared with other state-of-the-art fiber-optic temperature sensors^8,10,12. Previously reported sensors exhibit narrower temperature measurement ranges than the proposed device, alongside higher temperature deviations. The specific performance parameters are shown in Table 3. Since the operating principles of these sensors are not entirely identical, this comparison is intended solely to demonstrate the competitiveness of the constructed sensors. Despite this limitation, the sensor constructed in this work demonstrates unique advantages for practical temperature monitoring. The fluorescence lifetime-based mechanism ensures robustness against environmental interference, while the achieved temperature range (−20 °C –120 °C), resolution (0.056 °C) and linear correlation (R² = 0.9985) meet the requirements of high-precision temperature measurement. Compared to similar works^10,13, our work outperforms Ref¹⁰ in temperature measurement range and sensitivity, and surpasses Ref¹³ in resolution, sensitivity, and linearity. When compared to fiber-optic temperature sensors based on different principles, our work also leads in temperature measurement range and linearity. It should be noted that the PD in this work remains discrete. This occupies additional space to some extent, limiting the sensor’s application in space-constrained scenarios. Integrating the PD onto the chip would further optimize the system’s power consumption and accuracy, potentially expanding its application scope.

Fig. 7.

Comparison with other outstanding work in terms of sensor measurement range and deviation.

Table 3.

Performance comparison of different fiber-optic temperature sensors ^*.

Refs.	Temperature range	Temperature resolution	Measurement sensitivity	Measurement deviation	Linear correlation	Sensing mechanism	Year
7	−10℃−120℃	N/A	~ 130.9 pm℃−1	N/A	N/A	WS	2015
8	20℃−105℃	N/A	116.1 pm℃−1	< 0.5℃	R2 = 0.998	WS	2025
9	10℃−80℃	N/A	123.7 pm℃−1	N/A	R2 = 0.99	WS	2017
10	0℃−90℃	0.01℃	0.0081 ms℃−1	0.45℃	N/A	FL	2016
11	20.1℃−101℃	N/A	−0.00515 ℃−1	N/A	R2 = 0.991	FIR	2022
12	35.3℃−45.3℃	0.19℃	12.74 10− 3K− 1	0.276℃	R2 = 0.9989	FIR	2024
13	30℃−50℃	0.11℃	−0.553 ms°C− 1	N/A	R2 = 0.9967	FL	2025
This work	−20℃−120℃	0.056℃	0.0064 ms℃−1	0.075℃	R2 = 0.9985	FL	2025

^* The performance metrics of compared sensors correspond to distinct sensing mechanisms:.

(i) Sensors with unit pm°C^− 1 adopt wavelength-shift (WS)-based mechanisms;.

(ii) Sensors with unit ms°C^− 1 rely on fluorescence lifetime (FL)-based mechanisms;.

(iii) Sensors with unit K^− 1 or °C^− 1 are based on fluorescence intensity ratio (FIR) mechanisms.

Conclusions

In this paper, a single chip integrating both readout circuit and LED driver is designed and fabricated in 0.18-µm CMOS process. Based on this chip, a prototype of a high-precision fluorescent fiber-optic temperature sensor is developed. Discrete component count reduction effectively minimizes sensor noise, thereby enhancing the accuracy of fluorescence lifetime. Experimental results demonstrate that, compared with the discrete sensor, the noise voltage of the integrated sensor is reduced by 12.18%. Under the identical fluorescence lifetime extraction algorithm, the precision of the fluorescence lifetime is improved by 56.47%. Last but not least, the integrated sensor achieves a temperature resolution of 0.056 °C and an average sensitivity of 0.0064 ms℃⁻¹ within the range of −20 °C to 120 °C. This work provides valuable reference for the design of high-performance integrated fluorescent fiber-optic temperature sensors.

Author contributions

Y. X. participated in the design of the integrated chip and sensor, completed all experiments, and drafted the manuscript. J. Y. led the design of the integrated chip. Z. X. participated in the conceptual design and experiments for this work. X. L. conceived this work, participated in its design, and helped to draft the manuscript. Y. X., J. Y., Z. X. and X. L. reviewed and approved the manuscript.

Funding

This work was supported in part by the National Natural Science Foundation of China under Grant 62174052, and in part by the Postgraduate Scientific Research Innovation Project of Hunan Province under Grant CX20250752.

Data availability

All data in support of the findings of this paper are available within the article.

Declarations

Competing interests

The authors declare no competing interests.