Ultra-weak light detection system and method based on SPAD array and time-correlated photon counting

Abstract

This invention discloses an ultra-weak light detection system and method based on SPAD arrays and time-correlated photon counting, belonging to the field of ultra-weak light detection technology. It includes a trigger light source; a multi-channel SPAD array detection unit; a time-to-digital conversion module; a photon arrival time resolution unit, including a TCSPC model, a noise modeling and suppression module, and an adaptive convolutional filter; a data compression and fitting module; a spatial denoising filter; an output unit; and a main control processing unit. Based on the multi-channel SPAD array detection unit, time-to-digital conversion module, TCSPC model, noise modeling and suppression module, and adaptive convolutional filter, this invention achieves high-sensitivity imaging and quantitative detection of extremely low reflectivity materials (such as black silicon). The system features good real-time performance, high temporal resolution, high signal-to-noise ratio, strong on-chip processing capability, and high integration, while also solving the bottleneck of traditional TCSPC methods in processing large-scale photon data.

Description

Technical Field

[0001] This invention relates to the field of ultra-weak light detection technology, specifically to an ultra-weak light detection system and method based on SPAD array and time-correlated photon counting. Background Technology

[0002] With the development of high-sensitivity optical detection technologies such as quantum communication, biofluorescence imaging, deep space exploration, single-molecule imaging, and night vision imaging, the requirements for the detection capability of "ultra-weak light signals" are increasing. Ultra-weak light typically refers to situations where the photon flux density incident on the detector is extremely low (even less than one photon per nanosecond). In these application scenarios, traditional CCD (Charge-Coupled Device) or CMOS (Complementary Metal Oxide Semiconductor) image sensors are insufficient to meet the performance requirements.

[0003] Single-photon detectors have attracted attention due to their high sensitivity to extremely weak light signals, among which SPADs (Single-Photon Avalanche Diodes) have become a hot topic in research and application. SPADs operate in Geiger mode, enabling them to produce an avalanche response to every arriving photon. They not only have high single-photon response efficiency η but also nanosecond-level time resolution. However, they are also sensitive to interference sources such as dark current, thermal noise, and crosstalk, necessitating the introduction of noise suppression strategies to obtain usable signals.

[0004] Furthermore, due to the extremely low intensity of ultra-weak light signals, they exhibit significant randomness, making simple integral readings or intensity averaging methods highly susceptible to noise contamination, leading to misjudgments or data loss. To overcome these shortcomings, existing technologies employ Time-Correlated Single Photon Counting (TCSPC) technology. By comparing the time reference (trigger) of the excitation source with the arrival time of the probe photon using timestamps, photon signals can be effectively distinguished from background noise, thereby significantly improving the signal-to-noise ratio (SNR).

[0005] However, existing technologies still have the following shortcomings:

[0006] 1) SPAD devices mostly exist in the form of single-point detection or linear arrays, lacking scalable large-scale two-dimensional array design and having insufficient integration.

[0007] 2) TCSPC technology needs to process a large amount of data, and existing algorithms have problems such as poor real-time performance;

[0008] 3) For low reflectivity targets with strong light absorption characteristics (such as black silicon), the sensitivity of existing systems is insufficient and cannot meet the requirements for high precision.

[0009] Therefore, there is an urgent need to develop a SPAD array detection system that is suitable for low reflectivity materials, highly integrated, noise-resistant, and appropriate for TCSPC data processing.

[0010] Based on this, the present invention designs an ultra-weak light detection system and method based on SPAD array and time-correlated photon counting to solve the above problems. Summary of the Invention

[0011] To address the aforementioned shortcomings of existing technologies, this invention provides an ultra-weak light detection system and method based on SPAD arrays and time-correlated photon counting. This invention utilizes a multi-channel SPAD array detection unit, a time-to-digital conversion module, a TCSPC model, a noise modeling and suppression module, and an adaptive convolutional filter to achieve high-sensitivity imaging and quantitative detection of extremely low reflectivity materials (such as black silicon). The system features good real-time performance, high temporal resolution, high signal-to-noise ratio, strong on-chip processing capabilities, and high integration, while simultaneously overcoming the bottleneck of traditional TCSPC methods in processing large-scale photon data.

[0012] To achieve the above objectives, the present invention provides the following technical solution:

[0013] An ultra-weak light detection system based on SPAD arrays and time-correlated photon counting includes:

[0014] Trigger light source;

[0015] A multi-channel SPAD array detection unit is used to capture ultra-weak light signals and trigger a time-to-digital conversion module to record the photon arrival time;

[0016] The time-to-digital conversion module is used to convert the arrival time of each photon into a precise timestamp;

[0017] The photon arrival time resolution unit includes: a TCSPC model for calculating delay time and constructing a delay time histogram; a noise modeling and suppression module for background noise modeling and effective signal modeling, while introducing a weighted filter to suppress the background; and an adaptive convolution filter for outputting effective photon events.

[0018] The data compression and fitting module is used to construct two-dimensional images;

[0019] For a low-reflectivity target with strong light absorption characteristics, the spatial denoising filter performs Gaussian-Bayes filtering in the two-dimensional image space to smooth spatial noise in the image.

[0020] The output unit is used to output two-dimensional image results;

[0021] And the main control processing unit.

[0022] Furthermore, the multi-channel SPAD array detection unit adopts a two-dimensional SPAD array detector, and each SPAD pixel structure includes an avalanche diode body and an active hardening circuit; the active hardening circuit structure is as follows: bias voltage V BIAS One end is connected to the anode of the avalanche diode body, with a bias voltage V. BIAS The other end is connected to the enable (EN) terminal. The avalanche diode's cathode is grounded, and the avalanche diode's cathode is also connected to the drain of the field-effect transistor (FET). The FET's source is grounded. One end of resistor R is connected to the FET's drain, and the other end of resistor R is connected to the inverting input of the operational amplifier. The non-inverting input of the operational amplifier is connected to the reference voltage V. REF Connect the other end of EN to the output terminal.

[0023] Furthermore, the time-to-digital conversion module employs a time-to-digital converter, which includes a master counter, a delay chain, a sampling latch, and a Gray code converter.

[0024] Furthermore, the timestamp calculation formula is as follows:

[0025]

[0026] Where, N clk Master counter value, T is the system clock cycle. fine The result of the delay chain calculation;

[0027] The time resolution δt is:

[0028]

[0029] Where n is the TDC bit width.

[0030] Furthermore, the following algorithm is used in the TCSPC model:

[0031] Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is:

[0032]

[0033] Count all detected photon events and construct a delay time histogram:

[0034]

[0035] in, Delay time histogram Number of samples : No. In the second sampling, the detected first The arrival time of each photon Time reference of this excitation The delay time between; Approximation of the Dirac delta function; The time variable in the histogram;

[0036] The following algorithm is used in the noise modeling and suppression module:

[0037] Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is:

[0038]

[0039] in, Background noise over time The probability density at time , : Average count rate of background noise The time it takes for a photon to arrive;

[0040] Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope:

[0041]

[0042] Among them, P s (t): at time Signal probability density at that location Peak time of the effective signal 1: Standard deviation of the signal's temporal distribution : Actual arrival time of photons;

[0043] Introducing a weighted filter W(t) to suppress background:

[0044]

[0045] in, For regularization terms;

[0046] Using adaptive convolution filtering algorithms in adaptive convolution filters:

[0047] Define a time window function G(t) to enhance repetitive signals:

[0048]

[0049] The convolution result is:

[0050]

[0051] Where t-τ reflects the time interval from the past time τ to the current time t;

[0052] When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold.

[0053] Furthermore, the photon arrival time resolution unit adopts a hardware co-acceleration structure, namely an FPGA+ASIC hybrid architecture.

[0054] Furthermore, Gaussian-Bayes filtering is performed in the two-dimensional image input space denoising filter using the following formula:

[0055]

[0056] in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the Gaussian weight decay rate, and Ω represents the standard deviation parameter controlling the Gaussian weight decay rate. Z is the neighborhood window centered on the center, and Z is the normalization factor.

[0057] To better achieve the objectives of this invention, this invention also provides an ultra-weak light detection method based on SPAD array and time-correlated photon counting, comprising the following steps:

[0058] Step 1: Laser pulse triggering;

[0059] Step 2: The two-dimensional SPAD array detector receives the ultra-weak light signal and triggers an avalanche, and triggers the time-to-digital converter to record the photon arrival time;

[0060] Step 3: The time-to-digital converter converts the arrival time of each photon into a precise timestamp;

[0061] Step 4: The photon arrival time analysis unit calculates the delay time and constructs a delay time histogram through the TCSPC model. It performs background noise modeling and effective signal modeling through the noise modeling and suppression module. At the same time, a weighted filter is introduced to suppress the background, and an effective photon event is output through an adaptive convolution filter.

[0062] Step 5: Construct a two-dimensional image. If the target is a low-reflectivity object with strong light absorption characteristics (such as black silicon), then input the two-dimensional image into a spatial noise reduction filter to perform Gaussian-Bayes filtering and smooth the spatial noise of the image.

[0063] Step 6: Output the two-dimensional image result to the main control processing unit.

[0064] Furthermore, in step four, the following algorithm is used in the TCSPC model:

[0065] Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is:

[0066]

[0067] Count all detected photon events and construct a delay time histogram:

[0068]

[0069] in, Delay time histogram Number of samples : No. In the second sampling, the detected first The arrival time of each photon Time reference of this excitation The delay time between; Approximation of the Dirac delta function; The time variable in the histogram;

[0070] The following algorithm is used in the noise modeling and suppression module:

[0071] Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is:

[0072]

[0073] in, Background noise over time The probability density at time , : Average count rate of background noise The time it takes for a photon to arrive;

[0074] Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope:

[0075]

[0076] Among them, P s (t): at time Signal probability density at that location Peak time of the effective signal 1: Standard deviation of the signal's temporal distribution : Actual arrival time of photons;

[0077] Introducing a weighted filter W(t) to suppress background:

[0078]

[0079] in, For regularization terms;

[0080] Using adaptive convolution filtering algorithms in adaptive convolution filters:

[0081] Define a time window function G(t) to enhance repetitive signals:

[0082]

[0083] The convolution result is:

[0084]

[0085] Where t-τ reflects the time interval from the past time τ to the current time t;

[0086] When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold.

[0087] Furthermore, in step five, the two-dimensional image is further input into a spatial denoising filter to perform Gaussian-Bayes filtering, using the following formula:

[0088]

[0089] in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the Gaussian weight decay rate, and Ω represents the standard deviation parameter controlling the Gaussian weight decay rate. Z is the neighborhood window centered on the center, and Z is the normalization factor.

[0090] Compared to existing technologies, the advantages of this invention are as follows: This invention employs a two-dimensional SPAD array detector (SPAD) and a time-to-digital converter (TDC) to form the detection unit. Combined with a TCSPC model, noise modeling and suppression module, and adaptive convolutional filter, it effectively suppresses background noise, improves the ability to identify extremely weak signals, and exhibits excellent signal restoration performance under conditions of strong ambient light interference and high noise levels. It can adapt to the high absorption characteristics of low-reflectivity regions (such as black silicon surfaces), significantly enhancing the detection capability of weakly reflected light and achieving high-sensitivity imaging and quantitative detection of extremely low-reflectivity materials (such as black silicon). The system of this invention supports black silicon materials with reflectivity below 10... -3 Effective imaging under these conditions, with a minimum measurable reflectance of 7×10⁻⁶. -5 The system is highly integrated, supports on-chip parallel TCSPC processing, and features good real-time performance, high temporal resolution, high signal-to-noise ratio, and strong on-chip processing capabilities. It can employ a hardware-co-acceleration architecture, making it suitable for the rapid construction of large-scale 2D images. Attached Figure Description

[0091] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0092] Figure 1 This is a structural diagram of the ultra-weak light detection system based on SPAD array and time-correlated photon counting according to the present invention.

[0093] Figure 2 This is a structural diagram of an active quenching circuit.

[0094] Figure 3 This is a flowchart of a photon arrival time analysis unit.

[0095] Figure 4 This is a schematic diagram of the output results of black silicon sample detection using the present invention.

[0096] Figure 5 This is a schematic diagram of the chip physical layout of a two-dimensional SPAD array detector (SPAD) and a time-to-digital converter (TDC). Detailed Implementation

[0097] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0098] Example 1: Please refer to the accompanying drawings in the instruction manual. Figures 1-5 An ultra-weak light detection system based on SPAD array and time-correlated photon counting includes:

[0099] 1. The triggering light source is a laser, and the laser pulse emission is triggered by a TTL signal;

[0100] 2. Multi-channel SPAD array detection unit (SPAD for short) adopts a two-dimensional SPAD array detector to capture ultra-weak light signals and trigger the time-to-digital conversion module to record the photon arrival time;

[0101] To achieve high integration and high spatial resolution, a 256×256 two-dimensional SPAD array detector is fabricated using a CMOS (Complementary Metal-Oxide-Semiconductor) compatible process. Each SPAD pixel structure of the two-dimensional SPAD array detector includes an avalanche diode (AD) body and an active quenching circuit (AQC).

[0102] See active quenching circuit Figure 2 Bias voltage V BIAS One end is connected to the anode of the avalanche diode body, with a bias voltage V. BIAS The other end is connected to the enable (EN) terminal. The avalanche diode's cathode is grounded, and the avalanche diode's cathode is also connected to the drain of the field-effect transistor (FET). The FET's source is grounded. One end of resistor R is connected to the FET's drain, and the other end of resistor R is connected to the inverting input of the operational amplifier. The non-inverting input of the operational amplifier is connected to the reference voltage V. REF The other end of the EN terminal is connected to the output terminal, and the output terminal outputs the signal processed by the operational amplifier.

[0103] Each SPAD pixel in the two-dimensional SPAD array detector uses a silicon-based avalanche structure and operates at a voltage of V. op :

[0104] V op =V br +ΔV

[0105] Among them, V brΔV is the avalanche breakdown voltage, and ΔV is the overvoltage, with a value ranging from 2 to 5V.

[0106] When the SPAD reaches its breakdown condition, an avalanche response is generated, triggering a time-to-digital converter to record the time. When an incident photon triggers an avalanche in the SPAD, the voltage at the SPAD cathode drops rapidly. Upon detecting this voltage change, the operational amplifier immediately outputs a high-level signal to turn on the MOSFET, grounding the SPAD cathode and rapidly discharging the avalanche current, thus terminating the avalanche process. After a preset quenching time t... q Afterwards, the MOSFET is disconnected, the SPAD is reconnected to the bias voltage VBIAS, and the system returns to the probe state, completing a full quench-reset cycle.

[0107] The quenching cycle is usually:

[0108]

[0109] in, For junction capacitance, To restore voltage.

[0110] The photon detection probability (PDP) of each SPAD pixel structure can be expressed as:

[0111]

[0112] Where λ is the photon wavelength, P in P is the incident light power. det This refers to the actual measured optical power.

[0113] 3. Time-to-digital converter module, which uses a time-to-digital converter (TDC) to convert the arrival time of each photon into a precise timestamp;

[0114] The time-to-digital converter (TDC) includes a master counter, a delay chain, a sample latch, and a Gray code converter;

[0115] The main counter uses the system clock as the timing reference to achieve a coarse quantization and recording of time.

[0116] The delay chain uses a pseudo-random time delay line (TDL) or digital phase-locked loop (DLL) structure to subdivide the system clock cycle at the sub-cycle level.

[0117] Sampling latch: When the trigger signal edge arrives, it quickly samples the current state of the delay chain, converting the continuous time signal into discrete digital state information.

[0118] Gray code converter: Converts the state information of the delay chain from linear encoding to Gray code format and outputs the time value.

[0119] The formula for calculating timestamps is:

[0120]

[0121] Where, N clk Master counter value, T is the system clock cycle. fine The result of the delay chain calculation;

[0122] The time resolution δt is:

[0123]

[0124] Where n is the TDC bit width, and to ensure time resolution accuracy at the sub-nanosecond level, T is selected. clk =1 ns, n=10, then δt≈1 ps.

[0125] The time-to-digital conversion module can distinguish time differences of less than 10 ps.

[0126] 4. Photon arrival time resolution unit, including TCSPC model, noise modeling and suppression module and adaptive convolution filter, used to output valid photon events;

[0127] The following algorithm is used in the TCSPC model:

[0128] Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is:

[0129]

[0130] Count all detected photon events and construct a delay time histogram:

[0131]

[0132] in, : Delay time histogram, in time The count value indicates how many photons are in the time difference. It was detected at that time. The number of sampling cycles refers to how many excitation-detection cycles the system has performed. Each excitation may detect zero or one photon event. : No. In the second sampling, the detected first The arrival time of each photon Time reference of this excitation The delay time between them. An approximation of the Dirac delta function, used in... The histogram is accumulated and counted periodically. Since the actual data is discretely sampled, this... In reality, it is a time window function with a finite width, often represented as the count of a certain time bin. The time variable in the histogram is usually quantized as a time bin, and the resolution is determined by the TDC (time-to-digital converter).

[0133] The following algorithm is used in the noise modeling and suppression module:

[0134] Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is:

[0135]

[0136] in, Background noise over time The probability density at that time. : Average count rate of background noise (unit: events / time). : The time of photon arrival. This is the standard exponential distribution, used to describe the temporal distribution of random events occurring in the background noise.

[0137] Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope:

[0138]

[0139] Among them, P s (t): at time Signal probability density at that location The peak time of the effective signal, i.e., the expected arrival time. 1: The standard deviation of the signal's temporal distribution reflects its width. : Actual arrival time of photons. This formula follows a Gaussian (normal) distribution, simulating the distribution characteristics of real photon signals.

[0140] Introducing a weighted filter W(t) to suppress background:

[0141]

[0142] in, For regularization terms;

[0143] The adaptive convolutional filtering (ATCF) algorithm is used in the adaptive convolutional filter:

[0144] Define a time window function G(t) to enhance repetitive signals:

[0145]

[0146] The convolution result is:

[0147]

[0148] Where t-τ reflects the time interval from the past time τ to the current time t;

[0149] When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold.

[0150] Furthermore, the photon arrival time resolution unit adopts a hardware-coordinated acceleration structure, namely an FPGA+ASIC hybrid architecture. The ASIC is used to implement front-end precision timing tasks such as SPAD array integration, TDC logic (including the main counter, delay chain, etc.), and high-speed timestamp generation, while also performing preliminary noise threshold screening. The FPGA is responsible for processing logic such as TCSPC algorithm implementation, histogram construction, background modeling and filtering, adaptive convolution, image compression, and system data scheduling. The two work together through a high-speed interface to achieve real-time processing of large-scale parallel detection data and image output. The FPGA is used for timing control and real-time data processing, while the ASIC is used to embed the core logic of SPAD and TDC, improving data throughput. This hardware-coordinated acceleration structure is suitable for the rapid construction of large-scale two-dimensional images.

[0151] 5. Data Compression and Fitting Module (DCF), used to construct two-dimensional images;

[0152] 6. Spatial noise reduction filter, used to perform Gaussian-Bayes filtering in two-dimensional image space, smoothing spatial noise in the image and enhancing imaging effect;

[0153] Use the following formula:

[0154]

[0155] in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the rate of Gaussian weight decay, determining the smoothness (the larger the value, the stronger the smoothing). Ω represents the standard deviation parameter controlling the rate of Gaussian weight decay, determining the smoothness. The neighborhood window centered on the center is usually a window of size 1000. In the region, Z is a normalization factor that makes the sum of all weights equal to 1, ensuring that the output intensity value is reasonable.

[0156] 7. Output unit, used to output two-dimensional image results;

[0157] 8. The main control processing unit (MCU) is used to send the two-dimensional image results to the display unit for display, and to send the two-dimensional image results to the storage unit for storage through the data interface.

[0158] Example 2: Figures 1-5 As shown, the ultra-weak light detection method based on SPAD array and time-correlated photon counting includes the following steps:

[0159] Step 1: Laser pulse triggering;

[0160] Step 2: The two-dimensional SPAD array detector receives the ultra-weak light signal and triggers an avalanche, and triggers the time-to-digital converter to record the photon arrival time;

[0161] Each SPAD pixel structure of the two-dimensional SPAD array detector includes an avalanche diode body (AD) and an active quenching circuit (AQC).

[0162] See active quenching circuit Figure 2 Bias voltage V BIAS One end is connected to the anode of the avalanche diode body, with a bias voltage V. BIAS The other end is connected to the enable (EN) terminal. The avalanche diode's cathode is grounded, and the avalanche diode's cathode is also connected to the drain of the field-effect transistor (FET). The FET's source is grounded. One end of resistor R is connected to the FET's drain, and the other end of resistor R is connected to the inverting input of the operational amplifier. The non-inverting input of the operational amplifier is connected to the reference voltage V. REF The other end of the EN terminal is connected to the output terminal, and the output terminal outputs the signal processed by the operational amplifier.

[0163] Each SPAD pixel in the two-dimensional SPAD array detector uses a silicon-based avalanche structure and operates at a voltage of V. op :

[0164] V op =V br +ΔV

[0165] Among them, V br ΔV is the avalanche breakdown voltage, and ΔV is the overvoltage, with a value ranging from 2 to 5V.

[0166] When the SPAD reaches its breakdown condition, an avalanche response is generated, triggering a time-to-digital converter to record the time. When an incident photon triggers an avalanche in the SPAD, the voltage at the SPAD cathode drops rapidly. Upon detecting this voltage change, the operational amplifier immediately outputs a high-level signal to turn on the MOSFET, grounding the SPAD cathode and rapidly discharging the avalanche current, thus terminating the avalanche process. After a preset quenching time t... q Afterwards, the MOSFET is disconnected, the SPAD is reconnected to the bias voltage VBIAS, and the system returns to the probe state, completing a full quench-reset cycle.

[0167] The quenching cycle is usually:

[0168]

[0169] in, For junction capacitance, To restore voltage.

[0170] The photon detection probability (PDP) of each SPAD pixel structure can be expressed as:

[0171]

[0172] Where λ is the photon wavelength, P in P is the incident light power. det This refers to the actual measured optical power.

[0173] Step 3: The Time-to-Digital Converter (TDC) converts the arrival times of each photon into precise timestamps;

[0174] The time-to-digital converter (TDC) includes a master counter, a delay chain, a sample latch, and a Gray code converter;

[0175] The main counter uses the system clock as the timing reference to achieve a coarse quantization and recording of time.

[0176] The delay chain uses a pseudo-random time delay line (TDL) or digital phase-locked loop (DLL) structure to subdivide the system clock cycle at the sub-cycle level.

[0177] Sampling latch: When the trigger signal edge arrives, it quickly samples the current state of the delay chain, converting the continuous time signal into discrete digital state information.

[0178] Gray code converter: Converts the state information of the delay chain from linear encoding to Gray code format and outputs the time value.

[0179] The formula for calculating timestamps is:

[0180]

[0181] Where, N clk Master counter value, T is the system clock cycle. fine The result of the delay chain calculation;

[0182] The time resolution δt is:

[0183]

[0184] Where n is the TDC bit width, and to ensure time resolution accuracy at the sub-nanosecond level, T is selected. clk =1 ns, n=10, then δt≈1 ps.

[0185] The time-to-digital conversion module can distinguish time differences of less than 10 ps.

[0186] Step 4: The photon arrival time analysis unit calculates the delay time and constructs a delay time histogram through the TCSPC model. It performs background noise modeling and effective signal modeling through the noise modeling and suppression module. At the same time, a weighted filter is introduced to suppress the background, and an effective photon event is output through an adaptive convolution filter.

[0187] The following algorithm is used in the TCSPC model:

[0188] Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is:

[0189]

[0190] Count all detected photon events and construct a delay time histogram:

[0191]

[0192] in, Delay time histogram Number of samples : No. In the second sampling, the detected first The arrival time of each photon Time reference of this excitation The delay time between; Approximation of the Dirac delta function; The time variable in the histogram;

[0193] The following algorithm is used in the noise modeling and suppression module:

[0194] Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is:

[0195]

[0196] in, Background noise over time The probability density at time , : Average count rate of background noise The time it takes for a photon to arrive;

[0197] Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope:

[0198]

[0199] Among them, P s (t): at time Signal probability density at that location Peak time of the effective signal 1: Standard deviation of the signal's temporal distribution : Actual arrival time of photons;

[0200] Introducing a weighted filter W(t) to suppress background:

[0201]

[0202] in, For regularization terms;

[0203] The adaptive convolutional filtering (ATCF) algorithm is used in the adaptive convolutional filter:

[0204] Define a time window function G(t) to enhance repetitive signals:

[0205]

[0206] The convolution result is:

[0207]

[0208] Where t-τ reflects the time interval from the past time τ to the current time t;

[0209] When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold.

[0210] Step 5: Construct a two-dimensional image. If the target is a low-reflectivity object with strong light absorption characteristics (such as black silicon), the two-dimensional image is further input into a spatial noise reduction filter to perform Gaussian-Bayes filtering, smoothing spatial noise in the image and enhancing the imaging effect.

[0211] Use the following formula:

[0212]

[0213] in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the Gaussian weight decay rate, and Ω represents the standard deviation parameter controlling the Gaussian weight decay rate. Z is the neighborhood window centered on the center, and Z is the normalization factor.

[0214] Step 6: Output the two-dimensional image result to the main control processing unit (MCU). The two-dimensional image result can be sent to the display unit for display, or it can be sent to the storage unit for storage through the data interface.

[0215] Furthermore, due to the extremely low reflectivity of black silicon surfaces, approximately 10... -3 The following detection optimization methods can also be introduced:

[0216] (1) Laser beam incident angle in Set it to be close to the surface normal to reduce scattering loss;

[0217] (2) Increase the number of sampling times N and use a time superposition strategy to improve the cumulative statistical reliability of photons.

[0218] This invention employs a two-dimensional SPAD array detector (SPAD) and a time-to-digital converter (TDC) to form the detection unit. Combined with a TCSPC model, noise modeling and suppression module, and adaptive convolutional filter, it effectively suppresses background noise and improves the ability to identify extremely weak signals. It exhibits excellent signal restoration performance even under conditions of strong ambient light interference and high noise levels. It is adaptable to the high absorption characteristics of low-reflectivity regions (such as black silicon surfaces), significantly enhancing the detection capability of weakly reflected light and achieving high-sensitivity imaging and quantitative detection of extremely low-reflectivity materials (such as black silicon). This invention's system supports black silicon materials with reflectivity below 10. -3 Effective imaging under these conditions, with a minimum measurable reflectance of 7×10⁻⁶. -5The system is highly integrated, supports on-chip parallel TCSPC processing, and features good real-time performance, high temporal resolution, high signal-to-noise ratio, and strong on-chip processing capabilities. It can employ a hardware-co-acceleration architecture, making it suitable for the rapid construction of large-scale 2D images.

[0219] Experimental Example 1: The performance indicators of the system of the present invention are shown in Table 1.

[0220] Table 1

[0221]

[0222] Experiment Example 2: Testing Experiment of Low Reflectance Black Silicon Samples.

[0223] Laser: Pulse width 100 ps, ​​repetition frequency 10 MHz;

[0224] Reflectivity of black silicon target material: 0.07%;

[0225] Detector: The system of this invention;

[0226] Comparison system: Single-channel time-correlated photon counting (TCSPC) technology + photomultiplier tube (PMT).

[0227] The results are shown in Table 2.

[0228] Table 2

[0229]

[0230] It can be seen that the present invention can still accurately reconstruct the photon time distribution spectrum under ultra-weak signal conditions, proving its applicability under low reflectivity conditions.

[0231] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An ultra-weak light detection system based on SPAD array and time-correlated photon counting, characterized in that, include: Trigger light source; A multi-channel SPAD array detection unit is used to capture ultra-weak light signals and trigger a time-to-digital conversion module to record the photon arrival time; The time-to-digital conversion module is used to convert the arrival time of each photon into a timestamp; The photon arrival time resolution unit includes: a TCSPC model for calculating delay time and constructing a delay time histogram; a noise modeling and suppression module for background noise modeling and effective signal modeling, while introducing a weighted filter to suppress the background; and an adaptive convolution filter for outputting effective photon events. The following algorithm is used in the TCSPC model: Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is: ； Count all detected photon events and construct a delay time histogram: ； in, Delay time histogram Number of samples : No. In the second sampling, the detected first The arrival time of each photon Reference time of this excitation The delay time between; Approximation of the Dirac delta function; :time; The following algorithm is used in the noise modeling and suppression module: Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is: ； in, Background noise over time The probability density at time , : Average count rate of background noise; Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope: ； Among them, P s (t): at time Signal probability density at that location Peak time of the effective signal 1: Standard deviation of the signal's temporal distribution; Introducing a weighted filter W(t) to suppress background: ； in, For regularization terms; Using adaptive convolution filtering algorithms in adaptive convolution filters: Define a time window function G(t) to enhance repetitive signals: ； The convolution result is: ； Where t-τ reflects the time interval from the past time τ to the current time t; When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold. The data compression and fitting module is used to construct two-dimensional images; For low-reflectivity targets with strong light absorption characteristics, the spatial denoising filter performs Gaussian-Bayes filtering in the two-dimensional image space to smooth spatial noise in the image. The output unit is used to output two-dimensional image results; And the main control processing unit.

2. The ultra-weak light detection system based on SPAD array and time-correlated photon counting according to claim 1, characterized in that, The multi-channel SPAD array detection unit adopts a two-dimensional SPAD array detector. Each SPAD pixel structure includes an avalanche diode body and an active hardening circuit. The active hardening circuit structure is as follows: bias voltage V BIAS One end is connected to the anode of the avalanche diode body, with a bias voltage V. BIAS The other end is connected to the enable (EN) terminal. The avalanche diode's cathode is grounded, and the avalanche diode's cathode is also connected to the drain of the field-effect transistor (FET). The FET's source is grounded. One end of resistor R is connected to the FET's drain, and the other end of resistor R is connected to the inverting input of the operational amplifier. The non-inverting input of the operational amplifier is connected to the reference voltage V. REF Connect the other end of EN to the output terminal.

3. The ultra-weak light detection system based on SPAD array and time-correlated photon counting according to claim 1, characterized in that, The time-to-digital converter module uses a time-to-digital converter, which includes a main counter, a delay chain, a sampling latch, and a Gray code converter.

4. The ultra-weak light detection system based on SPAD array and time-correlated photon counting according to claim 3, characterized in that, The formula for calculating timestamps is: ； Where, N clk Master counter value, T is the system clock cycle. fine The result of the delay chain calculation; The time resolution δt is: ； Where n is the TDC bit width.

5. The ultra-weak light detection system based on SPAD array and time-correlated photon counting according to claim 1, characterized in that, The photon arrival time resolution unit adopts a hardware co-acceleration structure, namely an FPGA+ASIC hybrid architecture.

6. The ultra-weak light detection system based on SPAD array and time-correlated photon counting according to claim 5, characterized in that, Gaussian-Bayes filtering is performed in the spatial denoising filter of a two-dimensional image input using the following formula: ； in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the Gaussian weight decay rate, and Ω represents the standard deviation parameter controlling the Gaussian weight decay rate. Z is the neighborhood window centered on the center, and Z is the normalization factor.

7. A method for detecting ultra-weak light based on SPAD array and time-correlated photon counting, characterized in that, Includes the following steps: Step 1: Laser pulse triggering; Step 2: The two-dimensional SPAD array detector receives the ultra-weak light signal and triggers an avalanche, and triggers the time-to-digital converter to record the photon arrival time; Step 3: The time-to-digital converter converts the arrival time of each photon into a timestamp; Step 4: The photon arrival time analysis unit calculates the delay time and constructs a delay time histogram through the TCSPC model. It performs background noise modeling and effective signal modeling through the noise modeling and suppression module. At the same time, a weighted filter is introduced to suppress the background, and an effective photon event is output through an adaptive convolution filter. The following algorithm is used in the TCSPC model: Let the time of the i-th excitation light emission be t. i （0） The photon detection time is t i （k） The delay time is: ； Count all detected photon events and construct a delay time histogram: ； in, Delay time histogram Number of samples : No. In the second sampling, the detected first The arrival time of each photon Time reference of this excitation The delay time between; Approximation of the Dirac delta function; :time; The following algorithm is used in the noise modeling and suppression module: Background noise modeling: Assume the background noise is a Poisson process with an average count rate λ. b Its probability density function is: ； in, Background noise over time The probability density at time , : Average count rate of background noise The time it takes for a photon to arrive; Effective signal modeling: Assuming the effective signal peak value is around t0, the model is a Gaussian envelope: ； Among them, P s (t): at time Signal probability density at that location Peak time of the effective signal 1: Standard deviation of the signal's temporal distribution; Introducing a weighted filter W(t) to suppress background: ； in, For regularization terms; Using adaptive convolution filtering algorithms in adaptive convolution filters: Define a time window function G(t) to enhance repetitive signals: ； The convolution result is: ； Where t-τ reflects the time interval from the past time τ to the current time t; When S(t)≥θ, the photon event is considered a valid photon event, where θ is the signal extraction threshold. Step 5: Construct a two-dimensional image. For targets with strong light absorption and low reflectivity, the two-dimensional image is further input into a spatial denoising filter to perform Gaussian-Bayes filtering and smooth the spatial noise of the image. Step 6: Output the two-dimensional image result to the main control processing unit.

8. The ultra-weak light detection method based on SPAD array and time-correlated photon counting according to claim 7, characterized in that, In step five, the two-dimensional image is further input into a spatial denoising filter to perform Gaussian-Bayes filtering, using the following formula: ； in, This represents the grayscale or intensity value of the current pixel. Let I(i,j) be the position of the current pixel, and let I(i,j) be the number of pixels in the neighborhood. The grayscale value or intensity value, 2 represents the standard deviation parameter controlling the Gaussian weight decay rate, and Ω represents the standard deviation parameter controlling the Gaussian weight decay rate. Z is the neighborhood window centered on the center, and Z is the normalization factor.

Images