Single-guide antenna pulse terahertz time-domain spectroscopy system and detection method

Abstract

The application discloses a single light guide antenna pulse terahertz time-domain spectroscopy system and a detection method, and relates to the technical field of terahertz. The system comprises a laser pulse generation module, a single light guide antenna module, a compressed sampling module, a signal reconstruction module and a control unit. The application integrates a random light delay modulation unit and a single-pixel detector, realizes random time interval sampling of a terahertz signal, effectively breaks through the limitation of a traditional sampling mode, improves the flexibility and efficiency of signal acquisition, and combines a signal reconstruction module of an FPGA+GPU heterogeneous computing architecture. The system can adaptively construct a sparse base, accurately sparsely represent a measurement vector, and then efficiently reconstruct an original terahertz time-domain signal through an optimization objective function and an adaptive iterative solution algorithm. The application not only improves the accuracy of signal reconstruction, but also reduces the complexity of data processing, and provides strong support for the practical application of terahertz technology.

Description

Technical Field

[0001] This invention relates to the field of terahertz technology, specifically to a pulsed terahertz time-domain spectroscopy system and detection method using a single optical guide antenna. Background Technology

[0002] With the widespread application of the terahertz (THz) band in security inspection, biomedical imaging, non-destructive testing, and high-speed communication, the performance requirements for terahertz time-domain spectroscopy systems are increasing. Traditional terahertz time-domain spectroscopy systems typically rely on mechanical scanning or high-resolution analog-to-digital converters for signal acquisition, but these methods suffer from high equipment complexity, high cost, and limited sampling rates. In particular, in ultra-wideband terahertz signal detection, traditional methods struggle to balance the requirements of high time resolution and low hardware cost. Furthermore, terahertz signals are susceptible to environmental noise interference during transmission, leading to a decline in the quality of the reconstructed signal and limiting their application in real-time detection and complex scenarios. Therefore, developing an efficient, low-cost, and interference-resistant terahertz time-domain spectroscopy system has become a pressing technical challenge for the industry.

[0003] Traditional terahertz time-domain spectroscopy systems mainly rely on two technical approaches: one is to achieve time-resolved sampling through high-speed mechanical scanning devices, but the inertia of mechanical components limits the sampling rate, and the system is bulky and has poor stability; the other is to use high-resolution analog-to-digital converters for direct sampling, but the bandwidth and accuracy of the ADC cannot simultaneously meet the requirements of ultra-wideband terahertz signals, leading to a sharp increase in hardware costs. In addition, traditional signal reconstruction algorithms are usually based on fixed basis functions and cannot adapt to the dynamic attenuation characteristics of terahertz signals, resulting in distortion or noise amplification problems in the reconstructed signal. In real-time detection scenarios, traditional systems are prone to inter-module synchronization errors due to insufficient timing control accuracy, further reducing system reliability. To address these shortcomings, the industry urgently needs a terahertz time-domain spectroscopy solution that combines high efficiency, low cost, and adaptability. Therefore, a pulsed terahertz time-domain spectroscopy system and detection method with a single optical guide antenna were developed. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a pulsed terahertz time-domain spectroscopy system and detection method with a single optical guide antenna. By integrating laser pulse generation, random optical delay compression sampling, FPGA+GPU heterogeneous signal reconstruction and high-precision timing control modules, the system achieves efficient acquisition and adaptive reconstruction of terahertz signals. The system uses a MEMS micromirror array to break through the traditional sampling rate limitation and combines an adaptive sparse basis algorithm to optimize signal representation, thereby improving reconstruction accuracy and noise resistance.

[0005] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On the one hand, a pulsed terahertz time-domain spectroscopy system with a single optical guide antenna, the system comprising a laser pulse generation module, a single optical guide antenna module, a compressed sampling module, a signal reconstruction module, and a control unit;

[0006] The laser pulse generation module is a Ti:sapphire mode-locked laser that outputs femtosecond laser pulses and is equipped with an optical transmission system, a laser power monitoring and feedback system, and a temperature control subsystem.

[0007] The single optical guide antenna module is an optical guide antenna based on a semi-insulating gallium arsenide substrate, equipped with an adjustable DC bias power supply and an optical path alignment structure. The electrodes are fabricated using a special process and covered with an insulating layer.

[0008] The electrodes of the optical antenna are made of Au / Ge / Ni alloy material and are formed by electron beam evaporation, with a total thickness of 900nm, including an Au layer of 600nm, a Ge layer of 200nm, and a Ni layer of 100nm. The electrode surface is covered with a Si3N4 insulating film prepared by plasma-enhanced chemical vapor deposition, with a film thickness of 200nm and a breakdown voltage ≥500V.

[0009] The compressed sampling module consists of a random optical delay modulation unit and a single-pixel detector. The random optical delay modulation unit uses a MEMS micromirror array to achieve random time interval sampling, and the single-pixel detector is a liquid nitrogen-cooled mercury cadmium telluride detector.

[0010] The signal reconstruction module adopts an FPGA+GPU heterogeneous computing architecture, receives data collected by the compressed sampling module, processes it, and generates a reconstructed signal. The FPGA transmits the preprocessed data to the GPU memory in real time through the PCIe interface. The GPU calls the CUDA kernel to execute the sparse basis construction and iterative solution algorithm in parallel, and the calculation results are returned to the FPGA via DMA.

[0011] The control unit, based on the STM32H743 microprocessor, coordinates the timing of the entire system through a synchronous trigger signal and communicates with each module. The control unit sends synchronous clock pulses to the laser pulse generation module, the compression sampling module, and the signal reconstruction module through a hardware trigger signal line, with timing jitter ≤1ps.

[0012] Further, in the laser pulse generation module, the optical transmission system includes three groups of high-precision mirrors, one adjustable-focus lens, and one group of diaphragms. The surfaces of the mirrors are coated with an anti-reflection film in the 800 nm band. The focal length adjustment range of the adjustable-focus lens is 50 - 200 mm, and the aperture of the diaphragm can be continuously adjusted within the range of 0.1 - 5 mm. The laser power monitoring and feedback system includes an InGaAs photodetector, a power comparison circuit, and a pump source adjustment unit. The response wavelength range of the detector is 700 - 1100 nm. The temperature control subsystem consists of a thermocouple sensor, a PID temperature control circuit, and a semiconductor refrigerator. The temperature measurement accuracy of the sensor is ±0.05 °C, and the adjustment resolution of the temperature control circuit is 0.01 °C.

[0013] Furthermore, in the single photoconductive antenna module, the optical path alignment structure includes a three-dimensional adjustable translation stage, a laser collimator, and a position sensing detector. The adjustment range of the three-dimensional adjustable translation stage is ±5 mm for each of the X / Y / Z axes, and the adjustment accuracy is ≤1 μm. The output wavelength of the laser collimator is consistent with the laser center wavelength of the laser pulse generation module. The spatial resolution of the position sensing detector is ≤10 μm, and it can real-time feedback the focusing position of the laser spot on the electrode gap of the photoconductive antenna.

[0014] Furthermore, in the compressive sampling module: The random optical delay modulation unit is a MEMS micromirror array containing 1024 independently driven micromirror units. The size of each micromirror unit is 10 μm × 10 μm, and it uses an electrostatic drive method to achieve deflection within the range of ±10°, and the deflection response time is ≤1 μs. The drive circuit supporting the micromirror array enables 16-bit precision control of the deflection angle, and realizes random time interval sampling within the range of 0 - 1200 ps by adjusting the micromirror deflection angle. The time interval adjustment step is 1 ps.

[0015] Furthermore, in the compressive sampling module, the mercury cadmium telluride chip of the single pixel detector has a size of 50 μm × 50 μm, the cut-off wavelength is ≥12 μm, and the operating temperature is maintained at 77 K ± 0.1 K by a closed-loop controlled liquid nitrogen dewar. The detector is equipped with a cryogenic preamplifier, the response bandwidth covers 0.1 - 3 THz, and the output signal is transmitted to the subsequent processing module after 50 Ω impedance matching.

[0016] Furthermore, in the signal reconstruction module, the specific steps for processing the compressive sampling data and generating a reconstructed signal are as follows:

[0017] (1) The FPGA preprocesses the measurement vector, including data verification, format conversion, and preliminary filtering;

[0018] (2) The FPGA transmits the preprocessed data to the GPU. The GPU calls the adaptive sparse basis construction algorithm to generate a sparse basis that is adapted to the current terahertz signal. The sparse basis is used to sparsely represent the measurement vector and convert the original signal into a sparse wavelet coefficient vector.

[0019] (3) Based on the sparse representation results, the GPU constructs an optimization objective function that includes L1 norm regularization, data fidelity terms and variational mode constraints;

[0020] (4) The GPU calls the adaptive iterative solution algorithm to solve the optimization objective function. By dynamically adjusting the iteration step size, the wavelet coefficients and the reconstructed signal are updated alternately to complete the iterative calculation.

[0021] (5) After obtaining the reconstructed wavelet coefficients through iterative calculation, the reconstructed signal of the original terahertz time domain signal is obtained through inverse wavelet transform;

[0022] (6) Evaluate the quality of the reconstructed signal and calculate the root mean square error (RMSE) and correlation coefficient between the reconstructed signal and the reference signal. When the RMSE exceeds 2% or the correlation coefficient is less than 0.95, the reconstructing process can be triggered.

[0023] Furthermore, in the signal reconstruction module, the calculation formula for the adaptive sparse basis construction algorithm is as follows: Among them: Ψ THz (τ,t) is the adaptive sparse basis function for terahertz signals, a function of time offset τ and time t. τ is the time offset, determined dynamically by the signal attenuation; t is the time variable, dynamically adjusted with signal attenuation; k is the wavelet component order index, ranging from 0 to N-1; N is the total wavelet component order; ω... k Let ω be the weighting coefficient of the k-th wavelet component. Perform multi-scale decomposition on the terahertz signal and calculate the proportion of each wavelet component in the total signal energy. This proportion is the corresponding weighting coefficient, and ∑ω k =1, ψ(.) is the basic wavelet function, s k E is a scaling factor, determined based on the frequency distribution of the terahertz pulse. k The energy proportion of the k-th order component is determined by calculating the proportion of the energy of each wavelet component k in the total energy through energy analysis of the terahertz signal. β is the attenuation adjustment coefficient, which is determined by an optimization algorithm based on the attenuation law of the terahertz signal, and its value ranges from 0.1 to 0.5.

[0024] Furthermore, in the signal reconstruction module, the formula for calculating the objective function is: Where: r is the sparse wavelet coefficient vector, u hLet h be the h-th modal component, min is the minimum operation used to optimize the objective function, K is the total order of the modal components, h ranges from 1 to K, and α is the minimum value. h Let |r| be the adaptive L1 regularization coefficient corresponding to the h-th modal component. h ||1 is the h-th order wavelet coefficient vector r h The L1 norm, δ h Let λ be the modal constraint factor, λ be the data fidelity term weight, y be the measurement signal vector, R(t) be a transformation or operator related to time t, and Φ be the measurement matrix. It is a terahertz rarefaction base Ψ THz The false rebellion, The square of the L2 norm of a vector, h-th modal component u h The partial derivative with respect to time t, δ h It is the mode constraint factor corresponding to the h-th order mode component.

[0025] Furthermore, in the signal reconstruction module, the calculation formula of the adaptive iterative solution algorithm is as follows: Where: r (n) It is the wavelet coefficient vector at the nth iteration, r (n+1) It is the wavelet coefficient vector at the (n+1)th iteration, soft(·) is the soft thresholding function, and μ (n) It is the step size of the nth iteration. The wavelet coefficient vector r of the objective function J in the nth iteration (n) The gradient at μ (n+1) This is the step size for the (n+1)th iteration, and v is the step size adjustment coefficient, with a value ranging from 0.8. <v<1.2,||r (n+1) -r (n) ||2 is the L2 norm of the difference between the wavelet coefficient vectors of the (n+1)th and nth iterations, ||r (n) ||2 is the wavelet coefficient vector r of the nth iteration. (n) The L2 norm, α h is the adaptive L1 regularization coefficient corresponding to the h-th modal component, and n represents the number of iterations.

[0026] On the other hand, the pulsed terahertz time-domain spectral detection method using a single optical guide antenna involves the following specific steps:

[0027] Laser pulse generation and stabilization control: When the laser pulse generation module is working, the Ti:sapphire mode-locked laser outputs femtosecond laser pulses. The optical transmission system transmits, focuses, and adjusts the laser pulses. The laser power monitoring and feedback system monitors the laser power in real time and performs power stabilization control through the pump source adjustment unit. The temperature control subsystem maintains the laser's operating temperature stably.

[0028] Terahertz signal generation: The laser pulse enters the single optical guide antenna module, and the optical path alignment structure precisely focuses the laser spot on the gap between the electrodes of the optical guide antenna. The adjustable DC bias power supply provides bias voltage to the optical guide antenna, and the optical guide antenna generates a terahertz signal under the action of laser pulse irradiation and bias voltage.

[0029] Terahertz signal sampling and detection: The compressed sampling module starts working, the random optical delay modulation unit realizes random time interval sampling of the terahertz signal based on the MEMS micromirror array, and the single pixel detector detects the sampled terahertz signal and transmits the detection signal to the signal reconstruction module.

[0030] Signal reconstruction and quality assessment: The signal reconstruction module receives the probe signal, the FPGA preprocesses the measurement vector, and then transmits it to the GPU; the GPU calls the adaptive sparse basis construction algorithm to generate a sparse basis, and performs sparse representation of the measurement vector to obtain the wavelet coefficient vector; then it constructs the optimization objective function, calls the adaptive iterative solution algorithm to solve it, obtains the reconstructed signal through wavelet inverse transform, and evaluates the quality of the reconstructed signal. If the requirements are not met, it is reconstructed.

[0031] System timing coordination control: The control unit coordinates the working timing of each module in the entire system through synchronous trigger signals to ensure that each module works in coordination.

[0032] Compared with existing technologies, the pulsed terahertz time-domain spectroscopy system and detection method using a single optical guide antenna have the following advantages:

[0033] I. This invention integrates a random optical delay modulation unit and a single-pixel detector, enabling random time interval sampling of terahertz signals. This effectively overcomes the limitations of traditional sampling methods, improving the flexibility and efficiency of signal acquisition. Combined with a signal reconstruction module based on an FPGA+GPU heterogeneous computing architecture, the system can adaptively construct sparse bases to accurately represent measurement vectors. Furthermore, by optimizing the objective function and using an adaptive iterative solution algorithm, the original terahertz time-domain signal is efficiently reconstructed. This not only improves the accuracy of signal reconstruction but also reduces the complexity of data processing, providing strong support for the practical application of terahertz technology.

[0034] Second, this invention achieves precise timing coordination control of various modules, including laser pulse generation, terahertz signal generation, sampling detection, and signal reconstruction, through a control unit based on the STM32H743 microprocessor. This design ensures seamless connection and efficient collaboration between modules, improving the overall stability and reliability of the system. Furthermore, the system is equipped with a laser power monitoring feedback system and a temperature control subsystem, further enhancing the stability and accuracy of laser pulse generation and providing a solid guarantee for obtaining high-quality terahertz signals. These aspects collectively drive the development of terahertz technology towards higher performance and wider applications.

[0035] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0036] 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.

[0037] Figure 1 A schematic diagram of the pulsed terahertz time-domain spectroscopy system with a single optical guide antenna;

[0038] Figure 2 A flowchart of a pulsed terahertz time-domain spectral detection method using a single optical guide antenna;

[0039] Figure 3 This is a schematic diagram of signal flow in a pulsed terahertz time-domain spectroscopy system with a single optical guide antenna. Detailed Implementation

[0040] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0041] Example 1:

[0042] Pulse terahertz time-domain spectroscopy system with a single optical guide antenna.

[0043] The system consists of a laser pulse generation module, a single optical guide antenna module, a compressed sampling module, a signal reconstruction module, and a control unit. These modules work collaboratively to generate, detect, sample, and reconstruct terahertz signals. The specific details are as follows, and the process is as follows: Figure 1 As shown.

[0044] Laser Pulse Generation Module: Centered on a Ti:sapphire mode-locked laser, this module outputs femtosecond laser pulses, serving as the fundamental light source for the entire system. It provides the necessary excitation for subsequent terahertz signal generation. The equipped optical transmission system precisely guides the laser's propagation path, ensuring efficient and stable transmission of laser energy to the single optical guide antenna module, minimizing transmission losses and deviations. The laser power monitoring and feedback system monitors the laser's output power in real time, ensuring its stability and preventing power fluctuations from affecting the terahertz signal generation quality. The temperature control subsystem precisely controls the laser's operating temperature, keeping temperature fluctuations within a minimal range. Temperature variations affect the laser's output wavelength and pulse characteristics; a stable temperature environment is crucial for stable laser performance. During continuous 8-hour operation, the temperature control subsystem stabilizes the laser temperature within ±0.03℃, and the laser power monitoring and feedback system ensures power fluctuations are ≤±1.5%, providing a stable excitation source for terahertz signal generation.

[0045] Single-beamguide antenna module: Centered on a semi-insulating gallium arsenide substrate, this is a key component for terahertz signal generation and reception. An adjustable DC bias power supply provides a suitable bias electric field for the beamguide antenna. Adjusting the bias voltage optimizes the intensity and waveform of the terahertz signal, improving signal generation efficiency. The optical path alignment structure enables precise focusing of the laser spot across the gap between the beamguide antenna electrodes, ensuring maximum laser energy is applied to the antenna, enhancing the excitation efficiency and stability of the terahertz signal. Real-time feedback of the focusing position facilitates timely adjustments. The electrodes utilize Au / Ge alloys. The / Ni alloy material is deposited using an electron beam evaporation process, with a total thickness of 900nm, including a 600nm Au layer, a 200nm Ge layer, and a 100nm Ni layer. The electrode surface is covered with a 200nm thick Si3N4 insulating film prepared by plasma-enhanced chemical vapor deposition, with a breakdown voltage ≥500V. This ensures good conductivity and adhesion to the substrate, while the insulating layer effectively prevents breakdown and leakage between electrodes, improving the antenna's reliability and lifespan, and ensuring stable operation even under high bias voltages.

[0046] The compressed sampling module, consisting of a random optical delay modulation unit and a single-pixel detector, is responsible for the efficient acquisition of terahertz signals. The random optical delay modulation unit uses a MEMS micromirror array to achieve random time interval sampling. This sampling method can acquire signals at a rate much lower than the Nyquist sampling rate, greatly reducing the amount of data while retaining key information of the signal and improving sampling efficiency. The single-pixel detector uses a liquid nitrogen-cooled mercury cadmium telluride detector. Liquid nitrogen cooling can reduce the noise level of the detector and improve its sensitivity to weak terahertz signals, enabling it to detect even weaker signals. The detector is equipped with a cryogenic preamplifier that amplifies the detected weak signal while ensuring the signal's bandwidth characteristics. After impedance matching, the signal is transmitted to subsequent modules to ensure that the signal is not distorted or attenuated during transmission. Based on a MEMS micromirror array, a random optical delay modulation unit enables random sampling within the 0-1200ps time range. The acquisition time of a single terahertz signal is reduced from 50ms in traditional mechanical scanning to 6.2ms, improving the sampling efficiency by about 8 times. With a liquid nitrogen-cooled mercury cadmium telluride detector, the system can detect a minimum terahertz signal power of 50pW.

[0047] Signal reconstruction module: Adopting an FPGA+GPU heterogeneous computing architecture, it undertakes the important task of recovering the original terahertz signal from compressed sampled data. The FPGA first preprocesses the measurement vector, including data verification to ensure data integrity, format conversion to meet the requirements of subsequent processing, and preliminary filtering to remove some noise, providing a high-quality data foundation for subsequent signal reconstruction.

[0048] The FPGA transmits the preprocessed data to the GPU, which then calls an adaptive sparse basis construction algorithm to generate a sparse basis adapted to the current terahertz signal. The calculation formula for the adaptive sparse basis construction algorithm is as follows: Among them: Ψ THz (τ,t) is the adaptive sparse basis function for terahertz signals, a function of time offset τ and time t. τ is the time offset, determined dynamically by the signal attenuation; t is the time variable, dynamically adjusted with signal attenuation; k is the wavelet component order index, ranging from 0 to N-1; N is the total wavelet component order; ω... k Let ω be the weighting coefficient of the k-th wavelet component. Perform multi-scale decomposition on the terahertz signal and calculate the proportion of each wavelet component in the total signal energy. This proportion is the corresponding weighting coefficient, and ∑ω k =1, ψ(.) is the basic wavelet function, s k E is a scaling factor, determined based on the frequency distribution of the terahertz pulse. kThe energy proportion of the k-th order component is determined by calculating the proportion of energy of each wavelet component k in the total energy through energy analysis of the terahertz signal. β is the attenuation adjustment coefficient, which is determined by optimization algorithm based on the attenuation law of the terahertz signal, with a value range of 0.1 < β < 0.5. This sparse basis is used to convert the original signal into a sparse wavelet coefficient vector. The purpose of this step is to represent the complex terahertz signal in a more concise form, which is convenient for subsequent processing and reconstruction.

[0049] Next, the GPU constructs an optimization objective function that includes L1 norm regularization, data fidelity terms, and variational mode constraints. The formula for calculating the optimization objective function is as follows: Where: r is the sparse wavelet coefficient vector, u h Let h be the h-th modal component, min is the minimum operation used to optimize the objective function, K is the total order of the modal components, h ranges from 1 to K, and α is the minimum value. h Let |r| be the adaptive L1 regularization coefficient corresponding to the h-th modal component. h ||1 is the h-th order wavelet coefficient vector r h The L1 norm, δ h Let λ be the modal constraint factor, λ be the data fidelity term weight, y be the measurement signal vector, R(t) be a transformation or operator related to time t, and Φ be the measurement matrix. It is a terahertz rarefaction base Ψ THz The false rebellion, The square of the L2 norm of a vector, h-th modal component u h The partial derivative with respect to time t, δ h It is the mode constraint factor corresponding to the h-th order mode component. This objective function can ensure the similarity between the reconstructed signal and the original signal, while making the reconstructed signal have sparsity and good modal characteristics.

[0050] The GPU uses an adaptive iterative solution algorithm to solve the objective function. The calculation formula for the adaptive iterative solution algorithm is as follows: Where: r (n) It is the wavelet coefficient vector at the nth iteration, r (n+1) It is the wavelet coefficient vector at the (n+1)th iteration, soft(·) is the soft thresholding function, and μ (n) It is the step size of the nth iteration. The wavelet coefficient vector r of the objective function J in the nth iteration (n) The gradient at μ (n+1) This is the step size for the (n+1)th iteration, and v is the step size adjustment coefficient, with a value ranging from 0.8. <v<1.2,||r (n+1) -r (n)||2 is the L2 norm of the difference between the wavelet coefficient vectors of the (n+1)th and nth iterations, ||r (n) ||2 is the wavelet coefficient vector r of the nth iteration. (n) The L2 norm, α h denoted as the adaptive L1 regularization coefficient corresponding to the h-th modal component, and n represents the iteration number. By dynamically adjusting the iteration step size, the wavelet coefficients and the reconstructed signal are updated alternately to complete the iterative calculation. This process can gradually approach the optimal solution and improve the accuracy of the reconstructed signal.

[0051] After obtaining the reconstructed wavelet coefficients, the reconstructed signal of the original terahertz time-domain signal is obtained through inverse wavelet transform. Finally, the quality of the reconstructed signal is evaluated. When the evaluation index fails to meet the standard, the reconstruction process is automatically started to ensure that the output reconstructed signal has high quality and meets the needs of subsequent analysis and application. When reconstructing a standard terahertz pulse signal with a center frequency of 1THz and a pulse width of 50ps, the FPGA+GPU heterogeneous architecture makes the root mean square error (RMSE) of the reconstructed signal ≤1.2% and the correlation coefficient ≥0.98, which is significantly better than the traditional fixed basis function algorithm. In a noise environment with a signal-to-noise ratio (SNR) of 10dB, it can still maintain an RMSE ≤2.5% and a correlation coefficient ≥0.95, which shows outstanding noise resistance.

[0052] The control unit, based on the STM32H743 microprocessor, is the central control hub of the entire system. It coordinates the timing of the entire system through synchronous trigger signals, ensuring precise timing coordination between laser pulse generation, terahertz signal generation, sampling, and reconstruction, thus preventing signal errors and loss due to timing misalignments. Simultaneously, the control unit communicates with each module through multiple interfaces, enabling real-time acquisition of operational status information such as laser power, temperature, and detector status. Figure 3 As shown, the system sends corresponding control commands based on this information to achieve precise control and adjustment of each module, ensuring the stable and efficient operation of the entire system.

[0053] In summary, the pulsed terahertz time-domain spectroscopy system with a single optical guide antenna consists of a laser pulse generation unit, a single optical guide antenna, compressed sampling, signal reconstruction, and a control unit. Each module has a clear division of labor: the laser module provides a stable femtosecond laser, the antenna module is responsible for the generation and reception of terahertz signals, the compressed sampling module efficiently acquires signals, the reconstruction module reconstructs signals using an adaptive sparse basis construction algorithm, and the synchronous trigger signal of the control unit ensures that the system timing jitter is ≤1ps, and the terahertz signal amplitude drift is ≤2% over 8 hours of continuous operation. When performing transmission detection on a 50μm thick polyethylene film, the peak position deviation of the reconstructed spectrum is ≤0.3ps, the relative error of the absorption peak intensity is ≤3%, and the characteristic absorption at 0.5THz and 1.2THz is successfully identified, meeting the precise detection requirements in practical applications.

[0054] Example 2:

[0055] A method for detecting pulsed terahertz time-domain spectra using a single optical guide antenna.

[0056] This detection method is applicable to the aforementioned pulsed terahertz time-domain spectroscopy system with a single optical guide antenna. By executing a series of sequential steps, it achieves accurate capture and reconstruction of the terahertz signal. The specific details of each step are as follows, and the process is as follows: Figure 3 As shown

[0057] System startup and parameter initialization: First, the entire system is powered on. The control unit will perform a comprehensive initialization operation on the laser pulse generation module, single optical guide antenna module, compressed sampling module, and signal reconstruction module. The purpose of this step is to set the initial operating parameters for each module, such as the initial power range of the laser, the initial value of the bias voltage of the optical guide antenna, the trigger mode of the sampling module, and the initial calculation conditions of the reconstruction module. This ensures that all modules start running from a unified reference state, avoiding deviations in the subsequent detection process due to chaotic initial parameters, and laying the foundation for the smooth progress of the entire detection process.

[0058] Laser pulse generation and transmission control: The laser pulse generation module is activated under the command of the control unit, and the Ti:sapphire mode-locked laser begins to output femtosecond laser pulses. The optical transmission system precisely guides the laser pulses to the single optical guide antenna module through internal reflectors and lens assemblies. Its function is to reduce energy loss and directional deviation of the laser during transmission, ensuring that the laser accurately reaches the target position. At the same time, the laser power monitoring and feedback system tracks the changes in laser power in real time. Once the power exceeds the set range, it is immediately corrected to ensure the stability of laser power, because power fluctuations directly affect the generation intensity of terahertz signals. The temperature control subsystem continuously maintains the laser's operating temperature stably, avoiding temperature changes that cause the laser wavelength or pulse width to drift, ensuring the stability of laser pulse characteristics. The synergistic effect of laser power monitoring and temperature control ensures that the laser power fluctuation is ≤±1.5% and the temperature is stable at ±0.03℃, providing a stable excitation source for subsequent terahertz signal generation.

[0059] Terahertz signal excitation and reception: After the laser pulse arrives at the single optical guide antenna module, the optical path alignment structure will precisely focus the laser spot on the electrode gap of the optical guide antenna through three-dimensional adjustment and platform calibration. Its function is to maximize the laser energy to act on the antenna material and improve the excitation efficiency of the terahertz signal. At this time, the adjustable DC bias power supply provides a suitable bias electric field for the optical guide antenna. Under the combined action of the laser and the bias electric field, the optical guide antenna will generate a terahertz signal. At the same time, the module will also receive externally transmitted terahertz signals (such as terahertz signals modulated by the sample under test). This step is the core of the entire detection process and directly determines the signal source quality of subsequent sampling and reconstruction.

[0060] Terahertz signal compression sampling processing: The compression sampling module starts working under the synchronous control of the control unit. The random optical delay modulation unit achieves random time interval sampling of the terahertz signal through the rapid deflection of the MEMS micromirror array. Its function is to break through the limitations of traditional sampling methods, significantly reduce the amount of sampled data while retaining the key features of the signal, improve sampling efficiency and reduce data storage pressure. The single-pixel detector detects the sampled terahertz signal and converts the optical signal into an electrical signal. Due to the use of liquid nitrogen cooling, the noise level of the detector is greatly reduced, enabling it to capture weak terahertz signals more sensitively. The signal is then processed by a preamplifier and impedance matching to ensure that the electrical signal is not distorted during transmission, providing high-quality raw data for subsequent signal reconstruction. Through the combination of random sampling of the MEMS micromirror array and cryogenic detector, the acquisition time of a single signal is shortened to 6.2ms, and a weak signal of 50pW can be detected, ensuring efficient and high-sensitivity signal acquisition.

[0061] Signal Reconstruction and Quality Verification: The compressed sampling module transmits the processed electrical signal to the signal reconstruction module. The FPGA first preprocesses the received data, including data verification to remove invalid data, format conversion to adapt to the GPU's computing requirements, and preliminary filtering to remove high-frequency noise. This process purifies the data, preparing it for subsequent high-precision reconstruction. Subsequently, the data is transmitted to the GPU, which uses an adaptive sparse basis construction algorithm to generate a sparse basis that matches the current terahertz signal. The calculation formula for the adaptive sparse basis construction algorithm is as follows: The original signal is converted into a sparse wavelet coefficient vector. This step simplifies the signal representation and highlights its key features. Next, the GPU constructs an optimization objective function that includes L1 norm regularization, data fidelity terms, and variational mode constraints. The formula for calculating the optimization objective function is as follows: The function aims to suppress noise and preserve the modal characteristics of the signal while ensuring the similarity between the reconstructed signal and the original signal. Then, the GPU calls an adaptive iterative solution algorithm to solve the objective function. The calculation formula for the adaptive iterative solution algorithm is as follows: By dynamically adjusting the iteration step size, the wavelet coefficients and reconstructed signal are gradually optimized, and finally the reconstructed signal of the original terahertz time-domain signal is obtained through inverse wavelet transform. Finally, the system evaluates the quality of the reconstructed signal, calculates its root mean square error and correlation coefficient with the reference signal. If the evaluation is not up to standard, the reconstruction process is automatically started to ensure that the output reconstructed signal can accurately reflect the characteristics of the original terahertz signal and meet the needs of subsequent analysis. The reconstruction results of the standard terahertz pulse show that the RMSE ≤ 1.2% and the correlation coefficient ≥ 0.98. In the actual detection of polyethylene film, the characteristic absorption peak was successfully identified, and the peak position deviation was ≤ 0.3 ps. The relative standard deviation (RSD) of the peak intensity of the same quartz sample in 30 consecutive detections was ≤ 1.8%, which proves the high accuracy and repeatability of the reconstruction and verification steps.

[0062] Detection Completion and System Shutdown: After a detection is completed, the control unit sends a stop command to each module. The laser pulse generation module stops outputting lasers, and each module stops working in sequence. Finally, the main power supply of the system is turned off. The purpose of this step is to standardize the end of the detection process, avoid damage caused by long-term idle operation of each module, and ensure that the data generated during the detection process is properly preserved, providing a guarantee for subsequent data review and analysis.

[0063] In summary, the pulsed terahertz time-domain spectral detection method using a single optical guide antenna proceeds through the following steps: system startup initialization, laser generation and transmission, terahertz signal excitation and reception, compressed sampling, signal reconstruction and verification, and system shutdown. Each step is crucial: initialization lays the foundation, laser transmission ensures the stability of the excitation source, signal excitation and reception is the core, compressed sampling efficiently acquires data, and reconstruction and verification ensure signal quality. This method is standardized and orderly, relying on the various modules of the system to achieve accurate detection and reconstruction of terahertz signals.

[0064] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A pulsed terahertz time-domain spectral system with a single optical guide antenna, characterized in that, The system includes a laser pulse generation module, a single optical guide antenna module, a compressed sampling module, a signal reconstruction module, and a control unit; The laser pulse generation module is a Ti:sapphire mode-locked laser that outputs femtosecond laser pulses and is equipped with an optical transmission system, a laser power monitoring and feedback system, and a temperature control subsystem. The single optical guide antenna module is an optical guide antenna based on a semi-insulating gallium arsenide substrate, equipped with an adjustable DC bias power supply and an optical path alignment structure. The electrodes are fabricated using a special process and covered with an insulating layer. The compressed sampling module consists of a random optical delay modulation unit and a single-pixel detector. The random optical delay modulation unit uses a MEMS micromirror array to achieve random time interval sampling, and the single-pixel detector is a liquid nitrogen-cooled mercury cadmium telluride detector. The signal reconstruction module adopts an FPGA+GPU heterogeneous computing architecture, receives data collected by the compressed sampling module, processes it, and generates a reconstructed signal. The specific steps in the signal reconstruction module for processing the compressed sampled data and generating the reconstructed signal are as follows: The FPGA performs preprocessing on the measurement vectors, including data verification, format conversion, and preliminary filtering. The FPGA transmits the preprocessed data to the GPU. The GPU calls the adaptive sparse basis construction algorithm to generate a sparse basis that adapts to the current terahertz signal. The sparse basis is used to sparsely represent the measurement vector, and the original signal is converted into a sparse wavelet coefficient vector. Based on the sparse representation results, the GPU constructs an optimization objective function that includes L1 norm regularization, data fidelity terms, and variational mode constraints. The GPU calls an adaptive iterative solution algorithm to solve the optimization objective function. By dynamically adjusting the iteration step size, it alternately updates the wavelet coefficients and the reconstructed signal to complete the iterative calculation. After obtaining the reconstructed wavelet coefficients through iterative calculation, the reconstructed signal of the original terahertz time-domain signal is obtained through inverse wavelet transform. The quality of the reconstructed signal is evaluated, and the root mean square error (RMSE) and correlation coefficient between the reconstructed signal and the reference signal are calculated. When the RMSE exceeds 2% or the correlation coefficient is lower than 0.95, the reconstructing process is triggered. The calculation formula for the adaptive sparse basis construction algorithm is as follows: ,in: The adaptive sparse basis function of the terahertz signal is related to the time offset. and time The function, This is the time offset, determined based on the dynamic changes in signal attenuation. It is a time variable that dynamically adjusts as the signal decays. It is the order index of the wavelet component, with values ​​ranging from 0 to... , This represents the total order of the wavelet components. Let be the weighting coefficients for the k-th order wavelet component. Perform multi-scale decomposition on the terahertz signal and calculate the proportion of each wavelet component in the total signal energy; this proportion is the corresponding weighting coefficient. , It is the basic wavelet function. The scaling factor is determined based on the frequency distribution of the terahertz pulse. The energy proportion of the k-th order component is determined by calculating the proportion of energy of each wavelet component k in the total energy through energy analysis of the terahertz signal. β is the attenuation adjustment coefficient, determined by an optimization algorithm based on the attenuation law of terahertz signals, with a value range of 0.1 < β < 0.

5. The formula for calculating the optimization objective function is as follows: Where: r is a sparse wavelet coefficient vector. For the h-th modal component, It is a minimum value operation used to optimize the objective function. The total order of the modal components, The value range is 1 to , For the first The adaptive L1 regularization coefficients corresponding to the first-order modal components It is the first wavelet coefficient vector The L1 norm, For modal constraint factors, Weights for data fidelity items, To measure the signal vector, It is with time Some related transformation or operator, It is a measurement matrix. For adaptive sparse basis, The square of the L2 norm of a vector, No. First-order modal components Regarding time The partial derivatives, It is the first Modal constraint factors corresponding to the first modal components; The calculation formula for the adaptive iterative solution algorithm is as follows: ; ,in: It is the wavelet coefficient vector at the nth iteration. It is the wavelet coefficient vector at the (n+1)th iteration. It is a soft threshold function. It is the step size of the nth iteration. It is the wavelet coefficient vector of the objective function J in the nth iteration. gradient at, It is the step size of the (n+1)th iteration. This is the step size adjustment coefficient, and its value range is... , It is the L2 norm of the difference between the wavelet coefficient vectors of the (n+1)th and nth iterations. It is the wavelet coefficient vector of the nth iteration. L2 norm, For the first The adaptive L1 regularization coefficients corresponding to the first modal components, where n represents the number of iterations; The control unit, based on the STM32H743 microprocessor, coordinates the timing of the entire system and communicates with each module through synchronous trigger signals.

2. The pulsed terahertz time-domain spectral system with a single optical guide antenna according to claim 1, characterized in that, The laser pulse generation module includes an optical transmission system comprising three sets of high-precision reflectors, one adjustable focus lens, and one aperture. The reflectors are coated with an 800nm ​​band anti-reflection film. The adjustable focus lens has a focal length adjustment range of 50-200mm, and the aperture of the aperture is continuously adjustable within the range of 0.1-5mm. The laser power monitoring and feedback system includes an InGaAs photodetector, a power comparison circuit, and a pump source adjustment unit. The detector has a response wavelength range of 700-1100nm. The temperature control subsystem consists of a thermocouple sensor, a PID temperature control circuit, and a semiconductor cooler. The sensor has a temperature measurement accuracy of ±0.05℃, and the temperature control circuit has an adjustment resolution of 0.01℃.

3. The pulsed terahertz time-domain spectral system with a single optical guide antenna according to claim 1, characterized in that, In the single optical guide antenna module, the optical path alignment structure includes a three-dimensional adjustable translation stage, a laser collimator, and a position sensor detector. The adjustment range of the three-dimensional adjustable translation stage is ±5mm for each of the X / Y / Z axes, and the adjustment accuracy is ≤1μm. The output wavelength of the laser collimator is consistent with the laser center wavelength of the laser pulse generation module. The spatial resolution of the position sensor detector is ≤10μm, which can provide real-time feedback on the focusing position of the laser spot on the gap between the optical guide antenna electrodes.

4. The pulsed terahertz time-domain spectral system with a single optical guide antenna according to claim 1, characterized in that, In the compressed sampling module: the random optical delay modulation unit is a MEMS micromirror array containing 1024 independently driven micromirror units, each with a size of 10μm×10μm. It adopts an electrostatic driving method to achieve deflection within a range of ±10°, and the deflection response time is ≤1μs. The driving circuit of the micromirror array supports 16-bit precision deflection angle control. By adjusting the micromirror deflection angle, random time interval sampling within a range of 0-1200ps is achieved, and the time interval adjustment step is 1ps.

5. The pulsed terahertz time-domain spectral system with a single optical guide antenna according to claim 1, characterized in that, In the compressed sampling module, the single-pixel detector's mercury cadmium telluride chip has a size of 50μm×50μm, a cutoff wavelength ≥12μm, and its operating temperature is maintained at 77K±0.1K by a closed-loop controlled liquid nitrogen Dewar. The detector is equipped with a low-temperature preamplifier with a response bandwidth covering 0.1-3THz, and the output signal is transmitted to the subsequent processing module after being impedance matched by 50Ω.

6. A pulsed terahertz time-domain spectral detection method using a single optical guide antenna, wherein the detection method employs the pulsed terahertz time-domain spectral system of the single optical guide antenna as described in any one of claims 1-5, characterized in that, The specific steps of this detection method are as follows: Laser pulse generation and stabilization control: When the laser pulse generation module is working, the Ti:sapphire mode-locked laser outputs femtosecond laser pulses. The optical transmission system transmits, focuses, and adjusts the laser pulses. The laser power monitoring and feedback system monitors the laser power in real time and performs power stabilization control through the pump source adjustment unit. The temperature control subsystem maintains the laser's operating temperature stably. Terahertz signal generation: The laser pulse enters the single optical guide antenna module, and the optical path alignment structure precisely focuses the laser spot on the gap between the electrodes of the optical guide antenna. The adjustable DC bias power supply provides bias voltage to the optical guide antenna, and the optical guide antenna generates a terahertz signal under the action of laser pulse irradiation and bias voltage. Terahertz signal sampling and detection: The compressed sampling module starts working, the random optical delay modulation unit realizes random time interval sampling of the terahertz signal based on the MEMS micromirror array, and the single pixel detector detects the sampled terahertz signal and transmits the detection signal to the signal reconstruction module. Signal reconstruction and quality assessment: The signal reconstruction module receives the probe signal, the FPGA preprocesses the measurement vector, and then transmits it to the GPU; the GPU calls the adaptive sparse basis construction algorithm to generate a sparse basis, and performs sparse representation of the measurement vector to obtain the wavelet coefficient vector; then it constructs the optimization objective function, calls the adaptive iterative solution algorithm to solve it, obtains the reconstructed signal through wavelet inverse transform, and evaluates the quality of the reconstructed signal. If the requirements are not met, it is reconstructed. System timing coordination control: The control unit coordinates the working timing of each module in the entire system through synchronous trigger signals to ensure that each module works in coordination.

Images