Method for weak signal collection and noise suppression of light receiving assembly

Abstract

The present application relates to the field of optical communication technology, especially to a weak signal collection and noise suppression method of an optical receiving assembly, comprising: monitoring the amplitude state of an analog-to-digital conversion output sequence to identify a quiet interval, constructing a two-dimensional quiet section sampling matrix and performing a time-frequency transform to obtain a power spectrum estimation sequence, and determining a demarcation frequency; extracting power spectrum data points below the demarcation frequency for linear fitting, establishing a power spectrum density model of low-frequency correlated noise and integrating to obtain an equivalent power contribution, adding the equivalent power of a thermal noise white component to obtain a total noise equivalent power, and calculating a comprehensive noise effective value; extracting a peak amplitude sequence of each frame in the matrix, fitting a generalized extreme value distribution in combination with the comprehensive noise effective value; performing a point-by-point comparison determination of a starting boundary based on a detection threshold, and triggering a backtracking correction output decision result based on a threshold jump variable. The present application can accurately characterize the noise spectrum structure and adaptively derive an effective detection threshold.

Description

Technical Field

[0001] This invention relates to the field of optical communication technology, and in particular to a method for weak signal acquisition and noise suppression in optical receiving components. Background Technology

[0002] In the field of access networks and sensing optical communication, optical receiving components bear the core responsibility of converting extremely weak optical signals into decisionable electrical signals. With the evolution of PON networks towards XGS-PON, the receiver sensitivity specification on the ONU side has been compressed to... The order of magnitude corresponds to approximately [amount] of the input fiber optical power. OLT / ONU equipment deployed in the field must be in to It operates stably over a wide temperature range. In the above scenario, avalanche photodiodes, due to their internal multiplication gain, can amplify weak photocurrents to a processable level, making them a key front-end device in the receiving link. Their back-end transimpedance amplifier is responsible for converting current signals into voltage signals and sending them to the subsequent digital processing layer.

[0003] The common practice in existing digital processing layers is to extract a fixed-length time-domain window of sampling points from the ADC digitization sequence during the silent interval of the uplink burst frame, and then calculate the mean square error of this batch of samples. As the noise basis estimate, then take ( Using an empirical coefficient as the signal detection threshold, a single comparison decision is then performed on each sampling point to distinguish between valid weak light signals and noise false alarms. This approach has good stability in a laboratory at room temperature and is relatively simple to implement in engineering.

[0004] Patent document CN117978291B discloses a low-noise optical communication module for optical communication technology. It includes an optical receiving section, a signal gain control section, and a signal adaptive adjustment section. The optical receiving section defines the ambient noise power spectral density, establishes a photoelectric detection response using a nonlinear differential equation, and employs a noise suppression model to suppress background noise. The signal gain control section dynamically adjusts the signal gain based on the established photoelectric detection response, and uses the dynamically adjusted signal gain to process the received signal that has suppressed background noise, obtaining a gain-adjusted received signal. The signal adaptive adjustment section obtains the demodulated signal through multi-symbol maximum likelihood detection and calls an adaptive forward error correction coding model to perform error correction coding on the decoded signal. By defining the ambient noise power spectral density and combining wavelet transform to post-process and suppress mixed noise in the received signal, it improves the quality of the received signal to a certain extent. However, this type of scheme still focuses on filtering the noise components in the received signal at the algorithm level, and does not propose corresponding acquisition and parameter adaptation strategies for the physical mechanism of the fundamental change in the noise spectrum structure of the core front-end device of the optical receiving component (such as avalanche photodiode) under harsh conditions such as high temperature.

[0005] When the device operates in a high-temperature environment for an extended period, the lattice defect density in the avalanche photodiode's multiplication region increases significantly with increasing temperature. This fundamentally alters the spectral structure of its dark current noise, resulting in a correlation component in the low-frequency band where the power spectral density continuously increases as the frequency decreases. This type of noise. This low-frequency correlated component exhibits a slowly varying, non-stationary baseline drift in the time domain, and its statistical characteristics are distinctly different from white Gaussian noise. This is because the data is calculated based on a single time-domain window... It can only reflect the local statistical mean square error within the window and cannot capture the peak excitation amplitude caused by low-frequency components over a longer time scale. The peak range of noise often significantly exceeds Range, but single window The estimated numerical value is almost equivalent to that of white noise at room temperature. This phenomenon causes the detection threshold set accordingly to be systematically too low, failing to cover the true peak excitation range of noise under high-temperature conditions. This results in... Before the arrival of extremely weak signal frames, the low-frequency excitation peaks of the noise itself have repeatedly triggered the threshold comparator, generating a large number of false alarms, making it impossible to reliably distinguish the true boundary of the weak light signal frame from the noise background. Summary of the Invention

[0006] To address the existing technical problems of systematic inaccuracy in time-domain single-window variance estimation and the inability of weak signal detection threshold to cover the true peak excitation of noise caused by low-frequency correlation noise of dark current in avalanche photodiode optical receiving components under high-temperature conditions, this invention provides a method for weak signal acquisition and noise suppression of optical receiving components.

[0007] This invention provides a method for weak signal acquisition and noise suppression in an optical receiving component, employing the following technical solution:

[0008] S1. Monitor the amplitude state of the analog-to-digital conversion output sequence to identify the silent interval between uplink burst frames. Collect samples within the silent interval and accumulate them across multiple frames to construct a two-dimensional silent segment sampling matrix. Perform time-frequency transformation on the matrix to obtain a power spectrum estimation sequence. Determine the boundary frequency between the white noise segment and the low-frequency correlation component segment based on the power spectrum estimation sequence. S2. Extract power spectrum data points below the boundary frequency and perform linear fitting to obtain the slope and intercept of the line. Based on the slope and intercept of the line, establish a power spectral density model of low-frequency correlated noise and perform integration to obtain the equivalent power contribution of low-frequency correlated noise. Add the equivalent power contribution to the equivalent power of the thermal noise white component to obtain the total noise equivalent power. Calculate the comprehensive noise effective value based on the total noise equivalent power. S3. Extract the sequence of peak amplitudes of each frame in the two-dimensional silent segment sampling matrix, and fit a generalized extreme value distribution to the sequence with the comprehensive noise effective value to obtain multiple parameters. With the target false alarm rate as a constraint, solve the detection threshold from the fitted generalized extreme value distribution. S4. Perform point-by-point comparison on the analog-to-digital conversion output sequence of the current burst frame using the detection threshold to determine the starting boundary of the effective weak signal frame, maintain the frame-by-frame update difference sequence of the detection threshold in previous iterations and calculate historical statistical features. When the threshold jump variable deviates from the historical statistical features, trigger backtracking correction, and re-execute the decision on the cached data using the currently updated detection threshold to output the corrected decision result.

[0009] By using a cross-frame spectral decomposition path to accurately separate low-frequency correlated noise components from the thermal noise background, the effective value of the comprehensive noise truly reflects the noise excitation amplitude under the current operating conditions, thus solving the problem that low-frequency noise is missed by time-domain statistical methods under high-temperature conditions.

[0010] Preferably, monitoring the amplitude state of the analog-to-digital conversion output sequence to identify the silence interval between uplink burst frames includes: calculating the mean and standard deviation of the absolute values ​​of all analog-to-digital conversion samples within the preceding valid frame; calculating the signal-to-noise ratio (SNR) estimate of the preceding frame based on the mean and standard deviation, and determining the decision coefficient based on the SNR estimate of the preceding frame; when the instantaneous absolute value of the analog-to-digital conversion output sequence is lower than the dynamic boundary determined by the mean, decision coefficient, and standard deviation within multiple consecutive sampling clock cycles, the silence interval between uplink burst frames is finally obtained.

[0011] By dynamically calculating the signal-to-noise ratio estimate using the mean and standard deviation of the preceding valid frames, and adaptively determining the decision coefficient accordingly, and combining multiple criteria where the instantaneous absolute value of the noise is continuously lower than the dynamic boundary for multiple consecutive clock cycles, the system effectively avoids misjudging silence intervals due to noise jitter or instantaneous spikes. This improves the robustness and accuracy of silence interval identification and provides a reliable temporal positioning basis for the pure construction of the subsequent noise sampling matrix.

[0012] Preferably, the power spectrum estimation sequence is expanded in a logarithmic coordinate system to plot the power spectrum curve; the median power density of the horizontal segment of the power spectrum curve is used as the thermal noise basis; in the logarithmic coordinate system, the first significant inflection point of the slope of the power spectrum curve is detected, and the frequency position where the absolute value of the slope suddenly increases from near zero to exceed the preset slope threshold is determined as the boundary frequency.

[0013] By using the median power density of the horizontal segment in the power spectrum curve in logarithmic coordinates as the thermal noise basis and detecting the first significant inflection point of the slope to objectively determine the boundary frequency, the white noise segment and the low-frequency correlated component segment are accurately separated. This allows the boundary frequency to automatically drift and update with changes in device temperature and other operating conditions, thus providing a clear physical basis for selecting the fitting range of the subsequent low-frequency noise power spectrum model.

[0014] Preferably, a power spectral density model of low-frequency correlated noise is established based on the slope and intercept of a straight line, and an integral operation is performed to obtain the equivalent power contribution of the low-frequency correlated noise. This includes: using the product of the intercept and the negative power of the slope of the straight line to obtain the power spectral density model of the low-frequency correlated noise; determining the lowest resolvable frequency that can be covered by cross-frame acquisition; and performing a standard power function integral operation on the power spectral density model within the frequency band from the lowest resolvable frequency to the boundary frequency to finally obtain the equivalent power contribution of the low-frequency correlated noise.

[0015] By performing parametric fitting and analytical integration on the low-frequency power spectrum, the power contribution of low-frequency correlated noise in the entire correlated frequency band is fully accumulated into the effective value of the comprehensive noise, thereby correcting the systematic underestimation problem caused by the traditional time-domain variance estimation not covering the low-frequency correlated components.

[0016] Preferably, the equivalent power contribution is added to the equivalent power of the thermal noise white component to obtain the total noise equivalent power, and the comprehensive noise effective value is calculated based on the total noise equivalent power, including: obtaining the analog-to-digital conversion Nyquist bandwidth; multiplying the thermal noise floor with the analog-to-digital conversion Nyquist bandwidth to obtain the equivalent power of the thermal noise white component; summing the equivalent power contribution of the low-frequency correlated noise with the equivalent power of the thermal noise white component to obtain the total noise equivalent power; and performing a square root operation on the total noise equivalent power to finally obtain the comprehensive noise effective value.

[0017] The comprehensive noise RMS value fully reflects the total excitation amplitude of the combined effect of thermal noise and low-frequency noise under the current operating conditions. It provides an accurate and quantifiable input benchmark for deriving the detection threshold based on the real noise power, avoiding the problem of systematic omission of low-frequency contribution caused by traditional methods that rely solely on thermal noise or local time-domain variance.

[0018] Preferably, the generalized extreme value distribution is fitted to the sequence by combining the comprehensive noise effective value to obtain multiple parameters, including: using the comprehensive noise effective value as the initial scale reference quantity to set the iterative boundary conditions for fitting the extreme value distribution; performing maximum likelihood estimation on the peak amplitude of each frame in the sequence under the constraint of the iterative boundary conditions; extracting the statistical features of the extreme value distribution through maximum likelihood estimation, and finally obtaining the location parameters, scale parameters and shape parameters of the generalized extreme value distribution.

[0019] Preferably, the detection threshold is obtained from the fitted generalized extreme value distribution with the target false alarm rate as a constraint, including: obtaining the system bit error rate target and the single frame bit length; dividing the system bit error rate target by the single frame bit length to calculate the upper limit of the independent noise false alarm probability allowed for each bit slot, and using it as the target false alarm rate; substituting the target false alarm rate into the inverse function of the cumulative distribution function of the generalized extreme value distribution for analytical solution, and finally obtaining the detection threshold.

[0020] By replacing the Gaussian assumption with a generalized extreme value distribution and directly anchoring the derivation constraint of the detection threshold to the false alarm rate objectively derived from the system's bit error rate target, the value of the detection threshold always covers the true statistical tail of the noise peak, thus avoiding a large number of false alarms caused by underestimation of the tail under high-temperature conditions.

[0021] Preferably, the starting boundary of a valid weak signal frame is determined by performing a point-by-point comparison of the analog-to-digital conversion output sequence of the current burst frame with a detection threshold, including: obtaining the expected length of the shortest consecutive valid code string of the received code pattern, and determining the minimum number of consecutive sampling points based on the expected length; comparing the amplitude of each sampling point in the analog-to-digital conversion output sequence of the current burst frame with the detection threshold one by one; and finally forming the starting boundary of a valid weak signal frame when a segment in the sequence is detected in which the amplitude of a continuous number of sampling points not less than the minimum number of consecutive sampling points exceeds the detection threshold.

[0022] Preferably, maintaining the frame-by-frame update difference sequence of the detection threshold for each detection and calculating historical statistical features includes: constructing a sliding window of a preset length; recording the frame-by-frame update difference of the detection threshold within the sliding window to form a frame-by-frame update difference sequence; performing statistical operations on each difference value in the frame-by-frame update difference sequence to finally obtain historical statistical features including the mean and standard deviation of the difference sequence.

[0023] By constructing a sliding window of a preset length, the frame-by-frame update difference of the detection threshold is recorded within the window to form a difference sequence. Then, statistical operations are performed on this sequence to obtain quantified historical statistical features including the mean and standard deviation. This transforms the dynamic change process of the detection threshold into a measurable statistical distribution model. This statistical feature can objectively reflect the amplitude range and central tendency of the normal fluctuation of the threshold under the current working conditions, providing a numerical judgment benchmark for subsequent judgment on whether the current threshold jump variable belongs to the normal range. This gives the adaptive backtracking triggering mechanism a clear mathematical basis and avoids the problem of false triggering or missed triggering caused by relying on human experience to set fixed thresholds.

[0024] Preferably, when the current threshold jump variable deviates from historical statistical characteristics, backtracking correction is triggered. The current updated detection threshold is used to re-execute the decision on the cached data to output the corrected decision result. This includes: calculating the sum of the mean of the difference sequence and the standard deviation of the difference sequence at a preset multiple, as the jump judgment threshold; determining whether the absolute value of the current threshold jump variable is greater than the jump judgment threshold; if it is greater, it is determined that a noise state change has occurred in the current working condition, thereby triggering backtracking correction, extracting the original sampled data of the preceding frame retained in the cache, and re-executing the comparison decision on the original sampled data using the current updated detection threshold, and finally obtaining the corrected decision result.

[0025] The present invention has the following technical effects: Since the silence criterion boundary, dividing frequency, spectral slope, extreme value distribution parameters, false alarm rate constraints, and backtracking trigger criteria in the scheme are all objectively derived from measured data, these parameters can automatically drift and update with changes in temperature and signal power level, achieving adaptive tracking of the detection threshold as operating conditions change. For the boundary frame decision problem, an adaptive trigger criterion based on historical jump distribution is used for backtracking correction to ensure the accuracy of boundary frame decisions. Backtracking processing is limited to frame boundaries where sudden changes in operating conditions occur, effectively controlling processing overhead and making the entire weak signal acquisition and noise suppression link feasible for continuous and stable operation under wide temperature conditions. Attached Figure Description

[0026] Figure 1 This is a flowchart of the weak signal acquisition and noise suppression method of the optical receiving component of the present invention.

[0027] Figure 2 This is a comparison chart of target detection probabilities over a wide temperature range in an embodiment of the present invention. Detailed Implementation

[0028] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0029] This invention discloses a method for weak signal acquisition and noise suppression in an optical receiving component, referring to... Figure 1 This includes the following steps: S1. Identify the silent interval and construct a sampling matrix to determine the boundary frequency.

[0030] In response to the start of the burst frame reception process, the amplitude status of the ADC output sequence is continuously monitored through frame delimitation logic to identify the silence interval between uplink burst frames. Specifically, this is based on the mean of the absolute values ​​of all ADC samples within the preceding valid frame. Standard deviation of the absolute value of samples within the same frame The signal-to-noise ratio estimate of the preceding frame was calculated. And take the decision coefficient The signal-to-noise ratio estimate of the preceding frame The ratio to a preset scaling factor. For example, the preset scaling factor could be 2, i.e. When continuous The instantaneous absolute value of the ADC output sequence within each sampling clock cycle is lower than that calculated from the mean. Decision coefficient and standard deviation Determined dynamic boundary (i.e.) When ), determine that the current period has entered a silent interval.

[0031] In response to meeting the silence criterion, all ADC samples are collected within the current silence interval, and across consecutive intervals... Frame accumulation constructs a two-dimensional silent segment sampling matrix The matrix rows correspond to the frame number, and the columns correspond to the intra-frame sampling positions. The lower limit needs to be determined by considering both the frame rate and the required minimum frequency resolution: if the frame rate is... The target minimum frequency resolution is Then it must satisfy For example, when the frame rate Pick Target minimum frequency resolution Pick hour, The lower limit is 100 frames.

[0032] By analyzing the matrix Perform a discrete Fourier transform along the time axis to obtain the power spectrum estimation sequence for each frequency component. Since the silent segment does not contain any effective optical signal, This fully reflects the spectral characteristics of the superposition of dark current noise from the avalanche photodiode and thermal noise from the transimpedance amplifier under current operating conditions. The power spectrum is expanded in logarithmic coordinates and... For the horizontal axis, Plot the power spectrum curve on the vertical axis to separate the two types of noise: thermal noise is approximately flat across the entire frequency range, appearing as a horizontal baseline; The correlation components exhibit a significant negative slope trend in the low-frequency range. The median power density in the horizontal frequency band is used as the thermal noise basis. This serves as a white noise reference for subsequent steps. In logarithmic coordinates, the white noise segment is determined by detecting the first significant inflection point of the power spectrum slope. The dividing frequency of the segment The specific criterion is as follows: In a logarithmic coordinate system, the first significant inflection point of the slope of the power spectrum curve is detected, and the frequency position where the absolute value of the slope suddenly increases from near zero to exceed the slope threshold pre-calibrated based on the device's inherent noise characteristics is determined as the boundary frequency. For example, the slope threshold mentioned above can be taken as... .

[0033] It should be noted that there are many other ways to determine the boundary frequency in this field, and those skilled in the art can choose different judgment criteria according to actual needs. This boundary frequency... It carries physical information about the current operating state of the avalanche photodiode, and its value is positively correlated with the device temperature, reflecting the physical state at high temperatures. The noise-dominant region is broadened and passed to subsequent steps to determine the frequency range for spectral linear fitting.

[0034] Thus, through cross Frame accumulation constructs a sampling matrix and performs time-frequency transformation to convert the dark current of the avalanche photodiode. The relevant components are separated from the thermal noise background at the frequency domain level, avoiding the inaccuracy of the basis estimation caused by the inability of the single-window time-domain statistical method to distinguish the two types of noise spectrum structures. This establishes a frequency domain data foundation for subsequent spectrum slope extraction and equivalent power calculation.

[0035] S2. Calculate the total noise equivalent power and the comprehensive noise effective value.

[0036] Extract the determined boundary frequency The following are power spectrum data points in the low-frequency band, and a least-squares linear fit is performed on these data points in a logarithmic coordinate system, i.e., fitting... right The linear relationship is used to extract the slope of the line. With intercept .in, The attenuation rate represents the power spectral density per decade, and its absolute value is positively correlated with the device temperature; intercept. for The extrapolated power spectrum value at point , with dimensions of .

[0037] Establish based on the fitted parameters Noise power spectral density model:

[0038] in, express The power spectral density function of noise, This represents the intercept of the linear fitting of the power spectrum. Represents frequency variables. This represents the slope of the power spectrum linear fit.

[0039] By applying the above formula in the frequency band Perform integration on the above to obtain Equivalent power contribution of noise in this frequency band :

[0040] right Performing a standard power function integral, and expanding it by substituting the upper and lower limits, yields:

[0041] in, The lowest resolvable frequency that can be covered by cross-frame acquisition is taken as the threshold. , This is the duration of a single frame. Due to the presence of thermally excited dark current within the avalanche photodiode device... The measured value is usually not equal to 1, and the above integral is physically bounded.

[0042] Based on the white component of thermal noise in the effective bandwidth equivalent power within ,Will and Adding them together, we obtain the total noise equivalent power under the current operating conditions:

[0043] in, For ADC Nyquist bandwidth, The power spectral density of the thermal noise floor, determined in the previous steps, is used to calculate the effective value of the overall noise. .because exist Overlay Equivalent power contribution, calculated under high-temperature conditions. The value will be correspondingly larger, thus quantitatively reflecting the additional contribution of low-frequency correlated noise to the peak excitation.

[0044] Thus, by performing parametric fitting and analytical integration on the low-frequency power spectrum, The power contribution of noise across the entire relevant frequency band is fully accumulated into the effective value of the comprehensive noise, thereby correcting the systematic underestimation caused by the failure to cover low-frequency correlation components in traditional time-domain variance estimation. This provides an accurate input for deriving the adaptive detection threshold based on the actual noise power.

[0045] S3. Fit a generalized extreme value distribution to solve for the detection threshold.

[0046] Extract cross The sequence of peak amplitudes of each frame in the silent segment sampling The overall effective value of noise As an initial scale reference, iterative boundary conditions are set for fitting the extreme value distribution. Under the constraints of the iterative boundary conditions, maximum likelihood estimation is performed on the peak amplitude of each frame in the sequence. The statistical features of the extreme value distribution are extracted through maximum likelihood estimation, and finally the location parameters are obtained. Scale parameters With shape parameters Because traditional methods use fixed coefficients. Multiply Obtaining the threshold implicitly assumes that the noise follows a white Gaussian distribution, while... Under significant noise conditions, the peak distribution tail of the noise time-domain sequence is thicker than that of the Gaussian distribution. Using the same... The multiple systematically underestimates the peak excitation probability; therefore, a generalized extremum distribution is introduced to replace the Gaussian assumption. The shape parameter of the generalized extremum distribution... This directly reflects the thickness of the tail section. An increase in the value indicates a thicker tail at the peak noise level under the current operating conditions. The actual statistical characteristics of the noise are consistent with those under high-temperature conditions. The increase is precisely The direct statistical manifestation of noise enhancement.

[0047] Target false alarm rate To constrain the detection threshold, it is solved from the fitted generalized extreme value distribution. , so that:

[0048] in, This represents a random variable composed of the peak amplitudes of each frame in the two-dimensional silent segment sampling matrix. Indicates the detection threshold. This indicates the probability that the peak amplitude of a single frame of sampled data exceeds a set detection threshold in a completely silent noise segment with no signal.

[0049] Target false alarm rate The system BER target is objectively determined. For the XGS-PON scenario, the system BER target is... The frame length of a single frame is If there are 10 bits, then the upper limit of the independent noise false alarm probability allowed for each bit slot is:

[0050] The relationship is based on the following: within a single frame, if the independent false alarm probability of each bit slot is... The cumulative misjudgment probability of the entire frame does not exceed To satisfy the system's BER target constraint. and All of these are known parameters of the system.

[0051] Will Substitute the inverse function of the generalized extreme cumulative distribution function to solve analytically. This derivation process is driven by the measured peak sequence of the current frame and automatically drifts and updates with temperature changes. Detection threshold. The dynamic threshold used for detecting valid signals in this frame is updated once per detection cycle. This update is applied when the ambient temperature rises. When noise is amplified, the peak sequence and All will increase, making the solution of the inverse function of the generalized extremum distribution... It automatically adjusts upwards to maintain the threshold above a reasonable coverage level of the noise peak excitation.

[0052] It should be noted that there are many other methods in this field that can fit heavy-tailed extreme value distributions. Those skilled in the art can choose different extreme value statistical models according to actual needs. The generalized extreme value distribution is only a preferred implementation method and does not constitute a limitation on the scope of application of this solution.

[0053] Thus, by replacing the Gaussian assumption with a generalized extreme value distribution, and by directly anchoring the derivation constraint of the detection threshold to the false alarm rate objectively derived from the system's BER target, the following can be achieved: The value always covers the true statistical tail of the noise peak, avoiding a large number of false alarms caused by underestimation of the tail under high temperature conditions, and realizing adaptive tracking of the detection threshold as the operating conditions change.

[0054] S4. Determine the valid starting boundary and trigger backtracking correction.

[0055] With dynamic threshold Perform point-by-point comparisons on the ADC sampling sequence of the current burst frame, in response to the occurrence of consecutive frames of at least [number missing] frames in the sequence. The amplitude at each sampling point continuously exceeded The segment is determined as the starting boundary of a valid weak signal frame. Minimum number of consecutive sampling points. Determined by the expected length of the shortest consecutive valid code string in the received code pattern, for the XGS-PON PRBS code pattern, the shortest consecutive "1" code string is statistically no shorter than 2 symbols. The value is no less than the number of sampling points corresponding to two sampling clock cycles. By introducing a continuous exceedance criterion, sporadic noise spikes are distinguished from real signal code strings in terms of their sustained characteristics in the time domain, thus avoiding misjudgment of frame start triggered by isolated noise pulses.

[0056] During the transition period of dynamic threshold updates, there is a jump between the old and new thresholds. For this boundary frame, the system retains all the original ADC sampling data from the previous frame and uses the newly calculated threshold... The decision is re-executed on the cached data to output a corrected decision result, which overwrites the previously output coarse decision result for the boundary frames. The cache depth is one frame group.

[0057] To control processing overhead, an adaptive triggering criterion is used to trigger backtracking correction. This is achieved by maintaining a criterion of length [missing information]. A sliding window to record each time Frame-by-frame update difference sequence And calculate the mean of the difference sequence. with standard deviation This process yields historical statistical features including the mean and standard deviation of the difference sequence. The sum of the mean of the difference sequence and the standard deviation of the difference sequence at a preset multiple is calculated and used as the threshold for determining a jump.

[0058] For example, the preset multiple mentioned above can be 2. Determine the threshold jump variable for this time. Whether it is greater than the transition threshold, in response to satisfying: Triggering backtracking correction, where , Let be the change in the detection threshold of the i-th frame relative to the previous frame. This represents the dynamic detection threshold calculated based on the noise sampling data of the i-th frame. Indicates the i-th The dynamic detection threshold used in frame 1; when the jump variable deviates significantly from the historical normal jump range, it indicates that a substantial noise state change has occurred in the current operating condition, and the processor will re-determine the boundary frame; if the jump variable is within the historical normal fluctuation range, the backtracking process is skipped.

[0059] Thus, by combining the continuous exceedance criterion with the adaptive backtracking triggering mechanism based on historical jump distribution, the accuracy of boundary frame decision is guaranteed while the backtracking correction process is strictly limited to the frame boundary where a real change in operating conditions occurs. This effectively controls the computational overhead of the digital processing layer, making the entire weak signal acquisition and noise suppression link feasible for continuous and stable operation under wide temperature conditions.

[0060] The technical effects of this invention can also be illustrated in conjunction with the accompanying drawings. Figure 2 This is a comparison chart of target detection probabilities over a wide temperature range in this embodiment of the invention. The horizontal axis represents ambient temperature (°C), and the vertical axis represents detection probability. The chart is presented in a clustered bar chart format. Linearly filled bars represent existing methods, while solid bars represent the method of this invention. As can be observed from the chart, the target detection probability of this invention is significantly higher than that of existing methods at all temperature points, and remains at a high level across the entire temperature range, effectively overcoming the problem that the detection probability of existing methods decreases significantly with increasing temperature.

[0061] The above are all preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Therefore, all equivalent changes made in accordance with the structure, shape and principle of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A method for weak signal acquisition and noise suppression in an optical receiving component, characterized in that, The steps include: S1. Monitoring the amplitude state of the analog-to-digital conversion output sequence to identify the silent interval between uplink burst frames. Samples are collected within the silent interval and accumulated across multiple frames to construct a two-dimensional silent segment sampling matrix. A time-frequency transformation is performed on this matrix to obtain a power spectrum estimation sequence. Based on the power spectrum estimation sequence, the boundary frequency between the white noise segment and the low-frequency correlated component segment is determined. S2. Power spectrum data points below the boundary frequency are extracted and linearly fitted to obtain the slope and intercept of the line. Based on the slope and intercept, a power spectral density model of the low-frequency correlated noise is established and integral operations are performed to obtain the equivalent power contribution of the low-frequency correlated noise. The equivalent power contribution is added to the equivalent power of the thermal noise white component to obtain the total noise equivalent power. The comprehensive noise effective value is calculated based on the total noise equivalent power; S3, extract the sequence of peak amplitudes of each frame in the two-dimensional silent segment sampling matrix, and fit the generalized extreme value distribution of the sequence with the comprehensive noise effective value to obtain multiple parameters. With the target false alarm rate as a constraint, the detection threshold is solved from the fitted generalized extreme value distribution; S4, perform point-by-point comparison on the analog-to-digital conversion output sequence of the current burst frame with the detection threshold to determine the starting boundary of the effective weak signal frame, maintain the frame-by-frame update difference sequence of the detection threshold in previous times and calculate the historical statistical features. When the threshold jump variable deviates from the historical statistical features, backtracking correction is triggered. The current updated detection threshold is used to re-execute the decision on the cached data to output the corrected decision result.

2. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, Monitoring the amplitude status of the analog-to-digital conversion output sequence to identify the silent interval between uplink burst frames includes: calculating the mean and standard deviation of the absolute values ​​of all analog-to-digital conversion samples within the preceding valid frames; The signal-to-noise ratio (SNR) estimate of the preceding frame is calculated based on the mean and standard deviation, and the decision coefficient is determined based on the SNR estimate of the preceding frame. When the instantaneous absolute value of the analog-to-digital conversion output sequence is lower than the dynamic boundary determined by the mean, decision coefficient, and standard deviation within multiple consecutive sampling clock cycles, the silence interval between uplink burst frames is finally obtained.

3. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, Determining the boundary frequency between the white noise segment and the low-frequency correlated component segment based on the power spectrum estimation sequence includes: expanding the power spectrum estimation sequence in a logarithmic coordinate system and plotting the power spectrum curve; using the median power density of the horizontal segment of the power spectrum curve as the thermal noise basis; detecting the first significant inflection point of the slope of the power spectrum curve in a logarithmic coordinate system, and determining the frequency position where the absolute value of the slope suddenly increases from near zero to exceed the preset slope threshold as the boundary frequency.

4. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, A power spectral density model of low-frequency correlated noise is established based on the slope and intercept of a straight line, and an integral operation is performed to obtain the equivalent power contribution of the low-frequency correlated noise. This includes: using the product of the intercept and the negative power of the slope of the straight line to obtain the power spectral density model of the low-frequency correlated noise; determining the lowest resolvable frequency that can be covered by cross-frame acquisition; and performing a standard power function integral operation on the power spectral density model within the frequency band from the lowest resolvable frequency to the boundary frequency to finally obtain the equivalent power contribution of the low-frequency correlated noise.

5. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, The equivalent power contribution is added to the equivalent power of the thermal noise white component to obtain the total noise equivalent power, and the comprehensive noise effective value is calculated based on the total noise equivalent power. This includes: obtaining the analog-to-digital conversion Nyquist bandwidth; multiplying the thermal noise floor with the analog-to-digital conversion Nyquist bandwidth to obtain the equivalent power of the thermal noise white component; summing the equivalent power contribution of the low-frequency correlated noise with the equivalent power of the thermal noise white component to obtain the total noise equivalent power; and performing a square root operation on the total noise equivalent power to finally obtain the comprehensive noise effective value.

6. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, By combining the effective value of the comprehensive noise to fit the generalized extreme value distribution of the sequence, multiple parameters are obtained, including: using the effective value of the comprehensive noise as the initial scale reference to set the iterative boundary conditions for fitting the extreme value distribution; performing maximum likelihood estimation on the peak amplitude of each frame in the sequence under the constraint of the iterative boundary conditions; extracting the statistical features of the extreme value distribution through maximum likelihood estimation, and finally obtaining the location parameters, scale parameters and shape parameters of the generalized extreme value distribution.

7. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, Using the target false alarm rate as a constraint, the detection threshold is solved from the fitted generalized extreme value distribution, including: obtaining the system bit error rate target and the single frame bit length; dividing the system bit error rate target by the single frame bit length to calculate the upper limit of the independent noise false alarm probability allowed for each bit slot, which is used as the target false alarm rate; substituting the target false alarm rate into the inverse function of the cumulative distribution function of the generalized extreme value distribution for analytical solution, and finally obtaining the detection threshold.

8. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, The starting boundary of a valid weak signal frame is determined by performing point-by-point comparisons on the analog-to-digital conversion output sequence of the current burst frame using a detection threshold. This includes: obtaining the expected length of the shortest consecutive valid code string of the received code pattern, and determining the minimum number of consecutive sampling points based on the expected length; comparing the amplitude of each sampling point in the analog-to-digital conversion output sequence of the current burst frame with the detection threshold one by one; and finally forming the starting boundary of a valid weak signal frame when a segment in the sequence is detected where the amplitude of a continuous number of sampling points or less exceeds the detection threshold.

9. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, Maintaining the frame-by-frame update difference sequence of detection thresholds for each detection and calculating historical statistical features includes: constructing a sliding window of a preset length; recording the frame-by-frame update difference of detection thresholds within the sliding window to form a frame-by-frame update difference sequence; performing statistical operations on each difference value in the frame-by-frame update difference sequence to finally obtain historical statistical features including the mean and standard deviation of the difference sequence.

10. The method for weak signal acquisition and noise suppression of the optical receiving component according to claim 1, characterized in that, When the current threshold jump variable deviates from historical statistical characteristics, backtracking correction is triggered. The current updated detection threshold is used to re-execute the decision on the cached data to output the corrected decision result. This includes: calculating the sum of the mean of the difference sequence and the standard deviation of the difference sequence at a preset multiple, as the jump judgment threshold; determining whether the absolute value of the current threshold jump variable is greater than the jump judgment threshold; if it is greater, it is determined that a noise state change has occurred in the current working condition, thereby triggering backtracking correction, extracting the original sampled data of the previous frame retained in the cache, and re-executing the comparison decision on the original sampled data using the current updated detection threshold, and finally obtaining the corrected decision result.

Images