Memory signal integrity enhancement system based on multi-dimension parameter adaptation

Abstract

The application discloses a memory signal integrity enhancement system based on multi-dimension parameter self-adaption and belongs to the technical field of signal transmission.The application comprises a multi-dimension signal acquisition module, a multi-dimension parameter analysis module, a self-adaptive adjustment module, a memory signal optimization module and a feedback monitoring module.The multi-dimension signal acquisition module combines multi-dimension parameter analysis, acquires data bus, address bus and control bus signals synchronously, adopts differentiated optimization means according to a dynamic adjustment strategy generated by a model and forms a customized optimization system with one parameter and one scheme, so that the optimized target signal is optimal in amplitude stability, timing accuracy, anti-noise capability and impedance matching degree.The feedback monitoring module combines the multi-dimension parameter analysis module, carries out full-dimension index monitoring and dynamic calibration and guarantees the continuous adaptability of the system to the memory running state and the long-term stability of signal integrity.

Description

Technical Field

[0001] This invention relates to the field of signal transmission technology, and in particular to a memory signal integrity enhancement system based on multi-dimensional parameter adaptation. Background Technology

[0002] Currently, issues such as signal amplitude attenuation, timing jitter, noise interference, and impedance mismatch during memory operation have become core bottlenecks restricting further performance release.

[0003] Traditional memory signal processing solutions often use fixed parameter adjustment modes, ignoring the coupling relationship between amplitude, timing, noise, and impedance. On the other hand, traditional solutions do not consider the differences in dynamic working scenarios of memory, and fixed adjustment parameters are difficult to adapt to the signal requirements of different scenarios. This leads to problems such as signal overshoot and increased bit error rate in high-load scenarios, while causing resource waste in low-load scenarios.

[0004] Meanwhile, existing technologies for analyzing memory signal parameters often rely on subjective experience to judge the weight of key parameters, lacking scientific quantitative basis. Most signal processing systems adopt an open-loop adjustment architecture, meaning that after signal optimization, no real-time feedback calibration mechanism is established, making it impossible to dynamically track changes in signal characteristics during memory operation. When memory hardware ages, external electromagnetic interference increases, or workload changes suddenly, existing adjustment strategies are difficult to adapt quickly, leading to a gradual decline in signal integrity and subsequently causing serious failures such as system blue screens and data loss. Summary of the Invention

[0005] The purpose of this invention is to provide a memory signal integrity enhancement system based on multi-dimensional parameter adaptation to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a memory signal integrity enhancement system based on multi-dimensional parameter adaptation, comprising:

[0007] The multi-dimensional signal acquisition module is configured to acquire raw signal data in real time during the operation of the memory. The raw signal data includes signal amplitude, signal timing, signal noise, and signal impedance data.

[0008] A multidimensional parameter analysis module is communicatively connected to the multidimensional signal acquisition module and is configured to receive the raw signal data, perform multidimensional parameter analysis on the raw signal data, and determine the key parameters affecting the integrity of the memory signal and the weight ratio of each key parameter.

[0009] An adaptive adjustment module is communicatively connected to the multidimensional parameter analysis module and is configured to construct a parameter adaptive adjustment model based on the key parameters and the weight ratio of each key parameter, and generate a dynamic adjustment strategy for memory signals.

[0010] The memory signal optimization module is communicatively connected to the adaptive adjustment module and is configured to optimize the original memory signal according to the dynamic adjustment strategy and output the target signal with enhanced integrity.

[0011] The feedback monitoring module is communicatively connected to the memory signal optimization module and the multidimensional parameter analysis module, respectively, and is used to monitor the integrity index of the target signal in real time, and feed the integrity index back to the multidimensional parameter analysis module to realize the dynamic calibration of the parameter adaptive adjustment model.

[0012] Furthermore, the process by which the multidimensional signal acquisition module acquires raw signal data includes:

[0013] The multi-dimensional signal acquisition module has a built-in high-precision signal sampling unit, and the sampling frequency of the high-precision signal sampling unit is not less than 8 times the highest operating frequency of the memory.

[0014] The high-precision signal sampling unit synchronously samples the signals on the memory data bus, address bus, and control bus at preset time intervals to obtain initial sampling data.

[0015] The initial sampled data is preprocessed to obtain the original signal data;

[0016] Add a timestamp and signal source identifier to the original signal data.

[0017] Furthermore, the process by which the multidimensional parameter analysis module performs multidimensional parameter analysis on the original signal data includes:

[0018] The original signal data is acquired, and the original signal data is classified into amplitude data, time series data, noise data, and impedance data.

[0019] For the amplitude-type data, calculate the signal amplitude fluctuation range, amplitude attenuation rate, and amplitude consistency deviation to determine the key amplitude-type parameters;

[0020] For the aforementioned timing data, analyze the signal rise time, fall time, signal delay, and timing jitter value to determine key timing parameters;

[0021] For the aforementioned noise data, the signal noise amplitude, noise frequency distribution, and noise duration are detected to determine key noise parameters;

[0022] For the aforementioned impedance data, measure the impedance change and impedance matching degree during signal transmission to determine key impedance parameters;

[0023] A parameter weight evaluation model is constructed based on the analytic hierarchy process (AHP). The key parameters of amplitude, time series, noise, and impedance are input, and the weight ratio of each key parameter is calculated.

[0024] Furthermore, when determining the key parameters of the amplitude class, all signal amplitude sample values ​​in the amplitude class data are extracted, and the signal amplitude fluctuation range is calculated as the first amplitude parameter;

[0025] Based on timestamp information, amplitude data is divided into several consecutive data segments in chronological order. The amplitude change rate of adjacent data segments is calculated, and the maximum value among all change rates is taken as the second amplitude parameter.

[0026] In the statistical amplitude data, the deviation of all sampled values ​​from the standard amplitude value is calculated, the deviation rate of each sampled value is calculated, and the average of all deviation rates is taken as the third amplitude parameter.

[0027] By integrating the first amplitude parameter, the second amplitude parameter, and the third amplitude parameter and removing outlier parameters, the key amplitude parameters are obtained.

[0028] When determining the key parameters of the timing class, the rising edge of the signal in the timing class data is identified, the rising edge trigger threshold is set to 10%-90% of the standard high level, the rising edge time interval from 10% high level to 90% high level for each rising edge is recorded, and the average value of all rising edge time intervals is taken as the first timing parameter.

[0029] Set the falling edge trigger threshold to 90%-10% of the standard high level, record the falling edge time interval from 90% high level to 10% high level for each falling edge, and take the average of all falling edge time intervals as the second timing parameter;

[0030] Using the memory clock signal as a reference, calculate the time difference between each data signal and the clock signal, and take the absolute value of all time differences as the third timing parameter;

[0031] The delay value of the same data signal within consecutive clock cycles is statistically analyzed, and the standard deviation of the delay value is calculated as the fourth timing parameter.

[0032] By integrating the first, second, third, and fourth timing parameters and removing outlier parameters, key timing parameters are obtained.

[0033] Furthermore, when determining the key parameters of the noise class, the pure noise signal is separated from the noise class data, the maximum amplitude and minimum amplitude of the pure noise signal are calculated, and the larger of the absolute values ​​of the maximum amplitude and minimum amplitude is taken as the first noise parameter.

[0034] Perform a Fourier transform on the pure noise signal to obtain the frequency spectrum of the noise. Statistically identify the frequency intervals in the spectrum whose energy accounts for more than 5% of the total energy. Record the start and end values ​​of the frequency intervals as the second noise parameter.

[0035] Based on timestamp information, identify time periods where the noise amplitude exceeds a preset noise threshold, calculate the duration of each time period, and take the sum of all durations as the third noise parameter.

[0036] By integrating the first noise parameter, the second noise parameter, and the third noise parameter and removing outliers, key noise parameters are obtained.

[0037] When determining the key parameters of the impedance class, all impedance sample values ​​are extracted from the impedance class data, the difference between adjacent sample values ​​is calculated, and the maximum and minimum values ​​among all differences are taken as the first impedance parameter.

[0038] Set the standard impedance value for memory, calculate the matching degree between each impedance sample value and the standard impedance value, and take the average value of all matching degrees as the second impedance parameter;

[0039] The stability of the first and second impedance parameters is tested. If the fluctuation range of the impedance change value is less than the preset fluctuation threshold and the change amplitude of the impedance matching degree is less than 5% within 100 consecutive sampling periods, it is confirmed as a key parameter of effective impedance.

[0040] Furthermore, the multidimensional parameter analysis module uses the analytic hierarchy process (AHP) to calculate the weight proportions of key parameters, specifically including:

[0041] A hierarchical model is constructed, which includes a target layer, a criterion layer, and a scheme layer. The target layer determines the weight ratio of key parameters. The criterion layer includes amplitude-type parameters, time-series parameters, noise-type parameters, and impedance-type parameters. The scheme layer includes amplitude-type key parameters, time-series key parameters, noise-type key parameters, and impedance-type key parameters.

[0042] A judgment matrix is ​​constructed by comparing each parameter in the criterion layer and each key parameter in the scheme layer pairwise.

[0043] Calculate the maximum eigenvalue and corresponding eigenvector of the judgment matrix, normalize the eigenvector, and obtain the initial weights of each parameter.

[0044] Furthermore, the adaptive adjustment module constructs a parameter adaptive adjustment model, specifically including:

[0045] Based on the weight ratio of each key parameter, the parameter adjustment priority is determined. Key parameters with a weight ratio higher than a preset threshold are designated as high-priority adjustment parameters, and the rest are designated as low-priority adjustment parameters.

[0046] For the high-priority adjustment parameters, a parameter-signal integrity mapping relationship library is established, which stores the signal integrity compliance status corresponding to different parameter values;

[0047] Based on the mapping relationship library, a parameter adaptive adjustment model is constructed using the BP neural network algorithm, with the real-time values ​​of high-priority adjustment parameters as the model input and the signal integrity optimization target as the model output.

[0048] The parameter adaptive adjustment model is trained, and low-priority adjustment parameters are used as model constraints and incorporated into the parameter adaptive adjustment model to generate the dynamic adjustment strategy.

[0049] Furthermore, the memory signal optimization module optimizes the original signal according to a dynamic adjustment strategy, specifically including:

[0050] The dynamic adjustment strategy is analyzed to determine the specific adjustment methods and adjustment thresholds for different key parameters. The adjustment methods include at least amplitude compensation, timing calibration, noise suppression, and impedance matching.

[0051] For amplitude-related key parameters, the built-in amplitude compensation circuit adjusts the gain of the original signal to stabilize the signal amplitude within the standard amplitude range. For timing-related key parameters, the timing calibration unit adjusts the signal transmission delay to reduce timing jitter to below the preset jitter threshold. For noise-related key parameters, the noise suppression module is activated, and an adaptive filtering algorithm is used to filter signal noise and reduce noise interference with signal integrity. For impedance-related key parameters, the impedance matching network adjusts the signal transmission impedance to achieve the preset impedance matching standard.

[0052] After adjusting all key parameters, the optimized signal is integrated and the target signal is output.

[0053] Furthermore, the feedback monitoring module monitors the target signal integrity index, specifically including:

[0054] A preset signal integrity evaluation index system is established, which includes at least standard indicators for signal eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability.

[0055] The feedback monitoring module collects the target signal in real time and extracts the indicator data of the target signal. The indicator data includes eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability data.

[0056] Compare the extracted indicator data with the standard indicators and calculate the indicator deviation value;

[0057] If the deviation value of the indicator is less than or equal to the allowable deviation threshold, the target signal integrity is determined to meet the standard, and a compliance feedback message is generated; if the deviation value of the indicator is greater than the allowable deviation threshold, the target signal integrity is determined to fail to meet the standard, a failure feedback message is generated, and the indicator items that exceed the standard range are marked.

[0058] The compliance or non-compliance feedback information and the marked indicator items are transmitted to the multidimensional parameter analysis module, which adjusts the weight ratio of key parameters and the parameter adaptive adjustment model based on the feedback information.

[0059] Furthermore, the signal eye diagram parameters in the feedback monitoring module include eye diagram opening, eye diagram crossover voltage, eye diagram extinction ratio, and eye diagram jitter; the process of the feedback monitoring module extracting eye diagram parameters specifically includes:

[0060] Eye diagrams are plotted on the target signal to obtain real-time eye diagrams;

[0061] The maximum opening region of the eye diagram is located by image recognition, and the vertical distance of this region is measured as the eye diagram opening degree; the crossover point of the high and low level signals in the eye diagram is determined, and the voltage value corresponding to the crossover point is measured as the eye diagram crossover point voltage; the ratio of the average power of the high level signal to the average power of the low level signal in the eye diagram is calculated to obtain the eye diagram extinction ratio; based on the jitter trajectory of the eye diagram, the peak-to-peak value of the jitter is counted as the eye diagram jitter value;

[0062] The extracted eye diagram opening, crossover voltage, extinction ratio, and jitter value are compared with preset standard eye diagram parameters to complete the eye diagram parameter evaluation.

[0063] Compared with the prior art, the beneficial effects of the present invention are:

[0064] 1. The multi-dimensional signal acquisition module of this invention combines multi-dimensional parameter analysis. By synchronously acquiring data bus, address bus, and control bus signals, it avoids the signal detail loss and data partiality problems caused by traditional single bus or low-frequency sampling. The multi-dimensional parameter analysis module further classifies the raw data into four categories: amplitude, timing, noise, and impedance. For each category of data, a hierarchical parameter extraction logic is designed, and the weight of each key parameter is quantified by combining the analytic hierarchy process. This not only solves the problems of partiality and subjectivity in traditional parameter extraction, but also provides an objective priority basis for subsequent adjustment, ensuring that the analysis results can comprehensively and accurately reflect the core factors affecting the integrity of memory signals.

[0065] 2. The adaptive adjustment module of this invention is dynamically executed in conjunction with the memory signal optimization module. Based on the weight ratio of key parameters, high and low priority adjustment parameters are divided. For high priority parameters, a BP neural network adaptive adjustment model is constructed. Combined with the parameter-signal integrity mapping relationship library, the nonlinear correlation between parameters and integrity targets is accurately captured, avoiding the defects of traditional average adjustment or linear models that cannot adapt to dynamic signals. Based on the dynamic adjustment strategy generated by the model, differentiated optimization methods are adopted for different parameter types, forming a customized optimization system with one solution for each parameter. This solves the problems of poor performance and poor parameter coordination of traditional general adjustment, ensuring that the optimized target signal achieves optimal performance in amplitude stability, timing accuracy, noise resistance, and impedance matching, meeting the signal requirements of high-frequency and high-bandwidth operation of memory.

[0066] 3. The feedback monitoring module of this invention, combined with the multi-dimensional parameter analysis module, ensures the system's continuous adaptability to memory operating status and the long-term stability of signal integrity through full-dimensional indicator monitoring and dynamic calibration. By acquiring target signals in real time and calculating indicator deviation values, it quickly identifies abnormal fluctuations in the adjusted signals, avoiding memory operation risks caused by lagging monitoring. Based on the deviation results, it generates differentiated feedback information and marks abnormal indicator items, providing a precise adjustment basis for the multi-dimensional parameter analysis module. This enables the module to promptly correct the weight ratio of key parameters and the adaptive adjustment model, ensuring that the system can continuously respond to the signal requirements of the memory operating scenario, maintain memory signal integrity at the optimal level in the long term, and ensure the stable and efficient operation of the memory and even the entire computer system. Attached Figure Description

[0067] Figure 1 This is a schematic diagram of the working process of the memory signal integrity enhancement system of the present invention. Detailed Implementation

[0068] 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 embodiments of the present invention, and not all embodiments. 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.

[0069] Please see Figure 1 The present invention provides the following technical solutions:

[0070] A memory signal integrity enhancement system based on multi-dimensional parameter adaptation includes:

[0071] The multi-dimensional signal acquisition module is configured to acquire raw signal data in real time during the operation of the memory. The raw signal data includes signal amplitude, signal timing, signal noise, and signal impedance data.

[0072] A multidimensional parameter analysis module is communicatively connected to the multidimensional signal acquisition module and is configured to receive the raw signal data, perform multidimensional parameter analysis on the raw signal data, and determine the key parameters affecting the integrity of the memory signal and the weight ratio of each key parameter.

[0073] An adaptive adjustment module is communicatively connected to the multidimensional parameter analysis module and is configured to construct a parameter adaptive adjustment model based on the key parameters and the weight ratio of each key parameter, and generate a dynamic adjustment strategy for memory signals.

[0074] The memory signal optimization module is communicatively connected to the adaptive adjustment module and is configured to optimize the original memory signal according to the dynamic adjustment strategy and output the target signal with enhanced integrity.

[0075] The feedback monitoring module is communicatively connected to the memory signal optimization module and the multidimensional parameter analysis module, respectively, and is used to monitor the integrity index of the target signal in real time, and feed the integrity index back to the multidimensional parameter analysis module to realize the dynamic calibration of the parameter adaptive adjustment model.

[0076] The process of the multidimensional signal acquisition module acquiring raw signal data includes:

[0077] The multi-dimensional signal acquisition module has a built-in high-precision signal sampling unit, and the sampling frequency of the high-precision signal sampling unit is not less than 8 times the highest operating frequency of the memory.

[0078] The high-precision signal sampling unit synchronously samples the signals on the memory data bus, address bus, and control bus at preset time intervals to obtain initial sampling data.

[0079] The initial sampled data is preprocessed to obtain the original signal data;

[0080] Add a timestamp and signal source identifier to the original signal data.

[0081] In the above embodiments, the multi-dimensional signal acquisition module acquires raw signal data. By setting a high-precision sampling unit with a sampling frequency no less than 8 times the highest operating frequency of the memory, it can completely capture the transient changes of signals during high-frequency operation of the memory. It synchronously acquires the memory data bus, address bus, and control bus signals, avoiding the parameter bias caused by single bus sampling. It ensures that the raw signal data can fully reflect the overall operating status of the memory. Through preprocessing, adding timestamps and source identifiers, it effectively removes interference noise in the initial sampled data. At the same time, it realizes accurate traceability and time sequence correlation of signal data, greatly reducing the computational complexity of subsequent multi-dimensional parameter analysis and improving the overall data processing efficiency of the system.

[0082] The multidimensional parameter analysis module performs multidimensional parameter analysis on the raw signal data, including:

[0083] The original signal data is acquired, and the original signal data is classified into amplitude data, time series data, noise data, and impedance data.

[0084] For the amplitude-type data, calculate the signal amplitude fluctuation range, amplitude attenuation rate, and amplitude consistency deviation to determine the key amplitude-type parameters;

[0085] For the aforementioned timing data, analyze the signal rise time, fall time, signal delay, and timing jitter value to determine key timing parameters;

[0086] For the aforementioned noise data, the signal noise amplitude, noise frequency distribution, and noise duration are detected to determine key noise parameters;

[0087] For the aforementioned impedance data, measure the impedance change and impedance matching degree during signal transmission to determine key impedance parameters;

[0088] A parameter weight evaluation model is constructed based on the analytic hierarchy process (AHP). The key parameters of amplitude, time series, noise, and impedance are input, and the weight ratio of each key parameter is calculated.

[0089] Specifically, when determining the key parameters of the amplitude class, all signal amplitude sample values ​​in the amplitude class data are extracted, and the signal amplitude fluctuation range is calculated as the first amplitude parameter.

[0090] Based on timestamp information, amplitude data is divided into several consecutive data segments in chronological order. The amplitude change rate of adjacent data segments is calculated, and the maximum value among all change rates is taken as the second amplitude parameter.

[0091] In the statistical amplitude data, the deviation of all sampled values ​​from the standard amplitude value is calculated, the deviation rate of each sampled value is calculated, and the average of all deviation rates is taken as the third amplitude parameter.

[0092] By integrating the first amplitude parameter, the second amplitude parameter, and the third amplitude parameter and removing outlier parameters, the key amplitude parameters are obtained.

[0093] When determining the key parameters of the timing class, the rising edge of the signal in the timing class data is identified, the rising edge trigger threshold is set to 10%-90% of the standard high level, the rising edge time interval from 10% high level to 90% high level for each rising edge is recorded, and the average value of all rising edge time intervals is taken as the first timing parameter.

[0094] Set the falling edge trigger threshold to 90%-10% of the standard high level, record the falling edge time interval from 90% high level to 10% high level for each falling edge, and take the average of all falling edge time intervals as the second timing parameter;

[0095] Using the memory clock signal as a reference, calculate the time difference between each data signal and the clock signal, and take the absolute value of all time differences as the third timing parameter;

[0096] The delay value of the same data signal within consecutive clock cycles is statistically analyzed, and the standard deviation of the delay value is calculated as the fourth timing parameter.

[0097] By integrating the first, second, third, and fourth timing parameters and removing outlier parameters, key timing parameters are obtained.

[0098] When determining the key parameters of the noise class, the pure noise signal is separated from the noise class data, the maximum amplitude and minimum amplitude of the pure noise signal are calculated, and the larger of the absolute values ​​of the maximum amplitude and minimum amplitude is taken as the first noise parameter.

[0099] Perform a Fourier transform on the pure noise signal to obtain the frequency spectrum of the noise. Statistically identify the frequency intervals in the spectrum whose energy accounts for more than 5% of the total energy. Record the start and end values ​​of the frequency intervals as the second noise parameter.

[0100] Based on timestamp information, identify time periods where the noise amplitude exceeds a preset noise threshold, calculate the duration of each time period, and take the sum of all durations as the third noise parameter.

[0101] By integrating the first noise parameter, the second noise parameter, and the third noise parameter and removing outliers, key noise parameters are obtained.

[0102] When determining the key parameters of the impedance class, all impedance sample values ​​are extracted from the impedance class data, the difference between adjacent sample values ​​is calculated, and the maximum and minimum values ​​among all differences are taken as the first impedance parameter.

[0103] Set the standard impedance value for memory, calculate the matching degree between each impedance sample value and the standard impedance value, and take the average value of all matching degrees as the second impedance parameter;

[0104] The stability of the first and second impedance parameters is tested. If the fluctuation range of the impedance change value is less than the preset fluctuation threshold and the change amplitude of the impedance matching degree is less than 5% within 100 consecutive sampling periods, it is confirmed as a key parameter of effective impedance.

[0105] In the above embodiments, the multi-dimensional parameter analysis module creatively achieves precise and systematic extraction of key parameters of memory signals. For amplitude data, through the hierarchical calculation of the first to third amplitude parameters, it covers both the static amplitude range of the signal and captures the dynamic amplitude attenuation trend and consistency deviation, solving the problem of parameter one-sidedness caused by focusing only on a single amplitude value in traditional methods. In the analysis of timing parameters, it covers rising edge, falling edge, delay, and timing jitter in multiple dimensions, accurately locating the core causes of timing anomalies. For noise parameters, it separates pure noise signals and combines frequency and time dimension analysis to achieve a panoramic presentation of noise characteristics. The stability verification design of impedance parameters effectively filters instantaneous impedance fluctuation interference, ensuring the reliability of key parameters. The application of the analytic hierarchy process scientifically quantifies the weight of each key parameter, avoids subjective experience judgment bias, and provides an objective and accurate priority basis for subsequent adaptive adjustment.

[0106] The multidimensional parameter analysis module uses the analytic hierarchy process (AHP) to calculate the weight proportions of key parameters, specifically including:

[0107] A hierarchical model is constructed, which includes a target layer, a criterion layer, and a scheme layer. The target layer determines the weight ratio of key parameters. The criterion layer includes amplitude-type parameters, time-series parameters, noise-type parameters, and impedance-type parameters. The scheme layer includes amplitude-type key parameters, time-series key parameters, noise-type key parameters, and impedance-type key parameters.

[0108] A judgment matrix is ​​constructed by comparing each parameter in the criterion layer and each key parameter in the scheme layer pairwise.

[0109] Calculate the maximum eigenvalue and corresponding eigenvector of the judgment matrix, normalize the eigenvector, and obtain the initial weights of each parameter.

[0110] In the above embodiments, the multidimensional parameter analysis module uses the analytic hierarchy process (AHP) to calculate the weight ratio. By constructing a three-layer structure model, the abstract, definite weight target is decomposed into a quantifiable and comparable hierarchical relationship, solving the problem of ambiguity in the relationship between the target and parameters in traditional weight calculation. Based on pairwise comparisons, a judgment matrix is ​​constructed to quantify the relative importance between parameters, avoiding weight bias caused by single-dimensional evaluation. Through the calculation of the maximum eigenvalue and eigenvector and normalization processing, qualitative comparison is transformed into quantitative weight, ensuring the rationality and reliability of the final weight ratio.

[0111] The adaptive adjustment module constructs a parameter adaptive adjustment model, specifically including:

[0112] Based on the weight ratio of each key parameter, the parameter adjustment priority is determined. Key parameters with a weight ratio higher than a preset threshold are designated as high-priority adjustment parameters, and the rest are designated as low-priority adjustment parameters.

[0113] For the high-priority adjustment parameters, a parameter-signal integrity mapping relationship library is established, which stores the signal integrity compliance status corresponding to different parameter values;

[0114] Based on the mapping relationship library, a parameter adaptive adjustment model is constructed using the BP neural network algorithm, with the real-time values ​​of high-priority adjustment parameters as the model input and the signal integrity optimization target as the model output.

[0115] The parameter adaptive adjustment model is trained, and low-priority adjustment parameters are used as model constraints and incorporated into the parameter adaptive adjustment model to generate the dynamic adjustment strategy.

[0116] In the above embodiments, the adaptive adjustment module constructs a parameter adaptive adjustment model, divides the adjustment priority based on the weight ratio, and can prioritize the key parameters that have the greatest impact on signal integrity, avoiding the waste of resources and inefficiency caused by average adjustment. The establishment of the parameter-signal integrity mapping relationship library establishes a direct correlation between parameter values ​​and integrity effects, providing rich training basis for the model. The BP neural network algorithm is used to construct the model, and its powerful nonlinear fitting ability is used to accurately capture the complex mapping relationship between high-priority parameters and integrity targets. Low-priority parameters are incorporated into the model as constraints, taking into account both the comprehensiveness and priority of adjustment. The generated dynamic adjustment strategy can accurately solve the core problem and avoid interference from secondary parameters, significantly improving the effectiveness of memory signal adjustment.

[0117] The memory signal optimization module optimizes the original signal based on a dynamic adjustment strategy, specifically including:

[0118] The dynamic adjustment strategy is analyzed to determine the specific adjustment methods and adjustment thresholds for different key parameters. The adjustment methods include at least amplitude compensation, timing calibration, noise suppression, and impedance matching.

[0119] For amplitude-related key parameters, the built-in amplitude compensation circuit adjusts the gain of the original signal to stabilize the signal amplitude within the standard amplitude range. For timing-related key parameters, the timing calibration unit adjusts the signal transmission delay to reduce timing jitter to below the preset jitter threshold. For noise-related key parameters, the noise suppression module is activated, and an adaptive filtering algorithm is used to filter signal noise and reduce noise interference with signal integrity. For impedance-related key parameters, the impedance matching network adjusts the signal transmission impedance to achieve the preset impedance matching standard.

[0120] After adjusting all key parameters, the optimized signal is integrated and the target signal is output.

[0121] In the above embodiments, the memory signal optimization module optimizes the original signal according to a dynamic adjustment strategy. By analyzing the adjustment strategy, it clarifies the adjustment method and threshold of each parameter, achieving customized optimization and avoiding the poor effect caused by traditional general adjustment. Differentiated optimization methods are adopted for different types of key parameters, and the amplitude compensation circuit ensures the stability of the signal amplitude, forming a full-dimensional optimization system covering amplitude, timing, noise, and impedance. This comprehensively solves the core interference factors of memory signal integrity. The integrated design of the optimized signal avoids the signal coordination problem caused by single parameter optimization, ensuring that the target signal is well adapted among parameters in all dimensions, significantly improving the stability, anti-interference ability, and transmission reliability of the memory signal, and meeting the signal requirements of high-frequency and high-bandwidth memory operation.

[0122] The feedback monitoring module monitors target signal integrity indicators, specifically including:

[0123] A preset signal integrity evaluation index system is established, which includes at least standard indicators for signal eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability.

[0124] The feedback monitoring module collects the target signal in real time and extracts the indicator data of the target signal. The indicator data includes eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability data.

[0125] Compare the extracted indicator data with the standard indicators and calculate the indicator deviation value;

[0126] If the deviation value of the indicator is less than or equal to the allowable deviation threshold, the target signal integrity is determined to meet the standard, and a compliance feedback message is generated; if the deviation value of the indicator is greater than the allowable deviation threshold, the target signal integrity is determined to fail to meet the standard, a failure feedback message is generated, and the indicator items that exceed the standard range are marked.

[0127] The compliance or non-compliance feedback information and the marked indicator items are transmitted to the multidimensional parameter analysis module, which adjusts the weight ratio of key parameters and the parameter adaptive adjustment model based on the feedback information.

[0128] The signal eye diagram parameters in the feedback monitoring module include eye diagram opening, eye diagram crossover voltage, eye diagram extinction ratio, and eye diagram jitter; the process of extracting eye diagram parameters by the feedback monitoring module specifically includes:

[0129] Eye diagrams are plotted on the target signal to obtain real-time eye diagrams;

[0130] The maximum opening region of the eye diagram is located by image recognition, and the vertical distance of this region is measured as the eye diagram opening degree; the crossover point of the high and low level signals in the eye diagram is determined, and the voltage value corresponding to the crossover point is measured as the eye diagram crossover point voltage; the ratio of the average power of the high level signal to the average power of the low level signal in the eye diagram is calculated to obtain the eye diagram extinction ratio; based on the jitter trajectory of the eye diagram, the peak-to-peak value of the jitter is counted as the eye diagram jitter value;

[0131] The extracted eye diagram opening, crossover voltage, extinction ratio, and jitter value are compared with preset standard eye diagram parameters to complete the eye diagram parameter evaluation.

[0132] In the above embodiments, the feedback monitoring module, by pre-setting a multi-dimensional evaluation index system including eye diagram parameters, bit error rate, etc., overcomes the limitations of single index monitoring, and can comprehensively and objectively evaluate the integrity of the target signal. By collecting the target signal in real time and calculating the index deviation value, it realizes the dynamic tracking of signal integrity, can quickly identify abnormal fluctuations in the adjusted signal, avoid memory operation risks caused by lagging monitoring, generate differentiated feedback information based on the deviation results and mark abnormal index items, providing accurate adjustment basis for the multi-dimensional parameter analysis module, so that the weight ratio of key parameters and the adaptive adjustment model can adapt to signal changes in a timely manner, improve the system's adaptability to the dynamic operating state of memory, and ensure that the memory signal integrity is maintained at the optimal level in the long term.

[0133] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A memory signal integrity enhancement system based on multi-dimensional parameter adaptive design, characterized in that, include: The multi-dimensional signal acquisition module is configured to acquire raw signal data in real time during the operation of the memory. The raw signal data includes signal amplitude, signal timing, signal noise, and signal impedance data. The multidimensional parameter analysis module is communicatively connected to the multidimensional signal acquisition module and is configured to receive the raw signal data, perform multidimensional parameter analysis on the raw signal data, and determine the key parameters affecting the integrity of the memory signal and the weight ratio of each key parameter. Specifically, the multidimensional parameter analysis classifies the data from four dimensions: amplitude, timing, noise, and impedance, and extracts the key parameters corresponding to each category of data. An adaptive adjustment module is communicatively connected to the multidimensional parameter analysis module and is configured to construct a parameter adaptive adjustment model based on the key parameters and the weight ratio of each key parameter, and generate a dynamic adjustment strategy for memory signals. The memory signal optimization module is communicatively connected to the adaptive adjustment module and is configured to optimize the original memory signal according to the dynamic adjustment strategy and output the target signal with enhanced integrity. The feedback monitoring module is communicatively connected to both the memory signal optimization module and the multidimensional parameter analysis module. It is used to monitor the integrity index of the target signal in real time and feed the integrity index back to the multidimensional parameter analysis module to achieve dynamic calibration of the parameter adaptive adjustment model.

2. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 1, characterized in that, The process by which the multidimensional signal acquisition module acquires raw signal data includes: The multi-dimensional signal acquisition module has a built-in high-precision signal sampling unit, and the sampling frequency of the high-precision signal sampling unit is not less than 8 times the highest operating frequency of the memory. The high-precision signal sampling unit synchronously samples the signals on the memory data bus, address bus, and control bus at preset time intervals to obtain initial sampling data. The initial sampled data is preprocessed to obtain the original signal data; Add a timestamp and signal source identifier to the original signal data.

3. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 1, characterized in that, The process by which the multidimensional parameter analysis module performs multidimensional parameter analysis on the original signal data includes: The original signal data is acquired, and the original signal data is classified into amplitude data, time series data, noise data, and impedance data. For the amplitude-type data, calculate the signal amplitude fluctuation range, amplitude attenuation rate, and amplitude consistency deviation to determine the key amplitude-type parameters; For the aforementioned timing data, analyze the signal rise time, fall time, signal delay, and timing jitter value to determine key timing parameters; For the aforementioned noise data, the signal noise amplitude, noise frequency distribution, and noise duration are detected to determine key noise parameters; For the aforementioned impedance data, measure the impedance change and impedance matching degree during signal transmission to determine key impedance parameters; A parameter weight evaluation model is constructed based on the analytic hierarchy process (AHP). The key parameters of amplitude, time series, noise, and impedance are input, and the weight ratio of each key parameter is calculated.

4. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 3, characterized in that, When determining the key parameters of the amplitude class, all signal amplitude sample values ​​in the amplitude class data are extracted, and the signal amplitude fluctuation range is calculated as the first amplitude parameter. Based on timestamp information, amplitude data is divided into several consecutive data segments in chronological order. The amplitude change rate of adjacent data segments is calculated, and the maximum value among all change rates is taken as the second amplitude parameter. In the statistical amplitude data, the deviation of all sampled values ​​from the standard amplitude value is calculated, the deviation rate of each sampled value is calculated, and the average of all deviation rates is taken as the third amplitude parameter. By integrating the first amplitude parameter, the second amplitude parameter, and the third amplitude parameter and removing outlier parameters, the key amplitude parameters are obtained. When determining the key parameters of the timing class, the rising edge of the signal in the timing class data is identified, the rising edge trigger threshold is set to 10%-90% of the standard high level, the rising edge time interval from 10% high level to 90% high level for each rising edge is recorded, and the average value of all rising edge time intervals is taken as the first timing parameter. Set the falling edge trigger threshold to 90%-10% of the standard high level, record the falling edge time interval from 90% high level to 10% high level for each falling edge, and take the average of all falling edge time intervals as the second timing parameter; Using the memory clock signal as a reference, calculate the time difference between each data signal and the clock signal, and take the absolute value of all time differences as the third timing parameter; The delay value of the same data signal within consecutive clock cycles is statistically analyzed, and the standard deviation of the delay value is calculated as the fourth timing parameter. By integrating the first, second, third, and fourth timing parameters and removing outlier parameters, key timing parameters are obtained.

5. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 3, characterized in that, When determining the key parameters of the noise class, the pure noise signal is separated from the noise class data, the maximum amplitude and minimum amplitude of the pure noise signal are calculated, and the larger of the absolute values ​​of the maximum amplitude and minimum amplitude is taken as the first noise parameter. Perform a Fourier transform on the pure noise signal to obtain the frequency spectrum of the noise. Statistically identify the frequency intervals in the spectrum whose energy accounts for more than 5% of the total energy. Record the start and end values ​​of the frequency intervals as the second noise parameter. Based on timestamp information, identify time periods where the noise amplitude exceeds a preset noise threshold, calculate the duration of each time period, and take the sum of all durations as the third noise parameter. By integrating the first noise parameter, the second noise parameter, and the third noise parameter and removing outliers, key noise parameters are obtained. When determining the key parameters of the impedance class, all impedance sample values ​​are extracted from the impedance class data, the difference between adjacent sample values ​​is calculated, and the maximum and minimum values ​​among all differences are taken as the first impedance parameter. Set the standard impedance value for memory, calculate the matching degree between each impedance sample value and the standard impedance value, and take the average value of all matching degrees as the second impedance parameter; The stability of the first and second impedance parameters is tested. If the fluctuation range of the impedance change value is less than the preset fluctuation threshold and the change amplitude of the impedance matching degree is less than 5% within 100 consecutive sampling periods, it is confirmed as a key parameter of effective impedance.

6. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 3, characterized in that, The multidimensional parameter analysis module uses the analytic hierarchy process (AHP) to calculate the weight proportions of key parameters, specifically including: A hierarchical model is constructed, which includes a target layer, a criterion layer, and a scheme layer. The target layer determines the weight ratio of key parameters. The criterion layer includes amplitude-type parameters, time-series parameters, noise-type parameters, and impedance-type parameters. The scheme layer includes amplitude-type key parameters, time-series key parameters, noise-type key parameters, and impedance-type key parameters. A judgment matrix is ​​constructed by comparing each parameter in the criterion layer and each key parameter in the scheme layer pairwise. Calculate the maximum eigenvalue and corresponding eigenvector of the judgment matrix, normalize the eigenvector, and obtain the initial weights of each parameter.

7. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 1, characterized in that, The adaptive adjustment module constructs a parameter adaptive adjustment model, specifically including: Based on the weight ratio of each key parameter, the parameter adjustment priority is determined. Key parameters with a weight ratio higher than a preset threshold are designated as high-priority adjustment parameters, and the rest are designated as low-priority adjustment parameters. For the high-priority adjustment parameters, a parameter-signal integrity mapping relationship library is established, which stores the signal integrity compliance status corresponding to different parameter values; Based on the mapping relationship library, a parameter adaptive adjustment model is constructed using the BP neural network algorithm, with the real-time values ​​of high-priority adjustment parameters as the model input and the signal integrity optimization target as the model output. The parameter adaptive adjustment model is trained, and low-priority adjustment parameters are used as model constraints and incorporated into the parameter adaptive adjustment model to generate the dynamic adjustment strategy.

8. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 1, characterized in that, The memory signal optimization module optimizes the original signal according to a dynamic adjustment strategy, specifically including: The dynamic adjustment strategy is analyzed to determine the specific adjustment methods and adjustment thresholds for different key parameters. The adjustment methods include at least amplitude compensation, timing calibration, noise suppression, and impedance matching. For amplitude-related key parameters, the built-in amplitude compensation circuit adjusts the gain of the original signal to stabilize the signal amplitude within the standard amplitude range. For timing-related key parameters, the timing calibration unit adjusts the signal transmission delay to reduce timing jitter to below the preset jitter threshold. For noise-related key parameters, the noise suppression module is activated, and an adaptive filtering algorithm is used to filter signal noise and reduce noise interference with signal integrity. For impedance-related key parameters, the impedance matching network adjusts the signal transmission impedance to achieve the preset impedance matching standard. After adjusting all key parameters, the optimized signal is integrated and the target signal is output.

9. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 1, characterized in that, The feedback monitoring module monitors the target signal integrity indicators, specifically including: A preset signal integrity evaluation index system is established, which includes at least standard indicators for signal eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability. The feedback monitoring module collects the target signal in real time and extracts the indicator data of the target signal. The indicator data includes eye diagram parameters, bit error rate, signal overshoot amplitude, signal undershoot amplitude, and signal stability data. Compare the extracted indicator data with the standard indicators and calculate the indicator deviation value; if the indicator deviation value is less than or equal to the allowable deviation threshold, it is determined that the target signal integrity meets the standard and a compliance feedback information is generated. If the deviation value of the indicator is greater than the allowable deviation threshold, the integrity of the target signal is determined to be substandard, a non-compliance feedback message is generated, and the indicator items that exceed the standard range are marked. The compliance or non-compliance feedback information and the marked indicator items are transmitted to the multidimensional parameter analysis module, which adjusts the weight ratio of key parameters and the parameter adaptive adjustment model based on the feedback information.

10. The memory signal integrity enhancement system based on multi-dimensional parameter adaptation as described in claim 9, characterized in that, The signal eye diagram parameters in the feedback monitoring module include eye diagram opening, eye diagram crossover voltage, eye diagram extinction ratio, and eye diagram jitter; the process of extracting eye diagram parameters by the feedback monitoring module specifically includes: Eye diagrams are plotted on the target signal to obtain real-time eye diagrams; The maximum opening region of the eye diagram is located by image recognition, and the vertical distance of this region is measured as the eye diagram opening degree; the crossover point of the high and low level signals in the eye diagram is determined, and the voltage value corresponding to the crossover point is measured as the eye diagram crossover point voltage; the ratio of the average power of the high level signal to the average power of the low level signal in the eye diagram is calculated to obtain the eye diagram extinction ratio; based on the jitter trajectory of the eye diagram, the peak-to-peak value of the jitter is counted as the eye diagram jitter value; The extracted eye diagram opening, crossover voltage, extinction ratio, and jitter value are compared with preset standard eye diagram parameters to complete the eye diagram parameter evaluation.

Images