Hard disk read-write channel crosstalk offset method and system and solid state disk

Abstract

The invention relates to the field of data transmission, in particular to a hard disk read-write channel crosstalk offset method and system and a solid state disk. The method comprises the following steps: monitoring signal jump edge characteristics of adjacent read-write channels of the solid state disk in real time, and obtaining dynamic characteristic parameters of a crosstalk source; and on the basis of a pre-trained interference mode library, predicting a corresponding crosstalk noise parameter according to the signal jump edge characteristics, and establishing association between the jump edge and crosstalk. According to the hard disk read-write channel crosstalk offset method and system and the solid state disk provided by the invention, the problem that the prior art cannot adapt to crosstalk dynamic changes in different data modes is effectively solved through dynamic monitoring, accurate prediction, reverse compensation and self-adaptive optimization architectures, the change of crosstalk noise in different data modes can be dynamically adapted, and the stability of the solid state disk is improved. The integrity of a data eye pattern of a receiving end is effectively kept, the bit error rate is reduced, the transmission stability of multichannel parallel read-write of the solid state disk is improved, and support is provided for reliable application of a high-frequency and high-bandwidth SSD.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of data transmission, in particular to a hard disk read-write channel crosstalk cancellation method and system and a solid state disk. BACKGROUND

[0002] With the rapid evolution of solid state disks towards high capacity and high bandwidth, multi-channel parallel read-write architecture has become a solution to improve data transmission rate, that is, by simultaneously enabling multiple independent read-write channels, data is dispersed to different channels for parallel transmission to improve overall throughput. However, in high-frequency signal transmission scenarios, such as PCIe 4.0 and above specifications, the signal transition edge of adjacent channels will generate dynamic crosstalk noise through the power plane, ground plane or parasitic capacitance, such as the transition from "0" to "1" or "1" to "0". This crosstalk noise will superimpose on the valid signal, resulting in a decrease in "eye height" and a decrease in "eye width" of the data eye diagram at the receiving end, and even closing, directly increasing the bit error rate and seriously affecting the reliability of data transmission.

[0003] Existing crosstalk cancellation schemes mostly use fixed parameter pre-emphasis or physical shielding design. Pre-emphasis uses to enhance the amplitude of the signal transition edge to offset transmission loss, but fixed pre-emphasis parameters cannot adapt to the changes in crosstalk noise under different data modes. For example, when transmitting continuous all "0" or all "1" data, the frequency of the signal transition edge is very low, and the crosstalk noise is weak. When transmitting alternating "0" and "1" data, the frequency of the transition edge is very high, and the crosstalk noise is significantly enhanced. Fixed pre-emphasis parameters cannot effectively cancel this dynamic crosstalk. Physical shielding can reduce the coupling between adjacent channels, but it increases the complexity and cost of PCB design, and cannot completely eliminate the crosstalk caused by power / ground coupling.

[0004] Therefore, how to accurately cancel the dynamic crosstalk of the multi-channel parallel read-write of the solid state disk to adapt to the changes in crosstalk noise under different data modes, maintain the integrity of the data eye diagram at the receiving end, and improve the stability of high-frequency transmission has become a technical problem to be solved in the field. SUMMARY

[0005] The present disclosure provides a hard disk read-write channel crosstalk cancellation method, system and solid state disk, which aims to overcome at least one of the deficiencies in the prior art.

[0006] To achieve the above purpose, the technical scheme disclosed by the present application is as follows: According to one aspect of the present disclosure, a hard disk read-write channel crosstalk cancellation method is provided, comprising the steps of: real-time monitoring the signal transition edge characteristics of adjacent read-write channels of the solid state disk to obtain dynamic characteristic parameters of the crosstalk source; According to the signal jump edge feature, corresponding crosstalk noise parameters are predicted based on a pre-trained interference mode library, and a correlation between the jump edge and crosstalk is established. A compensation signal opposite to the crosstalk noise parameters is injected into a pre-emphasis waveform at the sending end to offset the crosstalk noise. An adaptive filtering algorithm is used to adjust the coefficient value of the compensation signal in real time according to the error rate feedback at the receiving end, and the compensation effect is dynamically optimized.

[0007] Further, real-time monitoring of the signal jump edge feature is achieved through a crosstalk observer, which includes a jump edge detection unit and a channel correlation unit. The jump edge detection unit detects the amplitude, frequency and time point of the signal jump edge of each channel, and the channel correlation unit correlates the jump edge signals of adjacent channels.

[0008] Further, the pre-training method of the interference mode library includes collecting signal jump edge features and corresponding crosstalk noise parameters in historical data, and using a support vector machine or neural network algorithm to establish a mapping relationship between the jump edge features and the crosstalk noise parameters.

[0009] Further, the prediction of corresponding crosstalk noise parameters according to the signal jump edge feature includes inputting the amplitude, frequency and time difference of the jump edge of adjacent channels of the signal jump edge into the pre-trained interference mode library, and outputting the amplitude and phase of the corresponding crosstalk noise.

[0010] Further, the injection of a compensation signal opposite to the crosstalk noise parameters in the pre-emphasis waveform at the sending end is achieved by adjusting the output voltage amplitude and phase of the pre-emphasis circuit, which is connected to the driving transistor at the sending end. The error rate at the receiving end is obtained through a cyclic redundancy check circuit or a Reed-Solomon code check circuit.

[0011] Further, one step of adjusting the coefficient value of the compensation signal includes calculating the adjustment amount of the compensation coefficient according to the error rate after transmitting one frame of data, and updating the coefficient value of the compensation signal using the gradient descent method.

[0012] Further, the adjacent read-write channels are two or more channels with adjacent physical positions in the solid state disk, and the crosstalk observer simultaneously monitors the signal jump edge features of the two or more channels.

[0013] Further, the interference mode library is periodically updated, the mapping relationship is optimized by using new signal jump edge features and corresponding crosstalk noise parameters, and the update period is adjusted according to the use time of the solid state disk; the adaptive filtering algorithm is a least mean square algorithm or a normalized least mean square algorithm; the slope of the pre-emphasis waveform is adjusted according to the amplitude of the crosstalk noise parameter, and the greater the amplitude of the crosstalk noise parameter, the greater the slope of the pre-emphasis waveform.

[0014] According to another aspect of the present disclosure, a hard disk read-write channel crosstalk cancellation system is provided for implementing the hard disk read-write channel crosstalk cancellation method as described above, comprising: A crosstalk observation module is configured to monitor the signal jump edge features of adjacent read-write channels of the solid state disk in real time and obtain dynamic characteristic parameters of crosstalk sources. A crosstalk prediction module is connected to the crosstalk observation module and configured to predict corresponding crosstalk noise parameters based on a pre-trained interference mode library according to the signal jump edge features. A compensation injection module is connected to the crosstalk prediction module and a pre-emphasis circuit of the sending end and configured to inject a compensation signal opposite to the crosstalk noise parameters in the output waveform of the pre-emphasis circuit. An adaptive adjustment module is connected to the compensation injection module and a bit error rate detection circuit of the receiving end and configured to adopt an adaptive filtering algorithm to adjust the coefficient value of the compensation signal in real time according to the feedback result of the bit error rate detection circuit.

[0015] According to still another aspect of the present disclosure, a solid state disk is provided, comprising: A storage array integrated with multiple read-write channels; A main control unit connected to the storage array and the hard disk read-write channel crosstalk cancellation system as described above; And a multi-channel scheduling module connected to the crosstalk observation module and the main control unit, configured to analyze the crosstalk risk of each channel combination according to the adjacent channel signal jump edge features monitored by the crosstalk observation module, adjust the task allocation of each read-write channel by the main control unit, and preferentially use the channel combination with low crosstalk risk.

[0016] The present application has the following advantages: The hard disk read-write channel crosstalk cancellation method, system and solid state disk provided by the present application effectively solve the problem that the prior art cannot adapt to the dynamic changes of crosstalk under different data modes by using the architecture of dynamic monitoring, accurate prediction, reverse compensation and adaptive optimization, and the beneficial effects mainly lie in the following aspects: Specifically, by monitoring the signal jump edge characteristics of adjacent channels in real time, such as jump amplitude, frequency and correlation time difference of adjacent channels, the application can accurately capture the dynamic changes of crosstalk sources. The jump edge is the main source of crosstalk noise, and monitoring the jump edge characteristics is equivalent to directly obtaining the source information of crosstalk, which provides accurate input for subsequent crosstalk cancellation.

[0017] Further, based on the pre-trained interference mode library to predict the crosstalk noise parameters, the application establishes the mapping relationship between the jump edge characteristics and the crosstalk noise. The interference mode library is obtained by training historical data, which can quickly match the crosstalk noise type corresponding to the current jump edge characteristics, such as amplitude and phase, avoiding the disadvantages of blind compensation in traditional schemes, and realizing accurate prediction of crosstalk.

[0018] Further, the reverse compensation signal is injected into the pre-emphasis waveform at the sending end, and the application adopts the active cancellation method to directly cancel the crosstalk noise before it is generated, that is, the pre-emphasis waveform is an important component of the sending end signal, and the injection of the reverse compensation signal is equivalent to suppressing the crosstalk at the beginning of signal transmission, which is more efficient than the traditional receiving end correction scheme.

[0019] Further, the compensation coefficient is adjusted according to the bit error rate feedback through the adaptive filtering algorithm, and the application realizes dynamic optimization of crosstalk cancellation. The bit error rate is a direct reflection of the crosstalk cancellation effect, and the adaptive adjustment can correct the compensation parameters in real time to ensure that the compensation effect remains optimal when the data mode changes, such as switching from continuous all "0" to alternating "0" and "1", avoiding the limitations of fixed parameter schemes that are adjusted once and used continuously.

[0020] Through the synergistic effect of the above technical solutions, the application can dynamically adapt to the changes of crosstalk noise under different data modes, effectively maintain the integrity of the data eye diagram at the receiving end, reduce the bit error rate, and improve the transmission stability of the multi-channel parallel read and write of the solid state disk, providing support for the reliable application of high-frequency and high-bandwidth SSD.

[0021] The above description is only a summary of the technical solutions of the application. In order to more clearly understand the technical means of the application and can be implemented according to the content of the specification, the following describes the preferred embodiments of the application with reference to the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 The flowchart of the hard disk read-write channel crosstalk cancellation method in an embodiment of the application; Figure 2 The schematic diagram of the influence of crosstalk noise on data eye diagram in an embodiment of the application; Figure 3 The schematic diagram of the multi-channel signal jump edge characteristic monitoring in an embodiment of the application; Figure 4A pre-emphasis injection reverse compensation signal diagram in an embodiment of the present application; Figure 5 A bit error rate performance comparison diagram of crosstalk cancellation in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present application will be clearly and completely described with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all the embodiments of the present application.

[0024] In the embodiments of the present application, the words such as “exemplary” or “for example” are used to mean serving as an example, instance, or illustration. Any embodiment or design solution described as “exemplary” or “for example” in the embodiments of the present application should not be interpreted as being more preferred or having more advantages than other embodiments or design solutions. In fact, the words such as “exemplary” or “for example” are used in the specific manner to present the relevant concept.

[0025] The present application provides the following preferred embodiments: Embodiment one: to solve the problem that the crosstalk noise caused by signal transition of adjacent channels in the multi-channel read-write process of a solid state disk leads to signal integrity degradation and bit error rate increase, the present embodiment provides a hard disk read-write channel crosstalk cancellation method, which realizes effective cancellation of crosstalk noise through a cooperative mechanism of real-time monitoring of crosstalk source dynamic characteristics, association of transition edge and crosstalk mode, injection of reverse compensation signal, and adaptive optimization of compensation coefficient. As shown in the figure, the flow of the hard disk read-write channel crosstalk cancellation method is as follows: Figure 1 S100: real-time monitoring of signal transition edge characteristics of adjacent read-write channels of a solid state disk, and acquisition of dynamic characteristic parameters of a crosstalk source.

[0026] S200: based on a pre-trained interference mode library, prediction of corresponding crosstalk noise parameters according to the signal transition edge characteristics, and establishment of association of transition edge and crosstalk.

[0027] S300: injection of a compensation signal reverse to the crosstalk noise parameters in a pre-emphasis waveform at a sending end, and cancellation of the crosstalk noise.

[0028] S400: adaptive filtering algorithm is adopted to real-time adjust the coefficient value of the compensation signal according to the bit error rate feedback of a receiving end, and dynamically optimize the compensation effect.

[0029] Specifically, real-time monitoring of signal transition edge characteristics of adjacent read-write channels of a solid state disk, and acquisition of dynamic characteristic parameters of a crosstalk source. As shown in the figure, the signal transition edge characteristics of the adjacent read-write channels of the solid state disk are monitored in real time, and the dynamic characteristic parameters of the crosstalk source are acquired. Figure 3 ​As shown, the output signals of adjacent channels are sampled and analyzed in real time by a high-speed transition edge detection unit provided at the transmitting end. The high-speed transition edge detection unit uses a combination of a high-speed comparator and an edge detection circuit to compare the signals of adjacent channels with a preset threshold value. When the signal amplitude exceeds the threshold value, the edge detection circuit is triggered to record the start time of the transition edge, the polarity of the rising or falling edge, and the change rate of the transition amplitude. These parameters together constitute the dynamic characteristic parameters of the crosstalk source. The start time corresponds to the generation time of the crosstalk noise, the polarity reflects the direction of the crosstalk, and the change rate of the transition amplitude is related to the strength of the crosstalk. It should be understood that the frequency of real-time monitoring needs to be strictly matched with the data transmission rate. For example, in a 10 Gbps data transmission scenario, the monitoring period needs to be less than 100 ps to ensure that the transition edge characteristics of each data bit are captured, providing accurate input for subsequent crosstalk prediction.

[0030] Further, based on the pre-trained interference pattern library, the corresponding crosstalk noise parameters are predicted according to the signal transition edge characteristics. The interference pattern library is a database established in advance through experiments and simulations, which stores the corresponding relationship model of different transition edge characteristics and crosstalk noise parameters. Specifically, different transition situations of adjacent channels are simulated in a laboratory environment, including transition time difference, amplitude difference, polarity combination, etc. The corresponding crosstalk noise frequency, phase, amplitude, etc. are measured by a high-speed oscilloscope, and these data are input into a machine learning model such as a support vector machine for training to obtain the mapping relationship between transition edge characteristics and crosstalk noise parameters. In actual application, the transition edge characteristics monitored in real time are input into the model, and the specific parameters of the crosstalk noise generated by the current adjacent channel can be predicted, including frequency, phase, and amplitude. Establishing the association between transition edge and crosstalk essentially binds the generation mechanism of crosstalk noise with the signal activity of adjacent channels, enabling crosstalk prediction to shift from passive measurement to active association, providing accurate basis for subsequent targeted compensation. As shown, Figure 4 The prediction results of the interference pattern library directly guide the generation of compensation signals, which is a key link in the crosstalk cancellation process.

[0031] Further, a compensation signal opposite to the crosstalk noise parameter is injected into the pre-emphasis waveform at the transmitting end. The pre-emphasis waveform is a signal generated by the transmitting end to compensate for the high-frequency attenuation of the channel, which improves the integrity of the signal after transmission through the channel by boosting the high-frequency components of the data signal, such as the steep part of the transition edge. The process of injecting the opposite compensation signal needs to be deeply integrated with the pre-emphasis waveform generation process: according to the predicted crosstalk noise parameter, a signal opposite to the crosstalk noise is generated by a compensation signal generation circuit, such as a digital signal processor, which has the same frequency as the crosstalk noise, the opposite phase of the crosstalk noise phase, and the same amplitude as the crosstalk noise amplitude. Then, the compensation signal is superimposed into the pre-emphasis waveform to form a transmitting signal containing compensation information. When the signal encounters crosstalk noise generated by adjacent channels during transmission, the compensation signal and the crosstalk noise cancel each other out due to opposite phase and equal amplitude, thereby reducing the impact of crosstalk on the target channel signal. It can be understood that the timing of injecting the compensation signal needs to be strictly synchronized with the generation time of the crosstalk noise, for example, within 10ps after the adjacent channel transition edge is generated, to ensure that the compensation signal and the crosstalk noise completely overlap in the time domain, maximizing the cancellation effect. As shown in Figure 4 the original transmitting signal is pre-emphasized and injected with an opposite compensation signal, the impact of crosstalk noise is effectively suppressed, and the signal integrity is significantly improved.

[0032] Further, an adaptive filtering algorithm is used to adjust the coefficient value of the compensation signal in real time according to the bit error rate feedback of the receiving end. The core of the adaptive filtering algorithm is to dynamically adjust the coefficients of the compensation signal, such as amplitude coefficient and phase coefficient, through the feedback information of the receiving end, to cope with the dynamic changes of the crosstalk noise, including the changes of crosstalk strength caused by temperature changes and the changes of crosstalk frequency caused by data pattern changes. Specifically, the receiving end calculates the bit error rate of the current data frame through a bit error rate detection circuit, such as a parity check circuit or a cyclic redundancy check circuit, and transmits the bit error rate information to the compensation coefficient adjustment module of the transmitting end through the feedback channel. The compensation coefficient adjustment module uses the least mean square algorithm to adjust the amplitude coefficient and the phase coefficient of the compensation signal by taking the bit error rate as the error signal: when the bit error rate rises, it means that the amplitude of the compensation signal is insufficient or the phase deviates, and the algorithm will increase the amplitude coefficient or adjust the phase coefficient to make the compensation signal closer to the opposite signal of the crosstalk noise; when the bit error rate is too low, it means that the compensation is excessive and may cause signal distortion, and the algorithm will reduce the amplitude coefficient to avoid excessive compensation. It should be understood that the adaptive adjustment process is a closed-loop feedback system, and its response speed needs to match the change rate of the crosstalk noise: in the scenario where the crosstalk noise changes slowly with temperature, the adjustment period can be set to seconds; in the scenario where the crosstalk noise changes rapidly with data pattern, the adjustment period needs to be set to milliseconds to ensure the real-time performance of the compensation effect. As shown in Figure 5As shown, after adjusting the compensation coefficient by using the adaptive filtering algorithm, the bit error rate performance of the system is obviously improved compared with the baseline without adaptive adjustment, which reflects the effectiveness of adaptive optimization.

[0033] The benefits of the embodiment are that the precise positioning of crosstalk noise is achieved by real-time monitoring of the dynamic characteristics of the crosstalk source, the association of the jump edge and the crosstalk is established based on the pre-trained interference pattern library, the crosstalk prediction is changed from passive measurement to active association, the reverse compensation signal is injected into the pre-emphasis waveform at the sending end to offset the crosstalk, which changes from passive suppression at the receiving end to active prevention at the sending end, the adaptive filtering algorithm is used to dynamically adjust the compensation coefficient according to the bit error rate feedback to ensure the long-term stability of the compensation effect, and the crosstalk noise of the adjacent read-write channels is effectively offset through the synergistic effect of the above steps, the signal integrity is improved, and the bit error rate is reduced.

[0034] Embodiment two: To solve the problem that the single-channel jump information is insufficiently associated with the adjacent channel when monitoring the adjacent channel signal jump edge characteristics in real time, which leads to inaccurate positioning of the crosstalk source, the structure design of the crosstalk observer is further refined in this embodiment. The crosstalk observer includes a jump edge detection unit and a channel association unit, which cooperatively achieve precise capture and association of multi-channel jump characteristics.

[0035] The jump edge detection unit is used to detect the amplitude, frequency and time point of the jump edge of each channel signal. Specifically, this unit uses a high-speed comparator to compare the channel signal with a preset threshold in real time, and triggers a jump event when the signal amplitude exceeds the threshold, while recording the starting time point of the jump edge through a high-precision counter; for jump amplitude detection, the signal voltage difference before and after the jump is captured through a sample-and-hold circuit to obtain the quantized value of the jump amplitude; and the jump frequency is extracted by counting the time interval of consecutive jump events to calculate the number of jumps per unit time. For example, in a 10Gbps data transmission scenario, the jump frequency corresponds to the conversion rate of data bits, and the detection accuracy needs to be within 1% to ensure the accuracy of subsequent crosstalk prediction.

[0036] Further, the channel association unit is used to associate the jump edge signals of adjacent channels. This unit stores the jump time point of each channel as a timestamp, and compares the time difference of jump events of adjacent channels through a timestamp matching algorithm: when the time difference of jump events of adjacent channels is less than a preset threshold such as 50ps, it is determined that there is an association between them, that is, the jump of the current channel is the source of crosstalk of the adjacent channel. It should be understood that the core of channel association is to capture the synchronicity of adjacent channel jumps, that is, the strength of crosstalk noise is directly related to the time difference of adjacent channel jumps, and the smaller the time difference, the stronger the crosstalk. For example, Figure 3 As shown, the monitoring of multi-channel signal jump edges needs to integrate scattered jump events into associated crosstalk source characteristics through the channel association unit to provide complete input information for subsequent crosstalk prediction.

[0037] The benefit of the embodiment is that, through the cooperation of the jump edge detection unit and the channel association unit, the jump edge feature of a single channel is promoted to a multi-channel association feature, ensuring the accuracy and relevance of the dynamic feature parameters of the crosstalk source, and laying a solid foundation for subsequent crosstalk noise prediction.

[0038] Embodiment three: To solve the problem of low crosstalk prediction accuracy caused by insufficient training data or improper model selection of the interference mode library, the pre-training method of the interference mode library is further optimized in this embodiment. The pre-training process combines the collection of historical data and machine learning model training to establish an accurate mapping relationship between the jump edge feature and the crosstalk noise parameter.

[0039] Specifically, the signal jump edge feature and the corresponding crosstalk noise parameter in the historical data are collected. The historical data comes from laboratory simulation and simulation: in the laboratory environment, a multi-channel solid state disk read-write test platform is built to simulate the jump scene of adjacent channels under different channel spacing, data rate, voltage level, etc. The jump edge feature (such as amplitude, frequency, time point) of the target channel and the crosstalk noise parameter (such as amplitude, phase, frequency) generated by the adjacent channel are synchronously collected by a high-speed oscilloscope; in the simulation environment, a SPICE model is used to construct an inter-channel crosstalk model, and the corresponding crosstalk noise parameter is output by inputting the simulation signal of different jump edge features. The collected data needs to cover multiple extreme scenarios, such as maximum channel spacing, highest data rate, minimum jump time difference, etc., to ensure the generalization ability of the model.

[0040] Further, support vector machines or neural network algorithms are used to establish the mapping relationship. For support vector machines, the jump edge feature is used as the input vector and the crosstalk noise parameter is used as the output vector. The input space is mapped to a high-dimensional feature space through a kernel function (such as a radial basis function), an optimal classification hyperplane is found, and linear or nonlinear mapping is realized; for neural networks, a multi-layer perceptron structure is used, the input layer is the jump edge feature, the hidden layer processes the nonlinear relationship through an activation function (such as ReLU), and the output layer is the crosstalk noise parameter. The weights are adjusted through the back propagation algorithm to minimize the prediction error. It should be understood that support vector machines are suitable for small sample data scenarios and can effectively avoid overfitting; neural networks are suitable for large sample and strong nonlinear scenarios and can capture more complex mapping relationships. The choice of the two algorithms needs to be determined according to the size and complexity of the historical data.

[0041] The benefit of the embodiment is that, through the systematic collection of historical data and the targeted selection of machine learning algorithms, an accurate mapping relationship between the jump edge feature and the crosstalk noise parameter is established, the prediction accuracy and generalization ability of the interference mode library are improved, and reliable model support is provided for real-time crosstalk cancellation.

[0042] Embodiment Four: To solve the problem that the input features are incomplete or the output parameters are not targeted enough when predicting crosstalk noise parameters, resulting in inaccurate compensation signal generation, this embodiment further clarifies the prediction logic of crosstalk noise parameters: the amplitude of the signal transition edge, the frequency and the time difference of the adjacent channel transition edge are taken as the input, and the corresponding crosstalk noise amplitude and phase are output.

[0043] Firstly, the selection of input parameters is based on the physical mechanism of crosstalk generation: the amplitude of the transition edge reflects the energy intensity of the adjacent channel signal, the larger the amplitude, the higher the intensity of crosstalk noise; the frequency of the transition edge reflects the rate of data transmission, the higher the frequency, the more abundant the high-frequency components of crosstalk noise; the time difference of the adjacent channel transition edge reflects the synchronism of crosstalk, the smaller the time difference, the more serious the superposition of crosstalk noise and the target channel signal. These parameters are all from real-time monitoring of the crosstalk observer, ensuring the authenticity and timeliness of the input information.

[0044] Further, the selection of output parameters directly serves the generation of compensation signal: the amplitude of crosstalk noise determines the amplitude of compensation signal, both of which need to be equal to achieve complete cancellation; the phase of crosstalk noise determines the phase of compensation signal, both of which need to be opposite, i.e. the phase difference π, to achieve reverse superposition. In the prediction process, the input parameters are normalized, such as mapping the amplitude to the interval of 0-1 and the time difference to the interval of 0-100ps, inputting the pre-trained interference pattern library model, and the model outputs the quantized values of the corresponding crosstalk noise amplitude and phase. For example, when the input transition edge amplitude is 0.5V, the frequency is 10Gbps, and the adjacent channel transition time difference is 20ps, the model outputs the crosstalk noise amplitude of 0.1V and the phase of 180°, at this time the compensation signal needs to be set to amplitude 0.1V and phase 0°, which is opposite to the phase of crosstalk.

[0045] It can be understood that the corresponding relationship between input and output parameters needs to be bound with the demand of crosstalk cancellation, the amplitude and phase are the core parameters of compensation signal, and their accuracy directly determines the effect of crosstalk cancellation. As shown in Figure 4 , the generation of compensation signal needs to strictly follow the predicted crosstalk noise amplitude and phase, so the prediction logic of this embodiment is a necessary link in the crosstalk cancellation process.

[0046] The benefit of this embodiment is that by specifying the prediction input and output of crosstalk noise parameters, it realizes the accurate mapping of transition edge features and core parameters of crosstalk noise, provides direct and effective guidance for the generation of subsequent compensation signal, and ensures the targeting and accuracy of crosstalk cancellation.

[0047] Embodiment Five: To solve the problem of accurate injection of compensation signal and reliable acquisition of bit error rate in crosstalk noise cancellation, this embodiment further refines the operation logic of the pre-emphasis circuit at the sending end and the acquisition method of the bit error rate at the receiving end.

[0048] The pre-emphasis circuit of the sending end is directly connected to the signal input end of the driving transistor, and has an internal integrated voltage control module and a phase adjustment unit: the voltage control module adjusts the signal amplitude output to the driving transistor by changing the supply voltage of the pre-emphasis circuit, so that the amplitude of the compensation signal is completely matched with the amplitude of the crosstalk noise; the phase adjustment unit changes the phase of the compensation signal by fine-tuning the delay time of the signal, so as to ensure that the phase of the compensation signal is opposite to that of the crosstalk noise, thereby realizing reverse compensation. For example, when the amplitude of the crosstalk noise is 0.3V and the phase is π / 2, the voltage control module of the pre-emphasis circuit adjusts the output voltage to 0.3V, and the phase adjustment unit delays the phase of the compensation signal by π, so that the phase difference between the compensation signal and the crosstalk noise is π, achieving the effect of reverse cancellation. When the receiving end obtains the bit error rate, the CRC circuit performs cyclic redundancy check on each frame of data transmitted, counts the number of frames that fail the check, and then calculates the current bit error rate; the Reed-Solomon code check circuit determines the position and number of errors by comparing the received code word with the remainder of the generator polynomial during the decoding process, thereby obtaining the bit error rate.

[0049] It should be understood that the direct connection of the pre-emphasis circuit and the driving transistor enables the compensation signal to be directly injected into the sending signal, reducing the loss in the signal transmission process; and the use of CRC or Reed-Solomon code ensures the reliability of the bit error rate acquisition, providing an accurate basis for the subsequent adjustment of the compensation coefficient.

[0050] The benefit of the present embodiment is that the precise adjustment of the compensation signal amplitude and phase by the pre-emphasis circuit improves the cancellation effect of the crosstalk noise, and the reliable bit error rate acquisition at the receiving end provides a solid foundation for real-time optimization of the system.

[0051] Embodiment Six: To solve the problems of efficiency and accuracy of real-time adjustment of the compensation coefficient, the present embodiment further optimizes the update process of the compensation coefficient. After transmitting each frame of data, the receiving end feeds back the bit error rate of the current frame to the control unit of the sending end, which first calculates the difference between the current bit error rate and the target bit error rate, and then determines the adjustment direction of the compensation coefficient based on the difference.

[0052] Specifically, the control unit adopts the gradient descent method, takes the bit error rate as the objective function, and takes the compensation coefficient as the variable. The gradient of the objective function with respect to the compensation coefficient, i.e., the rate and direction of the change of the bit error rate with the compensation coefficient, is calculated. Subsequently, the control unit calculates the adjustment amount of the compensation coefficient according to the gradient direction and in combination with a preset learning rate: if the gradient is positive, it indicates that the compensation coefficient is too small and needs to be increased; if the gradient is negative, it needs to be decreased. The selection of the learning rate needs to take into account the convergence speed and stability: if it is too large, it may cause the coefficient to oscillate, and if it is too small, it may prolong the convergence time, so the learning rate needs to be dynamically set according to the transmission rate of the system and the change rate of the crosstalk.

[0053] It needs to be understood that the real-time adjustment after each frame of data enables the compensation coefficient to quickly adapt to the change of the crosstalk noise, for example, when the amplitude of the crosstalk noise increases, the adjustment amount of the compensation coefficient also increases accordingly, ensuring that the amplitude of the compensation signal always matches the crosstalk noise.

[0054] The benefit of the embodiment is that the real-time update of the compensation coefficient by the gradient descent method enables the compensation signal to dynamically track the change of the crosstalk noise, maintains the stability of the system bit error rate, and at the same time, the adjustment period of each frame of data takes into account the timeliness of the adjustment and the computational load of the system.

[0055] Embodiment Seven: To solve the problem of comprehensive monitoring and correlation of multi-channel crosstalk, the embodiment further defines the definition of adjacent channels and the monitoring method of the crosstalk observer.

[0056] Adjacent read-write channels refer to two or more channels that are physically adjacent in a solid state disk, for example, in the storage array of a solid state disk, the channels corresponding to adjacent NAND flash particles have signal transmission lines that are close in physical layout, the electromagnetic coupling between the signals is stronger, and the crosstalk noise is more obvious. The crosstalk observer internally integrates multiple transition edge detection units and a channel correlation unit: each transition edge detection unit corresponds to an adjacent channel and monitors the transition time, transition amplitude, and transition slope of the channel signal in real time; the channel correlation unit analyzes the correlation of these features, for example, when the signal of the target channel abnormally transitions, the channel correlation unit compares the transition time of the adjacent channels, and if the transition time of the adjacent channels coincides with the abnormal transition time of the target channel, it is determined that the abnormal transition is caused by the crosstalk of the adjacent channels, and the correlation between the transition features of the adjacent channels and the abnormal transition of the target channel is counted.

[0057] It needs to be understood that the simultaneous monitoring of the transition edge features of multiple adjacent channels can comprehensively obtain the source and features of the crosstalk noise, for example, when two adjacent channels simultaneously transition, the crosstalk noise to the target channel may be the superposition of the two, and the crosstalk observer can accurately extract the superimposed crosstalk features through correlation analysis.

[0058] The advantage of this embodiment is that by simultaneously monitoring multiple channels that are physically adjacent, the crosstalk observer can comprehensively acquire the characteristics of crosstalk noise. The correlation analysis of the channel correlation unit improves the accuracy of crosstalk source judgment and provides a more comprehensive basis for the generation of subsequent compensation signals.

[0059] Example 8: To address the timeliness of the interference pattern library and the dynamic matching between the pre-emphasized waveform and crosstalk noise, this example further refines the update mechanism of the interference pattern library, the application logic of the adaptive filtering algorithm, and the slope adjustment strategy of the pre-emphasized waveform.

[0060] The interference pattern library employs a periodic update mechanism, with the update cycle dynamically adjusted based on the usage time of the solid-state drives (SSDs). For newly manufactured SSDs, the physical characteristics of the storage chips and transmission lines are stable, resulting in a slow rate of change in crosstalk noise characteristics; the update cycle is set to once a month. After more than a year of use, as the number of write cycles to the storage chips increases and the transmission lines age, the amplitude, phase, and other parameters of crosstalk noise change more rapidly, shortening the update cycle to once a week. During the update process, the crosstalk observer inputs recently detected new signal transition edge characteristics, such as combinations of simultaneous rising and falling edge transitions in adjacent channels, along with corresponding crosstalk noise parameters including amplitude, phase, and transition time difference, into the interference pattern library. The library optimizes the mapping relationship using a clustering algorithm, retaining entries highly correlated with the current crosstalk characteristics and deleting or archiving older entries with low correlation. For example, when the transition time difference between adjacent channels decreases from 1 ns to 0.5 ns, the corresponding crosstalk noise amplitude increases from 0.2V to 0.4V. The interference pattern library records this new mapping relationship, replacing the original low-correlation entries to ensure the accuracy of subsequent crosstalk noise predictions.

[0061] Furthermore, the adaptive filtering algorithm employs either the Least Mean Square (LMS) algorithm or the Normalized Least Mean Square (NLMS) algorithm. Taking the NLMS algorithm as an example, it avoids the decrease in convergence speed caused by changes in the input signal amplitude by dividing the step size by the power of the input signal. The input to the adaptive filter is the signal from the adjacent channel monitored by the crosstalk observer, and the desired signal is the difference between the original transmitted signal and the received signal, i.e., the estimated crosstalk noise value. The algorithm iteratively adjusts the filter coefficients to minimize the mean square error between the output compensation signal and the desired crosstalk noise. The crosstalk noise parameters predicted by the interference pattern library are used as the initial coefficients for the adaptive filter, reducing the number of iterations and accelerating the convergence speed. For example, when the interference pattern library predicts that the crosstalk noise amplitude at a certain moment is 0.3V and the phase is π / 3, the adaptive filtering algorithm uses this parameter as the initial value and quickly adjusts the coefficients to make the compensation signal closer to the actual crosstalk noise, thus shortening the adaptation time.

[0062] Further, the slope of the pre-emphasis waveform is adjusted according to the amplitude of the crosstalk noise parameter, and the greater the amplitude of the crosstalk noise, the greater the slope of the pre-emphasis waveform. The output waveform of the pre-emphasis circuit is obtained by superimposing the original signal and the pre-emphasis pulse, and the slope of the pre-emphasis pulse determines the steepness of the transition edge of the compensation signal. When the amplitude of the crosstalk noise is large, its influence on the transition edge of the original signal is more obvious, such as making the rising edge of the original signal slow or overshooting, and increasing the slope of the pre-emphasis waveform can make the transition edge of the compensation signal steeper, specifically offsetting the steep transition part of the crosstalk noise and maintaining the transition edge characteristics of the original signal. For example, when the amplitude of the crosstalk noise is 0.4V, the slope of the pre-emphasis waveform is set to 0.8V / ns; when the amplitude of the crosstalk noise increases to 0.6V, the slope is adjusted to 1.2V / ns, so that the transition edge of the compensation signal is steeper, and the influence of the crosstalk noise on the original signal is more effectively offset.

[0063] The benefit of the embodiment is that, by periodically updating the interference pattern library, the timeliness of the mapping relationship is maintained, and the crosstalk characteristics change during the use of the solid state disk are adapted; the application of the adaptive filtering algorithm improves the accuracy of the compensation signal; the dynamic adjustment of the slope of the pre-emphasis waveform enhances the offsetting ability of different amplitude crosstalk noise, and the three synergistically improve the effect of crosstalk offsetting.

[0064] Embodiment Nine: To solve the problem of dynamic offsetting of crosstalk noise in the multi-channel reading and writing process of the solid state disk, the embodiment provides a closed-loop system integrating crosstalk monitoring, prediction, compensation and adjustment, which realizes accurate offsetting of crosstalk noise through the cooperative work of each module.

[0065] To visually show the actual influence of crosstalk noise on the signal integrity of the solid state disk reading and writing channel, as shown in Figure 2 , the embodiment simulates the distortion effect of crosstalk on the data eye diagram at a rate of 10Gbps, and clearly shows the adverse effect of crosstalk noise on signal detection, providing a problem basis for the design of the subsequent crosstalk offsetting system.

[0066] Figure 2Based on the actual application scenario parameters of solid state disk, the original signal is a random NRZ (non-return-to-zero) signal with a symbol rate of 10 Gbps. A 32-fold oversampling is adopted, and the signal amplitude is normalized to ±0.5V, corresponding to the actual signal swing of 1V. The crosstalk noise is composed of two parts: a 0.3V sinusoidal signal coupled by adjacent channels, with a frequency of 15GHz, close to the frequency of the original signal, a crosstalk coupling coefficient of about -20dB, and a 0.2V Gaussian white noise, with a variance of 0.05, simulating the inherent noise of the circuit. After superimposing the original signal and the crosstalk noise, the signal level is obviously offset, and the transition edge appears irregular jitter: the eye height of the original signal is about 0.8V (high-low level difference), the eye width is about 0.15ns (decision window width), and the jitter is less than 0.02ns (signal transition time deviation); while the eye height of the signal with crosstalk noise decreases to 0.4V, the eye width narrows to 0.08ns, and the jitter increases to 0.05ns. The closure of the eye diagram directly leads to the difficulty of the receiving end decision circuit to distinguish the high and low levels: under the same signal-to-noise ratio (SNR=10dB), the original signal bit error rate is about 10 -9 , and the signal with crosstalk noise bit error rate rises to 10 -6 , which cannot meet the strict requirement of solid state disk that the bit error rate is less than or equal to 10 -7 .

[0067] The signal distortion is caused by the dynamic superposition of crosstalk noise and original signal, and the strength, frequency, phase and other characteristics of crosstalk change in real time with the read-write state of adjacent channels (such as transition time, direction, amplitude). For example, when the reverse transition occurs in adjacent channels (such as the rising edge of channel 1 and the falling edge of channel 2 occurring at the same time), the crosstalk coupling degree is about 2dB higher than that of the same direction transition, which leads to further closure of the eye diagram. Therefore, accurately capturing the signal distortion characteristics caused by crosstalk, such as transition edge jitter and eye diagram parameter change, is the premise of realizing crosstalk cancellation. The design goal of the crosstalk observation module is to achieve this, which provides bottom layer data support for subsequent crosstalk prediction and compensation by real-time acquisition of these characteristics through high-speed sampling circuit.

[0068] As the perception layer of the system, the sampling rate of the crosstalk observation module is not less than 10GSa / s. The signal waveform of adjacent read-write channels is captured in real time through high-speed sampling circuit, and the signal transition edge characteristics are extracted, including the time difference of adjacent channel transition, the combination of transition direction and the ratio of transition amplitude. These characteristics directly reflect the dynamic changes of crosstalk source. As shown in Figure 3 , the crosstalk observation module is built-in transition edge detection unit, which marks the transition event of a single channel; the channel association unit associates the transition events of adjacent channels to form complete signal transition edge characteristics, which provides accurate input for subsequent crosstalk prediction.

[0069] Further, the crosstalk prediction module is connected to the crosstalk observation module, and the core thereof is a pre-trained interference mode library. The library establishes a mapping relationship between a signal jump edge feature and crosstalk noise parameters (amplitude, phase, and frequency) by collecting a large amount of measured data of a solid state disk under different use scenarios (such as high-load reading and writing and low-load standby). When receiving the feature output by the crosstalk observation module, the prediction module queries the closest entry in the library through a similarity matching algorithm, and outputs corresponding crosstalk noise parameters. For example, when it is monitored that the time difference of the jump edge of channel 1 and channel 2 is 0.5 ns and the jump directions are opposite, the interference mode library predicts that the crosstalk noise amplitude is 0.3 V and the phase is π / 2, and provides target parameters for compensation injection.

[0070] Further, the compensation injection module is connected to the crosstalk prediction module and a pre-emphasis circuit of the sending end, and the function thereof is to convert the crosstalk noise parameters into an executable compensation signal waveform. The module first determines the amplitude of the compensation signal according to the predicted crosstalk noise amplitude, and the amplitude is usually 1.1 times the crosstalk noise amplitude to reserve a margin, determines the phase of the compensation signal according to the phase, and the compensation signal is different from the crosstalk noise by π to achieve reverse cancellation, and adjusts the delay time of the compensation signal according to the frequency to ensure that the compensation signal and the crosstalk noise are superimposed in the time domain. As shown in FIG. 4, the compensation injection module injects the compensation signal into the pre-emphasis waveform of the original sending signal through the pre-emphasis circuit. After the original signal is disturbed by the crosstalk noise, a distortion signal is generated. The compensation signal and the distortion signal are superimposed, the influence of the crosstalk noise is offset, and the steepness and amplitude of the jump edge of the original signal are restored. Figure 4

[0071] Further, the adaptive adjustment module, as a closed-loop feedback layer of the system, is connected to the compensation injection module and a bit error rate detection circuit of the receiving end. The adaptive adjustment module adopts an adaptive filtering algorithm, takes the bit error rate detection result as a feedback signal, and adjusts the coefficient value of the compensation signal in real time. When the bit error rate increases, it indicates that the matching degree of the current compensation signal and the actual crosstalk noise decreases. The algorithm adjusts the filter coefficient by iteration, such as increasing the amplitude coefficient of the compensation signal or adjusting the phase offset, so that the waveform of the compensation signal is closer to the reverse value of the actual crosstalk noise. When the bit error rate decreases to a target value (such as 10 -7 below), the algorithm stops adjusting and keeps the current coefficient. For example, when the bit error rate increases from 10 -6 to 10 -5 , the adaptive adjustment module adjusts the amplitude coefficient of the compensation signal from 0.8 to 0.9, and enhances the offset effect.

[0072] The benefits of the embodiment are that the crosstalk source is dynamically perceived through the crosstalk observation module, the crosstalk noise is accurately predicted through the crosstalk prediction module, the compensation injection module realizes the early offset of the sending end, and the adaptive adjustment module realizes the closed-loop optimization of the receiving end, thereby forming a complete crosstalk offset link of perception, prediction, compensation, and optimization, and effectively improving the signal integrity of the reading and writing channel of the solid state disk.​

[0073] Embodiment ten: To solve the problem of dynamic avoidance of crosstalk risk in multi-channel parallel read and write of solid state disk, this embodiment provides a solid state disk based on the crosstalk cancellation system, which realizes intelligent allocation of channel tasks through a multi-channel scheduling module, combines crosstalk state monitoring and task allocation optimization, and forms a crosstalk control strategy of active avoidance and passive cancellation.

[0074] Specifically, the storage array integrates 8 parallel read and write channels, each channel corresponds to an independent NAND flash particle, supports a single-channel data rate of 10 Gbps, the physical distance between channels is 0.5 mm, and the measured crosstalk coupling coefficient of adjacent channels is about -30 dB. The read and write tasks of each channel are uniformly scheduled by the host unit, the host unit serves as the data transmission hub, is responsible for converting the host PCIe 4.0 interface instructions into specific operations of the storage array, and coordinates the work of the crosstalk cancellation system. The channel signal transition edge characteristics (such as transition time, direction, amplitude) monitored by the crosstalk observation module are transmitted to the host unit in real time, providing the host unit with bottom state basis for task allocation.

[0075] Further, the multi-channel scheduling module connects the crosstalk observation module and the host unit, and its function is to convert the crosstalk observation data into the decision basis for channel task allocation. The module calculates the crosstalk risk value of each channel combination through an algorithm, and the calculation logic is based on three key features of adjacent channel transition edges: time difference (the smaller the Δt, the higher the crosstalk superposition probability, and the higher the risk value), direction combination (the crosstalk coupling degree of opposite direction transition is higher than that of the same direction, and the risk value weighting coefficient is 1.2), and amplitude ratio (the larger the amplitude ratio of adjacent channel transition, the more crosstalk energy is transferred, and the risk value increases linearly). For example, when the time difference of the transition edges of channel 1 and channel 2 is 0.2 ns (Δt threshold is 0.3 ns), the transition direction is opposite, and the amplitude ratio is 0.25 (threshold is 0.2), the crosstalk risk value is calculated as 0.75 (full score is 1), which belongs to a high-risk combination.

[0076] Further, when the host control unit receives a read-write task of the host, such as a 1GB file write, 2 channels need to be allocated and executed in parallel. The multi-channel scheduling module first queries the usage state of each channel, idle or busy, then extracts the crosstalk risk value of the idle channel in the last 10ms, and sorts it from low to high according to the risk value. For example, the idle channels are 3, 4, 5 and 6, the risk values of channels 3 and 4 are 0.3, low risk, and the risk values of channels 5 and 6 are 0.6, medium risk. The scheduling module recommends using channels 3 and 4 to execute the task to the host control unit. If the task is executed in the process, the crosstalk observation module monitors the transition edge time difference of channel 3 to shorten to 0.15ns, and the risk value rises to 0.85. The scheduling module immediately sends a state update to the host control unit, and the host control unit transfers the subsequent unfinished task to idle channels 7 and 8, with a risk value of 0.2, to avoid excessive compensation pressure of the crosstalk cancellation system under high-risk combination.

[0077] It should be understood that the multi-channel scheduling module and the crosstalk cancellation system form a cooperative mechanism: when the crosstalk risk value of a certain channel combination is high but cannot be avoided, such as all channels being in a busy state, the compensation injection module of the crosstalk cancellation system will automatically adjust the compensation signal amplitude, such as from 0.3V to 0.35V, to ensure that the bit error rate remains within the target range of 10 -7 When there is an idle channel, the scheduling module preferentially selects a low-risk combination to reduce the sustained workload of the compensation system and prolong the service life of the compensation circuit. This cooperative mode not only solves the crosstalk cancellation problem under high-risk scenarios, but also avoids potential risks through task allocation, thereby improving the overall reliability of the system.

[0078] The benefit of the present embodiment is that the crosstalk risk avoidance is integrated into the task allocation process through the multi-channel scheduling module, forming a synergy with the crosstalk cancellation system, and further improving the signal integrity of the solid state disk. The dynamic adjustment mechanism of the scheduling module ensures that the channel task allocation is always based on the latest crosstalk state, avoiding the accumulation of crosstalk risk under the fixed channel allocation mode, and is suitable for high-load, multi-task enterprise-level solid state disk application scenarios, meeting the strict requirements for data transmission reliability.

[0079] Although the present application has been described above with specific reference to the preferred embodiments thereof, it is to be understood that this application is not limited to the embodiments described above, but that various modifications and changes can be made thereto without departing from the spirit and scope of the present application.

Claims

1. A method for hard disk read-write channel crosstalk cancellation, characterized by the steps of The method comprises the following steps: Real-time monitoring of the signal transition edge characteristics of adjacent read-write channels of a solid state disk, obtaining dynamic characteristic parameters of crosstalk sources; Based on the pre-trained interference mode library, the corresponding crosstalk noise parameters are predicted according to the signal transition edge characteristics, and the association between the transition edge and the crosstalk is established; Injecting a compensation signal opposite to the crosstalk noise parameters into the pre-emphasis waveform at the sending end to offset the crosstalk noise; Using an adaptive filtering algorithm to adjust the coefficient value of the compensation signal in real time according to the error rate feedback of the receiving end, and dynamically optimizing the compensation effect.

2. The method of claim 1, wherein the hard disk read / write channel cross-talk cancellation is performed by a hard disk read / write channel cross-talk cancellation module. The real-time monitoring of the signal transition edge characteristics is realized by a crosstalk observer, which comprises a transition edge detection unit and a channel association unit. The transition edge detection unit detects the amplitude, frequency and time point of the signal transition edge of each channel, and the channel association unit associates the transition edge signals of adjacent channels.

3. The method of claim 1, wherein the method further comprises: determining a read channel crosstalk cancellation (RCCC) parameter for the read channel; and determining a write channel crosstalk cancellation (WCCC) parameter for the write channel. The pre-training method of the interference mode library comprises collecting signal transition edge characteristics and corresponding crosstalk noise parameters in historical data, and using support vector machine or neural network algorithm to establish the mapping relationship between the transition edge characteristics and the crosstalk noise parameters.

4. The method of claim 1, wherein the hard disk read / write channel cross-talk cancellation is performed by a hard disk read / write channel cross-talk cancellation module. The prediction of the corresponding crosstalk noise parameters according to the signal transition edge characteristics comprises inputting the amplitude, frequency and time difference of the adjacent channel transition edge of the signal transition edge into the pre-trained interference mode library, and outputting the amplitude and phase of the corresponding crosstalk noise.

5. The method of claim 1, wherein the hard disk read / write channel cross-talk cancellation is performed by a hard disk read / write channel. The injection of the compensation signal opposite to the crosstalk noise parameters into the pre-emphasis waveform at the sending end is realized by adjusting the output voltage amplitude and phase of the pre-emphasis circuit, and the pre-emphasis circuit is connected to the driving transistor at the sending end. The error rate of the receiving end is obtained by a cyclic redundancy check circuit or a Reed-Solomon code check circuit.

6. The method of hard disk read / write channel crosstalk cancellation as claimed in claim 1, wherein, One step of adjusting the coefficient value of the compensation signal in real time comprises calculating the adjustment amount of the compensation coefficient according to the error rate after transmitting one frame of data, and updating the coefficient value of the compensation signal using the gradient descent method.

7. The method of hard disk read / write channel crosstalk cancellation as claimed in claim 2, wherein, The adjacent read-write channels are two or more channels with adjacent physical positions in the solid state disk, and the crosstalk observer simultaneously monitors the signal transition edge characteristics of the two or more channels.

8. The method of hard disk read / write channel crosstalk cancellation as claimed in claim 3, wherein, The interference mode library is updated periodically, the mapping relationship is optimized using new signal transition edge characteristics and corresponding crosstalk noise parameters, and the update period is adjusted according to the usage time of the solid state disk; the adaptive filtering algorithm is the least mean square algorithm or the normalized least mean square algorithm; the slope of the pre-emphasis waveform is adjusted according to the amplitude of the crosstalk noise parameters, and the greater the amplitude of the crosstalk noise parameters, the greater the slope of the pre-emphasis waveform.

9. A hard disk read / write channel cross-talk cancellation system for implementing the hard disk read / write channel cross-talk cancellation method according to any one of claims 1-8, characterized by The method comprises the following steps: A crosstalk observation module is used to real-time monitor the signal transition edge characteristics of adjacent read-write channels of a solid state disk, and obtain dynamic characteristic parameters of crosstalk sources; A crosstalk prediction module connected to the crosstalk observation module is used to predict the corresponding crosstalk noise parameters based on a pre-trained interference mode library according to the signal transition edge characteristics; A compensation injection module connected to the crosstalk prediction module and a pre-emphasis circuit at the sending end is used to inject a compensation signal opposite to the crosstalk noise parameters into the output waveform of the pre-emphasis circuit. An adaptive adjustment module, connected to the compensation injection module and an error rate detection circuit of the receiving end, is configured to adopt an adaptive filtering algorithm to adjust a coefficient value of the compensation signal in real time according to a feedback result of the error rate detection circuit.

10. A solid state drive, comprising: The method comprises: a storage array integrating multiple read-write channels; a host unit connected to the storage array and the hard disk read-write channel crosstalk cancellation system of claim 9; and a multi-channel scheduling module connected to the crosstalk observation module and the host unit; configured to analyze crosstalk risks of each channel combination according to adjacent channel signal transition edge features monitored by the crosstalk observation module, and adjust task allocation of each read-write channel by the host unit, and preferentially use channel combinations with low crosstalk risks.

Images