Cdr optimization control device for video transmission network

Abstract

The application discloses a kind of CDR optimization control device of video transmission network.It includes CDR fast controller, memory and processor;CDR fast controller is used to receive the error between reference clock and local clock output by fractional order clock data recovery module, according to the error of first period, fuzzy input vector is constructed, fuzzy inference is carried out to the fuzzy input vector based on fuzzy unit and fuzzy mapping table, and the recovery control vector of fractional order clock data recovery module is updated;The first optimization program of CDR fast controller is stored on memory, and the following first optimization processing is realized when the first optimization program is executed by processor: the transmission quality of the video data is monitored according to the second period greater than the first period, and the membership parameters of the fuzzy calculation unit in CDR fast controller and the fuzzy rules in the fuzzy mapping table are optimized using the first adaptive exploration utilization strategy in response to the transmission quality decline.The application can adapt to the increasingly complex and dynamic environment of video transmission network, and realize efficient, robust, adaptive clock data recovery.

Description

Technical Field

[0001] This invention relates to the field of high-speed video information transmission technology, and in particular to a CDR optimization control device for video transmission networks. Background Technology

[0002] High-speed video transmission is a key component of modern electronic systems, widely used in automotive electronics (ADAS, autonomous driving), remote monitoring (on-site monitoring, teleoperation vision), and high-end industrial automation (machine vision)—fields with extreme requirements for real-time performance, reliability, and image quality. These applications often involve harsh operating conditions such as wide temperature range variations, strong electromagnetic interference, and changing operating environments, while simultaneously needing to handle the continuously increasing data rates brought about by high-definition and even ultra-high-definition video. The stable and reliable transmission of video signals relies heavily on high-performance clock data recovery (CDR) technology to ensure accurate recovery of clock and data from signals that may contain noise and jitter, guaranteeing the final video quality. However, with the continuous improvement of application demands and the increasing complexity of working environments, traditional CDR technology faces unprecedented challenges.

[0003] The CDR (Continuous Data Retrieval) technology widely used in existing high-speed video transmission systems is largely based on classical loop control theory. However, facing increasingly complex and dynamically changing application environments, especially harsh conditions such as wide temperature ranges, variable operating conditions, and temperature disturbances, existing CDR systems lack rapid response and adaptability. This leads to problems such as decreased clock data recovery accuracy, limited dynamic signal tracking capabilities, and difficulty in guaranteeing long-term operational performance. These issues constitute the core bottleneck of existing video transmission, restricting its performance under harsh and dynamic conditions. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a CDR optimization control device for video transmission networks, aiming to improve the ability of clock data recovery to adapt to the increasingly complex and dynamically changing environment of video transmission networks, and to achieve efficient and robust clock data recovery.

[0005] To achieve the above objectives, according to a first aspect of the present invention, a CDR optimization control device for a video transmission network is provided, comprising a CDR fast controller, a memory, and a processor; wherein,

[0006] The CDR fast controller is used to receive the error between the reference clock output by the fractional clock data recovery module and the local clock, construct a fuzzy input vector according to the error according to a preset first cycle, perform fuzzy inference on the fuzzy input vector based on the fuzzification unit and the fuzzy mapping table, and update the recovery control vector of the fractional clock data recovery module; the reference clock is the reference clock in the video data received from the video transmission network. The memory stores a first optimization program for the CDR fast controller. When the processor executes the first optimization program, it performs the following first optimization process: The transmission quality of the video data is monitored according to a preset second cycle. In response to a decrease in transmission quality, the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table are optimized using a first adaptive exploration and utilization strategy. The second cycle is longer than the first cycle.

[0007] Furthermore, the fuzzy input vector includes the phase error of the first period and the rate of change of the phase error; the fuzzification unit includes a first fuzzification unit corresponding to the phase error and a second fuzzification unit corresponding to the rate of change of the phase error; Based on the fuzzification unit and fuzzy mapping table, fuzzy inference is performed on the fuzzy input vector to update the recovery control vector of the fractional clock data recovery module, including: The first fuzzification unit is used to obtain the first The first membership degree of the ambiguity set to which the phase error of the first period belongs is obtained by using the second ambiguity unit. The second membership degree of the phase error change rate in the first cycle is used to obtain the first fuzzy set to which the phase error belongs and the second fuzzy set to which the phase error change rate belongs, based on the first membership degree. It is a natural number; Using the combination of the first fuzzy set and the second fuzzy set as an index to query the fuzzy rule, we obtain... The control adjustment vector and their respective weights , ; According to the recovery control vector in the... The value of the first cycle The control adjustment vector and the corresponding weights The recovery control vector is obtained at the 1st... The value of the first cycle .

[0008] Furthermore, the fuzzification unit includes a calculator and a comparator; the calculator is used to calculate the membership value of the fuzzy set corresponding to the fuzzification unit, and the comparator is used to determine whether the corresponding component in the fuzzy input vector belongs to the fuzzy set based on the membership value.

[0009] Furthermore, the first optimization process employs a first adaptive exploration and utilization strategy to optimize the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table, including: Using the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table as spores, the initial value of the spore population is generated, and the following iterative processing is performed on the spore population: For the Round iteration generates the adaptive operator for this round. And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; like Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; If the new position First fitness is better than original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the first fitness of a spore is better than the best first fitness in the spore population, then the best spore in the spore population is updated to the first fitness. The spores described; The spore population completed the first... After a round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the highest fitness in the spore population is used to update the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The memory also stores a first neural network model, which the processor calls during the first optimization process to complete the first... The spores at position First fitness The prediction, the The spores at position First fitness Long-term bit error rate compared to the output of the first neural network model trained offline They are inversely proportional; the input to the first neural network model is the first... The location of the spores The output is the corresponding long-term bit error rate. .

[0010] Furthermore, the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table are used as spores to generate the initial values ​​of the spore population, including: Based on the size of the spore population and the dimensions of the spores Generate size is A chaotic matrix, wherein the columns of the chaotic matrix correspond to the dimensions of the spores, and the rows of the chaotic matrix correspond to the spores in the spore population; The elements of each column vector of the chaotic matrix are mapped to the value space of the corresponding dimension of the recovery control vector to obtain the values ​​of each dimension of the spore. After the mapping is completed, a total of The spores mentioned above The number of spores is the initial value of the spore population; No. The adaptive operator in the round iteration follows the formula as the number of iterations increases. Adjustment: ,in, and These are the upper and lower limits of the adaptive operator's values; The global exploration strategy is: The local utilization strategy is: ; in, It refers to the current position of three spores randomly selected from the spore population. It is the current position of the spore with the highest fitness in the spore population; It is an exploratory factor. It is the step size scaling factor. It is a random scalar step size generated based on the Levy distribution; It is a factor of utilization. It is the current position of a spore randomly selected from the spore population.

[0011] Furthermore, it also includes a temperature sensor for acquiring temperature and an interference sensor for acquiring interference frequency and interference amplitude; The memory stores a second optimization program for the CDR fast controller. When the processor executes the second optimization program, it performs the following second optimization process: If changes in environmental characteristics are detected and continue to exceed a preset time threshold, a second adaptive exploration and utilization strategy is adopted to optimize the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The environmental characteristics include temperature, interference frequency, and interference amplitude.

[0012] Furthermore, the second optimization process employs a second adaptive exploration and utilization strategy to optimize the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table, including: The membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table are used as spores to generate the initial value of the spore population, based on the number of pre-approval iteration stages. Determine the first Each iteration phase The value of the adaptive operator is , , Number Greater than 2 and less than the maximum number of iterations; The following iterative process was performed on the spore population: For the Round iteration, And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; like Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; If the new position The second fitness is better than the original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the second fitness of a spore is better than the optimal second fitness in the spore population, then the optimal spore in the spore population is updated to the second fitness. The spores described; The spore population completed the first... After one round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the second highest fitness in the spore population is used to update the recovery control vector of the fuzzy inference module. The memory also stores a first neural network model, which the processor calls during the first optimization process to complete the first... The spores at position Second fitness Prediction, the first The spores at position Second fitness Long-term bit error rate compared to the output of the offline-trained second neural network model They are inversely proportional; the input to the second neural network model is the first... The location of the spores The temperature The interference amplitude and the interference frequency The output is the corresponding long-term bit error rate. .

[0013] Furthermore, the memory stores the temperature feedforward compensation program of the CDR fast controller, and when the processor executes the temperature feedforward compensation program, it performs the following temperature compensation processing: Monitoring ambient temperature, in response to a detected abrupt change in ambient temperature, a temperature-dependent perturbation effect model is employed, based on the current temperature. Predict the output offset of the fractional clock data recovery module According to the output offset and the current frequency of the fractional clock data recovery module Temperature compensation is performed on the local clock in the fractional clock data recovery module.

[0014] According to a second aspect of the present invention, a clock data recovery device for a video transmission network is provided, including a fractional clock data recovery module and a CDR optimization control device for a video transmission network as described in the first aspect of the present invention.

[0015] Furthermore, the fractional clock data recovery module includes a digital fractional loop filter and a numerically controlled oscillator, and the recovery control vector of the fractional clock data recovery module includes the proportional gain, integral gain, and fractional order of the digital fractional loop filter; The fractional-order clock data recovery module adjusts the frequency of the local clock using the error between the reference clock and the local clock, and the recovery control vector, including: According to the digital fractional-order loop filter in the first... The proportional gain of the first cycle Integral gain and fractional order The coefficients of the first difference equation of the digital fractional-order loop filter are determined by looking up a table. coefficients of the second difference equation ; Based on the aforementioned error and the coefficients of the first difference equation coefficients of the second difference equation The control signal of the numerically controlled oscillator is obtained according to the following formula. : ; According to the main control signal Or based on the main control signal After feedforward compensation, the frequency of the output signal of the numerically controlled oscillator is adjusted. .

[0016] The embodiments of the present invention have at least one of the following advantages or beneficial effects: This invention provides a processing architecture for a CDR (Continuous Delayed Response) optimization control device in a video transmission network. On one hand, the CDR fast controller within the device enables real-time response to phase error changes on a microsecond to millisecond timescale, dynamically adjusting the local clock of the fractional-order loop filter module. On the other hand, a processor executes a first optimization program stored in memory to perform first optimization processing, continuously optimizing the membership parameters and fuzzy rules of the CDR fast controller in the background. This processing architecture avoids control conflicts, enabling the device of this invention to not only cope with instantaneous disturbances but also adapt to long-term drift, aging, and gradual environmental changes, achieving continuous performance optimization.

[0017] The optimization control device of this invention monitors changes in transmission quality. On the one hand, it employs a first optimization program with a first adaptive exploration and utilization strategy to periodically optimize the membership parameters and fuzzy rules of the CDR fast controller for implicit influencing factors of transmission quality. On the other hand, it employs a second optimization program with a second adaptive exploration and utilization strategy to optimize the membership parameters and fuzzy rules of the CDR fast controller for explicit influencing factors of transmission quality. This comprehensively enhances the adaptive capability of clock data recovery in video transmission networks to various influencing factors, ensuring the stability of long-term clock data recovery performance.

[0018] The optimized control in this embodiment of the invention integrates sensing modules such as temperature sensors and interference sensors, enabling real-time acquisition of physical quantities such as ambient temperature, interference frequency and amplitude. These hardware sensing signals are directly input to the second optimization program and temperature feedforward compensation program executed by the processor. After sensing changes in environmental characteristics, the fractional clock data recovery module can adaptively and dynamically adjust and directly compensate. This closed-loop hardware architecture of "perception-decision-execution" gives the optimized control device an inherent ability to resist environmental disturbances.

[0019] The optimized control device of this invention adopts a modular design, which facilitates hardware and software integration with the fractional clock data recovery module. It can meet the stringent real-time requirements of high-speed video transmission while ensuring high performance, and is suitable for widespread application in embedded systems such as automotive electronics and industrial vision.

[0020] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0021] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0022] Figure 1 This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the adaptive capability of the CDR fast controller in an optimized control device according to an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of the first optimization processing adaptive capability implemented in an optimization control device according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to another embodiment of the present invention; Figure 5 This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to another embodiment of the present invention.

[0025] Figure 6 This is a schematic diagram illustrating the peak suppression performance gain of the temperature compensation processing implemented in the optimized control device according to another embodiment of the present invention when dealing with instantaneous temperature pulse disturbances; Detailed Implementation Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0026] Figure 1 This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to an embodiment of the present invention. Figure 1 As shown, the CDR optimization control device for a video transmission network in this embodiment of the present invention includes a CDR fast controller, a memory, and a processor. The CDR fast controller receives the error between a reference clock output by a fractional clock data recovery module and a local clock, constructs a fuzzy input vector based on the error according to a preset first period, performs fuzzy inference on the fuzzy input vector based on fuzzification units and a fuzzy mapping table, and updates the recovery control vector of the fractional clock data recovery module. The reference clock is a reference clock in the video data received from the video transmission network. The memory stores a first optimization program for the CDR fast controller. When the processor executes the first optimization program, it performs the following first optimization process: monitors the transmission quality of the video data according to a preset second period, and in response to a decrease in transmission quality, optimizes the membership parameters of the fuzzification calculation units in the CDR fast controller and the fuzzy rules in the fuzzy mapping table using a first adaptive exploration and utilization strategy. The second period is longer than the first period.

[0027] Specifically, in this embodiment and some embodiments of the present invention, there are one or more memories and one or more processors. The memories and processors are connected via a bus, which may include any number of interconnected buses and bridges. The bus connects various circuits of one or more processors and one or more memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, through interfaces (internal interfaces, external interfaces), which are well known in the art. The interface provides an interface between the bus and the transceiver, such as a communication interface or a user interface. The transceiver may be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over a wireless medium via an antenna, which further receives data and transmits it to the processor.

[0028] Specifically, in this embodiment and some embodiments of the present invention, the processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. The memory can be used to store data and programs used by the processor during operation.

[0029] Specifically, such as Figure 1 As shown in this embodiment and some embodiments of the present invention, the CDR optimization control device of the video transmission network in this embodiment is integrated with the fractional clock recovery module for clock data recovery during high-speed video transmission. The fractional clock data recovery module (FOCDR) is the foundation for realizing clock data recovery during video transmission in this embodiment, and its input is the reference clock phase information extracted from the front end of the high-speed transceiver (SerDes) of the corresponding camera. Its output is a stable clock phase. (This is the recovery clock in a general sense; the specific feedback to the PD / PFD is the frequency-divided phase.) The FOCDR is subject to parameter regulation by a two-layer adaptive control module consisting of fast and slow control, and a feedforward module.

[0030] Specifically, such as Figure 1 As shown, the fractional clock data recovery module mainly consists of the following components: Phase / Frequency Detector (PD / PFD): Its core function is to compare the phase of the input reference clock. Phase with locally generated feedback clock and output phase error signal Feedback clock phase The output clock phase of the digitally controlled oscillator (DCO) The frequency is obtained by dividing the frequency using a frequency divider. The output of PD / PFD can be expressed as:

[0031] in, It is the phase detector gain, and this error signal Drive the entire CDR loop.

[0032] Fractional-Order Loop Filter (FO-LF): This filter receives the error signal from the PD / PFD. As input, and output control signal For a digitally controlled oscillator (DCO), the design of its transfer function introduces a fractional order. It allows for dynamic adjustment of parameters, both of which are determined in real time by the CDR fast controller. Its continuous-time domain transfer function... for:

[0033] in, It is proportional gain. It is integral gain, a fractional-order operator. Within the specified frequency range The interior can be approximated as: ; in, It is the approximation order, a fixed integer such as 5 or 7 pre-selected based on the system's requirements for approximation accuracy and the trade-offs in hardware resources. Bring into Bilinear transformation is applied in the middle and later stages. The z-domain transfer function of the fractional-order loop filter in the discrete-time domain is obtained, and its common denominator is rearranged into the following form:

[0034] The behavior of the filter in the discrete-time domain of the digital signal is described by a difference equation. In this embodiment, an IIR (Infinite Impulse Response) filter approximation method is used, which has advantages in resource consumption and operating speed, and is more suitable for high-performance implementation in hardware. The z-domain transfer function mentioned above... The equivalent time-domain difference equation has the following form: ; Generally, and equal ; and It is by , , , , , The function. Whenever the algorithm is updated. , , , , , Then, a new set is calculated. and Coefficients. In hardware, this is executed through a series of pipelined multiply-accumulate (MAC) units, implemented using combinational logic and lookup tables (LUTs), where the coefficients... and It is loaded into the corresponding hardware register in real time. The filtering characteristics of the filter are changed immediately in the next clock cycle. The final calculation result is output to the numerically controlled oscillator.

[0035] Digitally Controlled Oscillator (DCO): The DCO operates based on the main control signal output from the FO-LF. Or based on the main control signal After feedforward compensation, the output frequency of the signal is adjusted. Its nominal center frequency is... (Default value), gain is Compensation control signal Provided by the feedforward module, and in the main control signal The input is superimposed on the previous input and finally fed into the DCO to offset the DCO frequency shift caused by sudden temperature changes. This refers to the tuning sensitivity of the DCO, which represents the frequency change caused by a unit change in the total control signal, and the instantaneous frequency of the DCO output. for: ,or,

[0036] Among them, the clock phase of the DCO output It is the integral of its frequency:

[0037] Frequency divider: Located between the DCO output and the PD / PFD feedback input, its division ratio is... This can be set by the user or the system. The clock frequency after frequency division is... The corresponding phase Feedback is sent to PD / PFD and participates in the error. The calculation. The recovered DCO output clock can then be used. The input video data stream is sampled and processed to output a clean video data stream.

[0038] Understandably, the fractional clock data recovery module forms a negative feedback control loop through the aforementioned components, and its basic goal is to drive the phase error. Approaching zero, causing the feedback clock phase Accurately track the phase of the input reference clock This allows us to obtain a clean and stable clock from the raw data stream.

[0039] In summary, in this embodiment and some embodiments of the present invention, the recovery control vector of the fractional-order clock data recovery module includes the proportional gain, integral gain, and fractional order of the digital fractional-order loop filter. The fractional-order clock data recovery module adjusts the frequency of the local clock using the error between the reference clock and the local clock and the recovery control vector, including: According to the digital fractional-order loop filter in the first... The proportional gain of the first cycle Integral gain and fractional order The coefficients of the first difference equation of the digital fractional-order loop filter are determined by looking up a table. coefficients of the second difference equation ; Based on the aforementioned error and the coefficients of the first difference equation coefficients of the second difference equation The main control signal of the numerically controlled oscillator is obtained according to the following formula. : ; According to the main control signal Or based on the main control signal After feedforward compensation, the frequency of the output signal of the numerically controlled oscillator is adjusted. ,in, ,or, , This indicates the center frequency of the numerically controlled oscillator. This is the gain coefficient. It is the compensation control signal output by the feedforward module.

[0040] It is understood that in this embodiment and some embodiments of the present invention, adjusting the center frequency of the numerically controlled oscillator is equivalent to adjusting the frequency of the local clock.

[0041] Specifically, such as Figure 1As shown, in this embodiment and some embodiments of the present invention, the first optimization process implemented by the CDR fast controller and processor when executing the first optimization program adjusts the key parameters in the fractional clock data recovery module loop (as shown). , , These components work together to optimize the core performance of the fractional clock data recovery module. Specifically, the CDR fast controller responds quickly within the first cycle (microseconds to milliseconds) and suppresses instantaneous phase errors caused by data jitter and transient noise, ensuring the CDR loop can quickly lock onto and stably track the input signal. The first optimization process, within a longer second cycle (seconds to minutes), uses an optimization strategy to find a set of optimal fuzzy controller parameters to minimize the long-term average bit error rate (BER), thereby improving the overall performance, robustness, and reliability of clock data recovery under long-term operation.

[0042] Understandably, in this embodiment and some embodiments of the present invention, a fractional-order loop filter (FO-LF) is used, by introducing a fractional order. Compared to traditional CDRs, these parameters provide greater freedom for the entire system and offer more diverse loop characteristics, giving the optimization device of this embodiment greater adjustment space and providing a foundation for the device to withstand harsh conditions and operate under multiple working conditions.

[0043] Specifically, in this embodiment and some embodiments of the present invention, when the explicit influencing factors of transmission quality are stable (temperature and interference, etc.), a dual-time-scale fuzzy adaptive control is adopted. This involves the division of labor and efficient collaboration between the CDR fast controller and the first optimization process to dynamically optimize the performance of the fractional-order clock data recovery module. Considering the stringent requirements of high-speed video transmission for real-time response, and the speed limitations of complex optimization strategies in hardware implementation, the system employs a complementary approach between the CDR fast controller and the slower-time-scale first optimization process. The CDR fast controller uses a fuzzy control strategy to directly adjust the parameters of the core loop filter of the fractional-order clock data recovery module to achieve adjustment capabilities. The first optimization process uses an intelligent algorithm (a first adaptive exploration and utilization strategy) to continuously optimize the characteristics of the CDR fast controller based on long-term performance indicators, improving the system's adaptability to implicit influencing factors such as long-term drift and aging effects.

[0044] Understandably, the stability and continuity of the clock are crucial in high-speed video transmission. An overly sensitive control system might introduce unnecessary disturbances due to large-scale, high-frequency adjustments to the fractional-order clock data recovery module's control vector, negatively impacting the smooth transmission of the video signal. Therefore, specifically, in this embodiment and some embodiments of the present invention, the CDR fast controller is constructed using a fuzzy control strategy, leveraging the good adaptability of fuzzy control to complex systems and its inherent smooth control characteristics. Fuzzy logic processes imprecise and fuzzy information based on empirical knowledge, making the parameter adjustment process of the CDR fast controller more gradual and robust when responding to loop dynamics. This avoids system oscillations or performance degradation caused by overly drastic or frequent control actions, thereby ensuring the quality and reliability of video transmission. In this embodiment and some embodiments of the present invention, the CDR fast controller is used to respond to the dynamic changes of the fractional-order clock data recovery module on a small timescale of microseconds to milliseconds, adjusting the proportional gain of the fractional-order loop filter (FO-LF) in real time through fuzzy logic. Integral coefficient Fractional order And a series of control parameters.

[0045] Specifically, in this embodiment and some embodiments of the present invention, the fuzzy input vector consists of two key state variables that characterize the current dynamic characteristics of the fractional-order clock data recovery module loop: instantaneous phase error. and its rate of change , represented as a vector : ; in, Provided directly by the phase / frequency detector (PD / PFD), Depend on Calculated using first-order backward differential (e.g., within a CDR fast controller), it reflects the trend of phase error variation: ； Specifically, in this embodiment and some embodiments of the present invention, the CDR fast controller adjusts the recovery control vector of the fractional clock data recovery module through fuzzification, fuzzy inference based on fuzzy rules, and defuzzification, as detailed below.

[0046] A. Blurring The process of blurring is to... Mapping to predefined fuzzy sets, the CDR fast controller determines the degree to which an input value belongs to each fuzzy set (i.e., its membership degree). Specifically, for any input value, the CDR fast controller calculates its membership degree with all fuzzy sets in parallel, thus determining the membership degree of the current input to the fuzzy sets. The CDR fast controller can be designed as a parallel pipelined hardware logic module to ensure operating speed. For each input, its five corresponding triangular membership functions are implemented by hardware-implemented fuzzification units. Each fuzzification unit includes a calculator and a comparator; the calculator is used to calculate the membership degree value of the fuzzy set corresponding to the fuzzification unit. The comparator is used to determine whether a corresponding component in the fuzzy input vector belongs to the fuzzy set based on the membership value. Specifically, in this embodiment, the input is divided into 5 fuzzy sets, which is for illustrative purposes only and not as a limitation.

[0047] Specifically, targeting Five fuzzy sets were defined: {Negative Large (NB), Negative Small (NS), Zero (ZE), Positive Small (PS), Positive Large (PB)}; [This is for...] Five fuzzy sets are also defined: {Negative Large (DNB), Negative Small (DNS), Zero (DZE), Positive Small (DPS), Positive Large (DPB)}.

[0048] The membership degree of each fuzzy set is described by a triangular membership function. The triangular membership function consists of three vertex parameters. definition:

[0049] and These are the points on the left and right bases of the triangle (with a membership degree of 0). These are the vertices of the triangle (with a membership degree of 1), representing typical values ​​or centers of the fuzzy set (these vertex parameters are optimized and configured by slow control to adjust the control behavior of the CDR fast controller).

[0050] B. Fuzzy reasoning After fuzzification, the fuzzy input vector is evaluated to determine its membership in a fuzzy set, followed by fuzzy inference. Fuzzy inference is the core decision-making process of the CDR fast controller, which determines the adjustment direction and intensity of the output based on a pre-set fuzzy rule base. The core of this fuzzy rule base is a set of "IF-THEN" fuzzy rules, totaling [number missing]. These rules comprehensively cover the input variables. and All combinations of fuzzy states, and specify the output variable for each case. , , Adjustments to control parameters, such as: Based on the current degree of membership of the input, each rule is adjusted as follows: IF is AND is THEN is AND is AND .

[0051] That is, if < ,and If > is true, then < for ,and for ,and for >;Among them, It represents the specific fuzzy set to which the input belongs. The output represents the set of corresponding output adjustment values. According to the definition of membership functions, there can be overlap between membership functions, and the input may belong to different fuzzy sets simultaneously. Therefore, multiple rules will be triggered simultaneously during the calculation process. Thus, it is necessary to calculate the trigger strength of multiple rules to obtain the final output. For the ... The degree to which a rule is activated by the current input, i.e., its trigger strength. To calculate:

[0052] in, It is input In fuzzy sets membership degree It is input In fuzzy sets The membership degree. In this embodiment, the fuzzy rule base is implemented as a large lookup table (LUT) – a fuzzy mapping table. The trigger strength of each rule. It is implemented by a set of parallel comparators.

[0053] C. Defuzzification Defuzzification is the process of synthesizing the outputs of all activated rules into a final, precise output vector of the fuzzy controller. The process of constructing a control adjustment vector from control adjustment values ​​involves merging the fuzzy suggestions from all activated rules into a single executable output value. Use a weighted average formula to synthesize each trigger intensity: ； in, They are the first In the rules The specified output value.

[0054] The recovery control vector calculated by the CDR fast controller The actual operating parameters used to update the fractional clock data recovery module can be achieved with just a multiplier and an adder. The update process is as follows: ； The parameters obtained from these dynamic calculations are directly substituted into the mathematical model of the fractional-order loop filter. This adjustment changes the filtering characteristics (gain and phase response) of the fractional-order clock data recovery module in real time, thereby altering its output. Ultimately, the main control signal is adopted. Alternatively, the feedforward compensation of the main control signal may drive the DCO, affecting its output frequency. and phase This adjustment mechanism forms a closed-loop adaptive control system where a CDR (Continuous Dynamic Controller) performs rapid adjustments. The adjustment criterion is the output of a fuzzy rule base, which is based on the current phase error. and error change rate This is derived from a comprehensive judgment of the fuzzy state. When both the error and the rate of change of error are large and in the same direction, it may be necessary to increase... To speed up the response; when the error is close to zero or overshoot occurs, it may be necessary to reduce the speed. To enhance stability. Fractional order. and The adjustment criteria are also based on the output of the fuzzy rule base. Changing the value of affects the phase margin and gain margin of the loop filter, thus impacting the system's dynamic performance and its ability to suppress jitter at specific frequencies. When the system requires stronger low-frequency jitter tracking capability, The characteristics of the filter can be changed. The CDR fast controller adjusts these two parameters in real time, and each control cycle (the first cycle, synchronized with the CDR loop's operating clock) is adjusted according to the latest... and The update is performed. The goal is to bring the instantaneous phase error to zero as quickly as possible and maintain stability.

[0055] Therefore, specifically, in this embodiment and some embodiments of the present invention, the fuzzification unit includes a first fuzzification unit corresponding to the phase error and a second fuzzification unit corresponding to the phase error change rate; fuzzy inference is performed on the fuzzy input vector based on the fuzzification unit and the fuzzy mapping table to update the recovery control vector of the fractional clock data recovery module, including the following processing: The first fuzzification unit is used to obtain the first The first membership degree of the ambiguity set to which the phase error of the first period belongs is obtained by using the second ambiguity unit. The second membership degree of the phase error change rate in the first cycle is used to obtain the first fuzzy set to which the phase error belongs and the second fuzzy set to which the phase error change rate belongs, based on the first membership degree. It is a natural number; Using the combination of the first fuzzy set and the second fuzzy set as an index to query the fuzzy rule, we obtain... The control adjustment vector and their respective weights , ; According to the recovery control vector in the... The value of the first cycle The control adjustment vector and the corresponding weights The recovery control vector is obtained at the 1st... The value of the first cycle .

[0056] Figure 2 This is a schematic diagram illustrating the adaptive capability of the CDR fast controller in the optimized control device of this embodiment. Figure 2 This demonstrates the ability to rapidly adapt to changes in input signal characteristics (such as jitter amplitude / frequency) or the introduction of strong interference into the channel. Specifically, for example... Figure 2 As shown, the horizontal axis represents time, and the vertical axis represents instantaneous phase error. The graph includes two stages, representing normal signal / channel conditions and deteriorated signal / channel conditions, respectively. The red curve represents a traditional CDR system, which uses a fixed or limited-range recovery control vector. When signal / channel conditions deteriorate, the phase error increases significantly, resulting in large peak values, persistent oscillations, or long recovery times. The blue curve represents a clock data recovery device with a CDR fast controller. The CDR fast controller adjusts the fractional-order loop filter parameters in real time, which can more effectively suppress phase error fluctuations when signal / channel conditions deteriorate. The peak error is smaller, oscillations are quickly suppressed, and the clock data recovery device can adapt to new conditions and recover to a better operating state more quickly. (Refer to...) Figure 2It is understandable that when dynamic signal characteristics change (temperature change) or severe interference occurs in the channel, the optimized control device of the present invention applied in the clock data recovery device can ensure the instantaneous performance and robustness of the system under complex dynamic input.

[0057] Specifically, in this embodiment and some embodiments of the present invention, unlike the CDR fast controller, the first optimization process implemented by the processor executing the first optimization program operates on a longer time scale. The first optimization process improves the exploration and utilization strategy of the fungal growth optimizer using a first adaptive exploration and utilization strategy (an improved fungal growth optimizer, IFGO), adjusting the key parameters of the fuzzy controller membership function and the weights of the fuzzy rules in the CDR fast controller to optimize the overall performance of the entire clock data recovery device under long-term operation. The parameters adjusted by the first optimization process are the fuzzy controller input variables (phase error) in the CDR fast controller. and error change rate This refers to a set of defined parameters for the membership function and the weights corresponding to each rule in the fuzzy rule base of the CDR fast controller. (The phase error is also mentioned.) and error change rate For example, they each have 5 fuzzy sets. Each fuzzy set's triangular membership function is defined by three vertex parameters. The first optimization strategy, x, seeks the optimal combination of membership function parameters by minimizing the long-term average bit error rate (BER). This means the first optimization process will try different membership function shapes and positions, as well as output adjustments. Each parameter has a corresponding value, and these control parameters to be optimized together constitute a complete set of control parameters (i.e., the recovery control vector). Specifically, in this embodiment, each recovery control vector... There are a total of 105 control parameters, or 105 dimensions, including the membership parameters of each fuzzy control set, totaling (3×5)×2=30. There are also 25 fuzzy rules, each corresponding to 3 parameters. There are a total of 75 parameters.

[0058] Specifically, in this embodiment and some embodiments of the present invention, the first optimization process employs a first adaptive exploration and utilization strategy to optimize the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table, including the following processes: Using the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table as spores, the initial value of the spore population is generated, and the following iterative processing is performed on the spore population: For the Round iteration generates the adaptive operator for this round. And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; like Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; If the new position First fitness is better than original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the first fitness of a spore is better than the best first fitness in the spore population, then the best spore in the spore population is updated to the first fitness. The spores described; The spore population completed the first... After a round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the highest fitness in the spore population is used to update the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The memory also stores a first neural network model, which the processor calls during the first optimization process to complete the first... The spores at position First fitness The prediction, the The spores at position First fitness Long-term bit error rate compared to the output of the first neural network model trained offline They are inversely proportional; the input to the first neural network model is the first... The location of the spores The output is the corresponding long-term bit error rate. .

[0059] Understandably, the first neural network model trained generally predicts the long-term bit error rate. The fitness is not equal to zero; therefore, in this embodiment and some embodiments of the present invention, the fitness is... The formula excludes the long-term bit error rate. The case where it equals zero.

[0060] It is understood that in other embodiments of the present invention, the first The spores at position fitness Calculate using the following formula: ;in, For a minimum value, such as 10 to the power of -15, the long-term bit error rate in this formula is... It can be equal to zero.

[0061] Specifically, in this embodiment and some embodiments of the present invention, the first neural network model adopts a fully connected multilayer perceptron (MLP) structure, which aims to establish a nonlinear mapping relationship between the membership parameters of the fuzzy computing unit and the fuzzy rules in the fuzzy mapping table and the system performance index (long-term bit error rate), so as to meet the fitting requirements of the complex nonlinear system response in the second cycle.

[0062] Specifically, in this embodiment and some embodiments of the present invention, the number of input layer nodes of the first neural network model is set to... D ,in D The dimension of the spore is given. The output layer of the neural network model contains one neuron, which outputs the predicted long-term bit error rate. Between the input and output layers, there are three hidden layers, each with 0.5 to 2 times the number of nodes in the input layer (e.g., 64 to 256 nodes). The layers are fully connected, and a non-linear activation function (Sigmoid function) is introduced.

[0063] Specifically, to achieve offline training of the first neural network model, this embodiment and some embodiments of the present invention construct a system simulation platform containing a core circuit model and a channel model of the clock recovery system. In the simulation platform, multiple sets of spores (membership parameters of fuzzy computation units and fuzzy rules in the fuzzy mapping table) are randomly generated; the simulation is run and the long-term bit error rate within a corresponding preset second period (e.g., 1 day, 15 days, 1 month, 1 year, etc.) is recorded, thereby constructing a sample dataset. Further, the sample dataset is divided into a training set and a validation set. Backpropagation and gradient descent optimization are used, with the goal of minimizing the mean square error between the predicted and actual bit error rates, iteratively updating the network weights and biases until the error on the validation set converges to a preset threshold. Through the above settings, the first neural network model in this embodiment of the present invention can internalize the dynamic characteristics of the clock recovery system, thereby replacing circuit simulation in the optimization process, achieving fitness evaluation, and making the optimization process more efficient and feasible.

[0064] More specifically, in this embodiment and some embodiments of the present invention, the transmission quality of the serial data is monitored according to a preset second cycle. In response to a decrease in transmission quality, the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table are optimized using a first adaptive exploration and utilization strategy. Specifically, this includes processes S100a1 to S100a8.

[0065] Process S100a1 according to the size of the spore population. and the dimensions of the spores Generate size is A chaotic matrix, wherein the columns of the chaotic matrix correspond to the dimensions of the spores, and the rows of the chaotic matrix correspond to the spores in the spore population.

[0066] Specifically, a chaotic matrix is ​​generated using a Logistic chaotic mapping, and the iterative formula for each row of the chaotic matrix is ​​as follows: , , , ;in, It is the first chaotic matrix Line number The elements of the column.

[0067] In processing S100a2, the elements of each column vector of the chaotic matrix are mapped to the value space of the corresponding dimension of the recovery control vector to obtain the values ​​of each dimension of the spore. After the mapping is completed, a total of The spores mentioned above The number of spores is the initial value of the spore population.

[0068] Specifically, the first spores The Dimensional parameters By generating chaotic matrix elements Initialize by mapping to the above value range:

[0069] in, and It is the first of the recovery control vectors Upper and lower bounds of the dimension parameter.

[0070] Processing S100a3, iteration rounds , Process S100a4 and calculate the current round. The value of the adaptive operator : , in, It is the first The value of the adaptive operator in the next iteration. and These are the upper and lower limits of the adaptive operator's values.

[0071] In process S100a5, a random number is generated for each spore in the spore population, wherein the first... The random number corresponding to each spore is ,like Then the first The first spore executes the global exploration strategy S100a6; otherwise, the second spore... Each spore is subjected to treatment S100a7.

[0072] Processing S100a6, the first Each spore executes a global exploration strategy: ; in, It refers to the current position of three spores randomly selected from the spore population. It is the current position of the spore with the highest fitness in the spore population. It is an exploration factor that decreases with each iteration, a step size scaling factor, and a random scalar step size. Based on two random variables that follow a normal distribution and The generation and calculation formula is: ; in, Follows a mean of 0 and a standard deviation of The normal distribution It follows a standard normal distribution (mean 0, standard deviation 1). The calculation formula is as follows: ; in, The value is 1.6. It is the Gamma function.

[0073] Processing S100a7, the first Individual spores employ a localized utilization strategy: ; It is a utilization factor that decreases with each iteration. It is the current position of a spore randomly selected from the spore population.

[0074] Processing S100a8, if the new position The fitness is better than that of the original position. The fitness of the aforementioned, then the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the fitness of a spore is better than the optimal fitness in the spore population, then the optimal spore in the spore population is updated to the spore with the best fitness. The spores mentioned.

[0075] Among them, the The spores at position First fitness Calculate using the following formula: ; Among them, among them, This represents the first neural network model trained offline, and the input of the first neural network model is the first... The location of the spores The output is the corresponding long-term bit error rate. .

[0076] Process S100a9 and determine whether the iteration termination condition is met, such as reaching the maximum number of iterations or the rate of change of the optimal fitness being less than a preset threshold. If so, output the spore with the highest fitness in the spore population as the value of the target recovery control vector and send it to the fuzzy controller in the CDR fast controller. Otherwise, return to process S100a2.

[0077] Figure 3 This is a schematic diagram of the first optimization processing adaptive capability implemented in the optimized control device of this embodiment. For example... Figure 3 As shown, the first optimization process is executed according to a preset cycle to cope with implicit factors such as device aging or slow environmental temperature drift encountered by the system, maintaining the system's high performance. The horizontal axis represents multiple working cycles over a long period of time, and the vertical axis represents a key long-term performance indicator, the average bit error rate (BER). The red curve represents a system without slow control optimization, whose performance gradually deteriorates over time (due to accumulated parameter drift), manifested as a gradual increase in the BER. The blue curve represents a clock data recovery device that implements the first optimization process. The first optimization process actively combats the influence of implicit factors by periodically adjusting the membership function parameters of the CDR fast controller, enabling the system's long-term performance to remain at a relatively stable level. Although there may be minor fluctuations caused by executing the first optimization program to achieve the first optimization process, the first optimization process makes an indispensable contribution to ensuring the long-term stability and reliability of the system.

[0078] Figure 4This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to another embodiment of the present invention. Figure 4 As shown, the CDR optimization control device for the video transmission network in this embodiment of the present invention, in Figure 1 The device shown also includes a temperature sensor for acquiring temperature and an interference sensor for acquiring interference frequency and interference amplitude.

[0079] also, Figure 4 In the illustrated embodiment, a second optimization program for the CDR fast controller is stored in the memory, and when the processor executes the second optimization program, it performs the following second optimization process: If changes in environmental characteristics are detected and continue to exceed a preset time threshold, a second adaptive exploration and utilization strategy is adopted to optimize the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The environmental characteristics include temperature, interference frequency, and interference amplitude.

[0080] Understandably, the second optimization process is an adaptive control when the transmission environment changes significantly. When a clear and continuous change in environmental characteristics (such as temperature, interference amplitude, interference frequency variation amplitude and time exceeding the threshold) is detected, the second adaptive exploration and utilization strategy (another improved fungal optimization strategy, Improved Fungal Growth Optimizer, IFGO) is activated to find a suitable set of recovery control vectors for the new environmental conditions.

[0081] Specifically, in this embodiment and some embodiments of the present invention, when environmental characteristics change significantly or are subjected to severe interference, such as when the changes in temperature and interference exceed a preset change threshold and the duration exceeds a preset time threshold, the processor executes a second optimization program stored in the memory to achieve a second optimization process. The goal is to ensure that the minimum long-term average bit error rate (BER) is obtained under the constraints of the changed environmental characteristics (temperature, interference factors, etc.).

[0082] Specifically, this includes processing S200a to S200b.

[0083] In processing S200a, the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table are used as spores to generate the initial value of the spore population, based on the number of pre-review iteration stages. Determine the first Each iteration phase The value of the adaptive operator is , , Number Greater than 2 and less than the maximum number of iterations; For process S200b, perform iterative processing on the spore population as described in processes S200b1 to S200b4: Processing S200b1, for the first Round iteration, And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; If S200b2 is processed, Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; Process S200b3, if the new position The second fitness is better than the original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the second fitness of a spore is better than the optimal second fitness in the spore population, then the optimal spore in the spore population is updated to the second fitness. The spores described; Processing S200b4, the spore population completes the first... After one round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the second highest fitness in the spore population is used to update the recovery control vector of the fuzzy inference module.

[0084] Specifically, in this embodiment and some embodiments of the present invention, the long-term bit error rate output by the offline-trained second neural network model is compared with... They are inversely proportional. More specifically, calculate using the following formula: ; in, This represents the second neural network model trained offline, whose input is the first... The location of the spores The temperature The interference amplitude and the interference frequency The output is the corresponding long-term bit error rate. .

[0085] Understandably, in other embodiments of this invention, the second fitness is calculated according to the following formula. ), This represents the minimum value.

[0086] Specifically, in the embodiments and some embodiments of the present invention, the second neural network model page adopts a fully connected multilayer perceptron (MLP) structure, which aims to establish a nonlinear mapping relationship between the recovery control vector and the environmental state to the system performance index (long-term bit error rate) in order to meet the fitting requirements of complex nonlinear system response.

[0087] Specifically, in this embodiment and some embodiments of the present invention, the number of input layer nodes of the second neural network model is set to... D +3, of which D The dimension of the spore is 3 (e.g., 105 dimensions in this embodiment), and the number 3 corresponds to three environmental features: temperature, interference amplitude, and interference frequency. The output layer of the neural network model contains one neuron, which outputs the predicted long-term bit error rate. Between the input and output layers, there are three hidden layers, each with 0.5 to 2 times the number of nodes in the input layer (e.g., 64 to 256 nodes). The layers are fully connected, and a non-linear activation function (Sigmoid function) is introduced.

[0088] Specifically, to achieve offline training of the second neural network model, this embodiment and some embodiments of the present invention construct a system simulation platform containing a core circuit model and a channel model of the clock recovery system. In the simulation platform, random combinations of temperature, interference amplitude, and frequency covering the operating range are set, and multiple sets of spores (membership parameters of fuzzy computation units and fuzzy rules in the fuzzy mapping table) are randomly generated; the simulation is run and the corresponding long-term bit error rate is recorded, thereby constructing a sample dataset. Further, the sample dataset is divided into a training set and a validation set. Backpropagation and gradient descent optimization are used, with the goal of minimizing the mean square error between the predicted long-term bit error rate and the actual long-term bit error rate, iteratively updating the network weights and biases until the error on the validation set converges to a preset threshold. Through the above settings, the second neural network model in this embodiment of the present invention can internalize the dynamic characteristics of the clock system, thereby replacing circuit simulation in the optimization process, realizing fitness evaluation, and making the optimization process more efficient and feasible.

[0089] Figure 5 This is a schematic diagram of the main modules of a CDR optimization control device for a video transmission network according to another embodiment of the present invention. Figure 5 As shown, the CDR optimization control device for the video transmission network in this embodiment of the present invention, relative to... Figure 1 and Figure 4 The device has a memory storing a temperature feedforward compensation program for the CDR fast controller. When the processor executes the temperature feedforward compensation program, it performs the following temperature compensation processing: Monitoring ambient temperature, in response to a detected abrupt change in ambient temperature, a temperature-dependent perturbation effect model is employed, based on the current temperature. Predict the output offset of the fractional clock data recovery module According to the output offset and the current frequency of the fractional clock data recovery module Adjust the feedforward frequency of the local clock in the fractional clock data recovery module.

[0090] Understandably, in the temperature compensation process of this embodiment, open-loop feedforward compensation is performed for disturbances such as temperature abrupt changes that are fast, large in magnitude, direct and predictable, in order to offset their direct impact on the DCO output frequency with almost no delay, and to avoid them generating large instantaneous errors in the feedback loop.

[0091] Specifically, such as Figure 5 As shown, this embodiment and some embodiments of the present invention also employ temperature compensation processing to proactively and rapidly compensate for the direct impact of significant environmental disturbances on the fractional-order clock data recovery module, thereby reducing the burden on the feedback control loop. In harsh environments such as automotive applications, transient disturbances such as drastic temperature changes can cause rapid and large drifts. Traditional feedback control systems respond slowly to such sudden changes and may lead to large instantaneous errors. To address this issue and meet the transient robustness requirements of high-speed video transmission, this embodiment executes a temperature feedforward compensation program in the processor to introduce temperature compensation processing, aiming to achieve proactive and rapid disturbance cancellation. The optimized control device monitors the ambient temperature in real time. (From temperature sensor), and based on a temperature-dependent perturbation effect model, predict the impact of the current temperature The expected fractional clock data recovery module output offset This embodiment uses a second-order polynomial function as the model for the impact of this disturbance:

[0092] Among them, model coefficients These are fixed values ​​determined through a one-time offline characterization and calibration process. This process is described in detail below.

[0093] First, characterize the DCO: During the design verification or production testing phase of the DCO, place it in a temperature-controlled environment. Without closing the CDR main feedback loop or under a specific preset test mode, measure and record the actual output frequency or deviation from the nominal frequency of the DCO at multiple different ambient temperature points (covering the entire operating temperature range, which is typically -40 to +125°C in automotive applications). This will generate a raw dataset containing data pairs of (ambient temperature, DCO frequency deviation).

[0094] Secondly, model coefficient fitting is performed: standard curve fitting techniques are used, specifically least squares polynomial regression, to fit the second-order polynomial model to the collected original dataset. This fitting process calculates the coefficient set that best describes the DCO temperature drift characteristics. .

[0095] Finally, the coefficients are stored: the obtained set of coefficient values ​​is burned as fixed parameters or stored in the non-volatile memory of the optimization control device for use by the temperature feedforward compensation program during operation.

[0096] Specifically, in this embodiment and some embodiments of the present invention, the predicted output frequency after temperature perturbation... .

[0097] The fractional clock data recovery module requires compensation control signals to eliminate the influence of external interference. as follows:

[0098] In this way, the relative change in DCO frequency caused by temperature is adjusted. This offsets the fluctuations, preventing them from becoming drastic due to temperature shifts.

[0099] Figure 6 This is a schematic diagram illustrating the peak suppression performance gain of the temperature compensation processing implemented in the optimized control device according to another embodiment of the present invention when dealing with instantaneous temperature pulse disturbances. Specifically, as shown... Figure 6As shown, a comparison was made between the two cases with and without temperature compensation. The instantaneous phase error curves for the same instantaneous, short-duration temperature pulse disturbance (e.g., a rapid rise followed by a rapid fall in temperature) clearly demonstrate the effectiveness of temperature compensation in suppressing transient spike disturbances. The red curve shows the system response relying solely on the fractional clock data recovery (FOCDR) module and dual-time-scale feedback control (CDR fast controller + first optimization processing control) (without temperature compensation): during the temperature pulse, a significant instantaneous spike in phase error occurs, which the feedback control system then attempts to correct back to the baseline level. The blue curve shows the response of the complete optimization control unit (CDR fast controller + first optimization processing control + temperature compensation): because the temperature compensation process pre-adjusts the compensation control signal according to the rapid temperature change. The instantaneous spike impact caused by temperature pulses on phase errors is significantly weakened or almost eliminated, and the optimized control device exhibits stronger disturbance suppression capability. This demonstrates the crucial role of temperature compensation processing in rapidly suppressing and weakening predictable, short-term significant disturbances (especially temperature-related rapid transients), thereby effectively protecting the stability of clock data recovery.

[0100] Furthermore, it is understood that in this embodiment and some embodiments of the present invention, the temperature compensation process chooses to directly adjust the control signal of the DCO, rather than the loop filter parameters, because changing the control signal of the DCO can most quickly and directly correct the clock frequency, thereby rapidly offsetting the direct impact of temperature and other disturbances on the DCO frequency and quickly reducing phase error. At the same time, this adjustment of the DCO control signal does not change the filtering characteristics of the FO-LF. Therefore, the CDR fast controller, the first optimization process, and the second optimization process in this embodiment do not conflict with the temperature compensation process, ensuring the synergy and effectiveness of the overall control strategy of the optimized control device.

[0101] Based on the above detailed description of specific embodiments of the present invention, a clearer understanding of the present invention provides the following advantages: This invention provides a processing architecture for a CDR (Continuous Delayed Response) optimization control device in a video transmission network. On one hand, the CDR fast controller within the device enables real-time response to phase error changes on a microsecond to millisecond timescale, dynamically adjusting the local clock of the fractional-order loop filter module. On the other hand, a processor executes a first optimization program stored in memory to perform first optimization processing, continuously optimizing the membership parameters and fuzzy rules of the CDR fast controller in the background. This processing architecture avoids control conflicts, enabling the device of this invention to not only cope with instantaneous disturbances but also adapt to long-term drift, aging, and gradual environmental changes, achieving continuous performance optimization.

[0102] The device in this embodiment of the invention monitors changes in transmission quality. On the one hand, it employs a first optimization program with a first adaptive exploration and utilization strategy to periodically optimize the membership parameters and fuzzy rules of the CDR fast controller for implicit influencing factors of transmission quality. On the other hand, it employs a second optimization program with a second adaptive exploration and utilization strategy to optimize the membership parameters and fuzzy rules of the CDR fast controller for explicit influencing factors of transmission quality. This comprehensively enhances the adaptive capability of clock data recovery in video transmission networks to various influencing factors, ensuring the stability of long-term clock data recovery performance.

[0103] The device in this embodiment of the invention integrates sensing modules such as temperature sensors and interference sensors, enabling it to collect physical quantities such as ambient temperature, interference frequency and amplitude in real time. These hardware sensing signals are directly input to the second optimization program and temperature feedforward compensation program executed by the processor. After sensing changes in environmental characteristics, it can achieve adaptive dynamic adjustment and direct compensation of the fractional clock data recovery module. This closed-loop hardware architecture of "perception-decision-execution" gives the device in this embodiment of the invention an inherent ability to resist environmental disturbances.

[0104] The optimized control device of this invention adopts a modular design, which facilitates hardware and software integration with the fractional clock data recovery module. It can meet the stringent real-time requirements of high-speed video transmission while ensuring high performance, and is suitable for widespread application in embedded systems such as automotive electronics and industrial vision.

[0105] 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 changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A CDR optimization control device for a video transmission network, characterized in that, Includes a CDR fast controller, memory, and processor; among which, The CDR fast controller is used to receive the error between the reference clock output by the fractional clock data recovery module and the local clock, construct a fuzzy input vector according to the error according to a preset first cycle, perform fuzzy inference on the fuzzy input vector based on the fuzzification unit and the fuzzy mapping table, and update the recovery control vector of the fractional clock data recovery module; the reference clock is the reference clock in the video data received from the video transmission network. The memory stores a first optimization program for the CDR fast controller. When the processor executes the first optimization program, it performs the following first optimization process: The transmission quality of the video data is monitored according to a preset second cycle. In response to a decrease in transmission quality, the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table are optimized using a first adaptive exploration and utilization strategy. The second cycle is longer than the first cycle.

2. The apparatus according to claim 1, characterized in that, The fuzzy input vector includes the phase error and the rate of change of the phase error in the first period; the fuzzification unit includes a first fuzzification unit corresponding to the phase error and a second fuzzification unit corresponding to the rate of change of the phase error; Based on the fuzzification unit and fuzzy mapping table, fuzzy inference is performed on the fuzzy input vector to update the recovery control vector of the fractional clock data recovery module, including: The first fuzzification unit is used to obtain the first The first membership degree of the ambiguity set to which the phase error of the first period belongs is obtained by using the second ambiguity unit. The second membership degree of the phase error change rate in the first cycle is used to obtain the first fuzzy set to which the phase error belongs and the second fuzzy set to which the phase error change rate belongs, based on the first membership degree. It is a natural number; Using the combination of the first fuzzy set and the second fuzzy set as an index to query the fuzzy rule, we obtain... The control adjustment vector and their respective weights , ; According to the recovery control vector in the... The value of the first cycle The control adjustment vector and the corresponding weights The recovery control vector is obtained at the 1st... The value of the first cycle .

3. The apparatus according to claim 2, characterized in that, The fuzzification unit includes a calculator and a comparator; the calculator is used to calculate the membership value of the fuzzy set corresponding to the fuzzification unit, and the comparator is used to determine whether the corresponding component in the fuzzy input vector belongs to the fuzzy set based on the membership value.

4. The apparatus according to claim 1, characterized in that, The first optimization process employs a first adaptive exploration and utilization strategy to optimize the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table, including: Using the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table as spores, the initial value of the spore population is generated, and the following iterative processing is performed on the spore population: For the Round iteration generates the adaptive operator for this round. And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; like Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; If the new position First fitness is better than original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the first fitness of a spore is better than the best first fitness in the spore population, then the best spore in the spore population is updated to the first fitness. The spores described; The spore population completed the first... After a round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the highest fitness in the spore population is used to update the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The memory also stores a first neural network model, which the processor calls during the first optimization process to complete the first... The spores at position First fitness The prediction, the The spores at position First fitness Long-term bit error rate compared to the output of the first neural network model trained offline They are inversely proportional; the input to the first neural network model is the first... The location of the spores The output is the corresponding long-term bit error rate. .

5. The apparatus according to claim 4, characterized in that, Using the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table as spores, the initial values ​​of the spore population are generated, including: Based on the size of the spore population and the dimensions of the spores Generate size is A chaotic matrix, wherein the columns of the chaotic matrix correspond to the dimensions of the spores, and the rows of the chaotic matrix correspond to the spores in the spore population; The elements of each column vector of the chaotic matrix are mapped to the value space of the corresponding dimension of the recovery control vector to obtain the values ​​of each dimension of the spore. After the mapping is completed, a total of The spores mentioned above The number of spores is the initial value of the spore population; No. The adaptive operator in the round iteration follows the formula as the number of iterations increases. Adjustment: ,in, and These are the upper and lower limits of the adaptive operator's values; The global exploration strategy is: The local utilization strategy is: ; in, It refers to the current position of three spores randomly selected from the spore population. It is the current position of the spore with the highest fitness in the spore population; It is an exploratory factor. It is the step size scaling factor. It is a random scalar step size generated based on the Levy distribution; It is a factor of utilization. It is the current position of a spore randomly selected from the spore population.

6. The apparatus according to claim 1, characterized in that, It also includes a temperature sensor for acquiring temperature, and an interference sensor for acquiring interference frequency and interference amplitude; The memory stores a second optimization program for the CDR fast controller. When the processor executes the second optimization program, it performs the following second optimization process: If changes in environmental characteristics are detected and continue to exceed a preset time threshold, a second adaptive exploration and utilization strategy is adopted to optimize the membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table. The environmental characteristics include temperature, interference frequency, and interference amplitude.

7. The apparatus according to claim 6, characterized in that, The second optimization process employs a second adaptive exploration and utilization strategy to optimize the membership parameters of the fuzzification calculation unit in the CDR fast controller and the fuzzy rules in the fuzzy mapping table, including: The membership parameters of the fuzzification calculation unit and the fuzzy rules in the fuzzy mapping table are used as spores to generate the initial value of the spore population, based on the number of pre-approval iteration stages. Determine the first Each iteration phase The value of the adaptive operator is , , Number Greater than 2 and less than the maximum number of iterations; The following iterative process was performed on the spore population: For the Round iteration, And generate a random number for each spore in the spore population, wherein, For natural numbers, the first The random number corresponding to each spore is ; like Then the first The first spore executes a global exploration strategy; otherwise, the second spore... The spores employ a localized utilization strategy; wherein, in the spore population, the spores of the th... The spores mentioned above start from the beginning position of the current round. New locations after exploration or utilization ; If the new position The second fitness is better than the original position The fitness of the first The position of each spore is updated to the new position. Otherwise, retain the original position. Furthermore, if the new position If the second fitness of a spore is better than the optimal second fitness in the spore population, then the optimal spore in the spore population is updated to the second fitness. The spores described; The spore population completed the first... After one round of exploration or utilization, it is determined whether the iteration termination condition is met; if not, the next round of iteration continues; otherwise, the spore with the second highest fitness in the spore population is used to update the recovery control vector of the fuzzy inference module. The memory also stores a first neural network model, which the processor calls during the first optimization process to complete the first... The spores at position Second fitness Prediction, the first The spores at position Second fitness Long-term bit error rate compared to the output of the offline-trained second neural network model They are inversely proportional; the input to the second neural network model is the first... The location of the spores The temperature The interference amplitude and the interference frequency The output is the corresponding long-term bit error rate. .

8. The apparatus according to claim 6, characterized in that, The memory stores the temperature feedforward compensation program of the CDR fast controller. When the processor executes the temperature feedforward compensation program, it performs the following temperature compensation process: Monitoring ambient temperature, in response to a detected abrupt change in ambient temperature, a temperature-dependent perturbation effect model is employed, based on the current temperature. Predict the output offset of the fractional clock data recovery module According to the output offset and the current frequency of the fractional clock data recovery module Temperature compensation is performed on the local clock in the fractional clock data recovery module.

9. A clock data recovery device for a video transmission network, characterized in that, It includes a fractional clock data recovery module and a CDR optimization control device based on the video transmission network according to any one of claims 1-8.

10. The apparatus according to claim 9, characterized in that, in, The fractional clock data recovery module includes a digital fractional loop filter and a numerically controlled oscillator. The recovery control vector of the fractional clock data recovery module includes the proportional gain, integral gain, and fractional order of the digital fractional loop filter. The fractional-order clock data recovery module adjusts the frequency of the local clock using the error between the reference clock and the local clock, and the recovery control vector, including: According to the digital fractional-order loop filter in the first... The proportional gain of the first cycle Integral gain and fractional order The coefficients of the first difference equation of the digital fractional-order loop filter are determined by looking up a table. coefficients of the second difference equation ; Based on the aforementioned error and the coefficients of the first difference equation coefficients of the second difference equation The main control signal of the numerically controlled oscillator is obtained according to the following formula. : ； According to the main control signal Or based on the main control signal After feedforward compensation, the frequency of the output signal of the numerically controlled oscillator is adjusted. .

Images