Method and device for adjusting adaptive weight for equalizer, and equalization system
By dynamically configuring the adaptive weights, using the tap coefficient difference values of different equalizers to optimize the convergence speed and steady-state error, the problem of difficulty in optimization in the prior art is solved, and a more efficient and stable adaptive equalizer is achieved.
Patent Information
- Application Number
- CN202411899344.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-23
- Publication Date
- 2025-04-25
AI Technical Summary
In the prior art, it is difficult for the adaptive equalizer with constant weight to optimize the convergence speed and steady-state error at the same time. Larger adaptive weights can speed up the convergence speed but may lead to an increase in the steady-state error. Conversely, smaller adaptive weights can reduce the steady-state error but slow down the convergence speed.
By extracting the tap coefficients of different equalizers and comparing their differences to dynamically configure the adaptive weights, so that the tap coefficients of the equalizer can adaptively converge according to the working state, optimizing the convergence speed and steady-state error.
The convergence speed and steady-state error of the adaptive equalizer are achieved simultaneously. By dynamically adjusting the adaptive weight, the working state of the equalizer is automatically adapted to the system stability and efficiency.
Smart Images
Figure CN119892558A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data processing, and in particular to a method and device for adjusting adaptive weights of an equalizer, and an equalizing system. Background Art
[0002] In serial transmission systems (such as Ethernet, etc.), adaptive equalizers (such as forward feedback equalizers and decision feedback equalizers) are generally used to overcome signal distortion and inter-symbol interference during transmission. When data is transmitted in a cable, the dispersion characteristics of the cable may cause signal distortion, which varies with the signal spectrum and the length of the medium loop. In order to ensure the integrity and accuracy of the signal, the received signal must be compensated. The adaptive equalizer can adjust its coefficients at any time based on the measurement of channel characteristics to adapt to changes in these characteristics, thereby effectively eliminating inter-symbol interference.
[0003] An adaptive equalizer based on the LMS (Least Mean Square) algorithm is an effective solution. The LMS algorithm is designed based on the mean square error criterion to minimize the mean square estimation error. This algorithm iteratively adjusts the tap coefficients of the equalizer, i.e., the filter weights, to achieve the best compensation for the received signal. According to the LMS algorithm, the input signal, the error signal, and an adaptive weight (convergence factor) are used to update the weights of the equalizer.
[0004] The adaptive weight is a key factor in the LMS algorithm, which determines the algorithm convergence speed and steady-state error. In the design of a constant-weight adaptive equalizer, it is impossible to optimize both the convergence speed and the steady-state error. A larger adaptive weight can speed up the convergence speed, but may lead to a larger steady-state error. A smaller adaptive weight can reduce the steady-state error, but the convergence speed will be slower. If the adaptive weight is too large, even when the algorithm is close to convergence, it may cause the weight update to be too large, causing the system to oscillate or even diverge. If the adaptive weight is too small, it may cause the algorithm to require more iterations to converge, thus wasting computing resources. Summary of the invention
[0005] The present invention provides a method, a device and an equalization system for adjusting adaptive weights of an equalizer, which can simultaneously optimize the convergence speed and the steady-state error of the adaptive equalizer.
[0006] In one aspect of the present invention, a method for adjusting adaptive weights for an equalizer is provided. The method comprises: extracting a first tap coefficient of a first equalizer and a second tap coefficient of a second equalizer; dynamically configuring the adaptive weights according to a comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weights until the first tap coefficient and the second tap coefficient tend to be stable.
[0007] In another aspect of the present invention, a device for adjusting adaptive weights for an equalizer is provided. The device includes: an extraction module configured to extract a first tap coefficient of a first equalizer and a second tap coefficient of a second equalizer; and a weight configuration module configured to dynamically configure the adaptive weights according to a comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weights until the first tap coefficient and the second tap coefficient tend to be stable.
[0008] In another aspect of the present invention, an equalizing system is provided, which includes the above-mentioned device for adjusting adaptive weights for an equalizer, and a first equalizer and a second equalizer electrically coupled to the device.
[0009] According to the technical solution of the present invention, the tap coefficients of different equalizers are obtained to determine the working state of the current adaptive equalizer, the adaptive weights are dynamically configured according to the comparison results between the tap coefficients of different equalizers, and the tap coefficients are adaptively converged based on the configured adaptive weights, so that the tap coefficients of the equalizer can be dynamically adjusted according to the working state of the equalizer. Compared with the method of fixing the adaptive weights, the adaptive weights of the present invention can automatically adapt to the working state of the equalizer, and the convergence speed and steady-state error of the adaptive equalizer can be optimized at the same time by dynamically configuring the adaptive weights. BRIEF DESCRIPTION OF THE DRAWINGS
[0010] Figure 1 A flow chart of a method for adjusting adaptive weights for an equalizer according to an embodiment of the present invention; Figure 2 A schematic diagram of the change of the difference and the adaptive weight according to an embodiment of the present invention; Figure 3 A schematic diagram of changes in iterative updating of tap coefficients according to an embodiment of the present invention; Figure 4 A block diagram of a device for adjusting adaptive weights for an equalizer according to an embodiment of the present invention; Figure 5 FIG. 4 is an architecture diagram of a balancing system according to an embodiment of the present invention. DETAILED DESCRIPTION
[0011] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.
[0012] In the prior art, in the design of a constant-weight adaptive equalizer, it is impossible to optimize both the convergence speed and the steady-state error at the same time. A larger step size can speed up the convergence speed, but may lead to a larger steady-state error. A smaller step size can reduce the steady-state error, but the convergence speed will be slower. If the step size is too large, even when the algorithm is close to convergence, it may cause the weight update to be too large, causing the system to oscillate or even diverge. If the step size is too small, it may cause the algorithm to require more iterations to converge, thus wasting computing resources.
[0013] In order to solve at least the above technical problems, the present disclosure provides a method for adjusting adaptive weights for an equalizer. According to the present disclosure, tap coefficients of different equalizers are obtained to determine the current working state of the adaptive equalizer, adaptive weights are dynamically configured according to the difference between the tap coefficients of different equalizers, and the tap coefficients are adaptively converged based on the configured adaptive weights, so that the tap coefficients of the equalizer can be dynamically adjusted according to the working state of the equalizer. In this way, compared with the method of fixing adaptive weights, according to an embodiment of the present disclosure, the adaptive weights can automatically adapt to the working state of the equalizer, and the convergence speed and steady-state error of the adaptive equalizer can be optimized simultaneously by dynamically configuring the adaptive weights.
[0014] The adaptive weight in the present invention is the weight used by the equalizer to update the tap coefficient, and the tap coefficient is the weight used by the equalizer to adjust the signal strength. According to an embodiment of the present disclosure, the adaptive weight of the equalizer is adjusted based on the difference between the tap coefficients of the same level of the forward feedback equalizer and the decision feedback equalizer. The first tap coefficient of the forward feedback equalizer and the second tap coefficient of the decision feedback equalizer are extracted. As the adaptive process of the equalizer proceeds, the first tap coefficient and the second tap coefficient should tend to be stable, and ideally the first tap coefficient and the second tap coefficient tend to be equal. Therefore, by calculating the difference between the first tap coefficient and the second tap coefficient, the stability of the digital equalization system is judged, and then the difference between the tap coefficients is used as the judgment logic to adjust the size of the adaptive weight, thereby realizing the use of different adaptive weights according to the stable state of the equalizer.
[0015] Hereinafter, the technical solution according to the present disclosure will be described with reference to specific embodiments and in conjunction with the accompanying drawings.
[0016] Figure 1 1 is a flow chart showing a method 100 for adjusting adaptive weights for an equalizer according to an embodiment of the present disclosure. Figure 1The method 100 includes the following steps 102 to 104.
[0017] In step 102, a first tap coefficient of a first equalizer and a second tap coefficient of a second equalizer are extracted.
[0018] In a digital signal processing filter, each tap can be considered as an access cycle point in the signal stream. For example, if the current digital equalization system consists of 6 working levels of taps, the digital equalization system can eliminate the tail interference of 6 cycles of the signal. Each level of tap is configured with a tap coefficient, which is used to adjust the signal strength. The adaptive equalizer can train the tap coefficients that can overcome inter-symbol interference according to the current cable transmission conditions.
[0019] Specifically, the tap coefficients can be adaptively generated by the LMS (least mean square error) algorithm. The LMS algorithm is specifically: W(n+1)=W(n)+μ×e(n)×X(n), where W(n+1) represents the tap coefficients after iterative update, W(n) represents the tap coefficients before iterative update, μ represents the weight used to update the tap coefficients or the adaptive weights described in this article, e(n) represents the decision error between the desired signal and the actual signal, and X(n) represents the digital signal input to the equalizer. The adaptive process of the tap coefficient W is essentially an integration process, so in the adaptive process, the trend of the decision error e is: the absolute value of e changes from large to small, and when the decision error e tends to 0, the tap coefficient W tends to a stable value, and the tap coefficient W can be considered to be in a convergence state.
[0020] In step 104, adaptive weights are dynamically configured based on a comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weights until the first tap coefficient and the second tap coefficient tend to be stable.
[0021] In some embodiments, dynamically configuring the adaptive weight according to the comparison result between the first tap coefficient and the second tap coefficient includes: obtaining a preset coefficient compensation value, and compensating the difference between the first tap coefficient and the second tap coefficient according to the coefficient compensation value, so as to dynamically set the adaptive weight according to the compensated difference. In this way, the first equalizer is used to process the forward interference of the signal, and the second equalizer is used to process the backward interference of the signal. When the forward interference and the backward interference of the signal are not equal, the tap coefficients of the two equalizers are in an unequal state even in the convergence state. Therefore, the difference between the two tap coefficients is compensated by the preset coefficient compensation value to eliminate the difference between the two coefficients caused by the difference in forward and backward interference, thereby improving the compatibility of the equalizer with different signal processing.
[0022] In some embodiments, the coefficient compensation value is a fixed value and does not change with the tap coefficient. In some specific scenarios, if the front and rear tail interferences of the signal are consistent, the coefficient compensation value can be set to 0, and if the front and rear tail interferences of the signal are inconsistent, the coefficient compensation value can be set to the middle value of the signal difference.
[0023] In some embodiments, the first tap coefficient and the second tap coefficient are used to process the same periodic signal. Dynamically configuring the adaptive weight according to the comparison result between the first tap coefficient and the second tap coefficient includes: selecting the adaptive weight corresponding to the difference between the first tap coefficient and the second tap coefficient. In this way, under ideal conditions, the tap coefficients for processing the same periodic signal under different equalizers should tend to be equal when tending to a convergence state. Therefore, the stability of the current equalizer can be determined by the difference in tap coefficients between the two equalizers, so as to dynamically configure the appropriate adaptive weight according to the stability of the equalizer.
[0024] The convergence state of the tap coefficient represents the convergence state of the digital equalization system. When the tap coefficient tends to be stable, the digital equalization system tends to work stably. According to the present disclosure, since the first tap coefficient and the second tap coefficient that tend to be stable should tend to be equal under ideal conditions, the difference in tap coefficients between the forward feedback equalizer and the decision feedback equalizer can feedback the convergence degree of the current digital equalization system. When the digital equalization system is stable, the difference in tap coefficients should tend to 0. Since the stable value of the tap coefficient tends to be stable will vary with the length of the network cable, it is impossible to determine the final stable value of the tap coefficient, that is, it is impossible to determine the reference value for adjusting the adaptive weight. However, it is known that when the digital equalization system tends to be stable, the tap coefficients used to process the same periodic signal tend to be equal, so different adaptive weights can be selected according to the difference in tap coefficients.
[0025] In some embodiments, dynamically selecting the adaptive weight corresponding to the difference according to the difference includes: selecting adaptive weights in different ranges according to the size of the difference. In this way, when the equalizer is in an unstable state, the adaptive weight is set to a larger value to speed up the convergence of the equalizer. As the equalizer tends to converge, the adaptive weight is gradually reduced to improve the accuracy and stability of the adaptation.
[0026] In some embodiments, selecting adaptive weights in different ranges according to the size of the difference includes: if the absolute value of the difference is greater than or equal to a first threshold, selecting an adaptive weight in the first range; if the absolute value of the difference is less than the first threshold and greater than or equal to a second threshold, selecting an adaptive weight in the second range; if the absolute value of the difference is less than the second threshold, selecting an adaptive weight in the third range, wherein the first threshold is greater than the second threshold, the first range is greater than the second range, and the second range is greater than the third range. In this way, when the size of the difference is greater than the first threshold, it indicates that the equalizer is in an unstable state, and a larger adaptive weight should be selected to speed up the adaptive speed. When the size of the difference is less than the first threshold and greater than the second threshold, it indicates that the equalizer is gradually tending to a stable state, and a moderate adaptive weight should be selected to take into account both the adaptive speed and the accuracy. When the size of the difference is less than the second threshold, it indicates that the equalizer has tended to a stable state, and a smaller adaptive weight should be selected to obtain a more accurate tap coefficient.
[0027] Figure 2 is a schematic diagram showing a difference and a change in adaptive weight according to an embodiment of the present disclosure. In some embodiments, the difference is calculated based on the following formula: S=F1-D1+Vcomp; Where S is the difference, F1 is the first-stage tap coefficient of the forward feedback equalizer, and D1 is the first-stage tap coefficient of the decision feedback equalizer. In addition, Vcomp is a coefficient compensation value, which is applied to the scenario where the forward tail and the backward tail are inconsistent. It should be understood that for the nth-stage tap coefficient, Fn and Dn can be substituted into the above formula.
[0028] In a specific scenario, the first-stage tap coefficient F1 of the forward feedback equalizer and the first-stage tap coefficient D1 of the decision feedback equalizer are extracted for adaptive weight adjustment. Since the backward tailing of the signal is more serious than the forward tailing, that is, the backward tailing needs to span more cycles, in order to allow the adaptive difference between the tap coefficients to clearly characterize the stable state of the system, the initial coefficient value of the tap coefficient F1 used for iterative update of the forward feedback equalizer can be set to 0, the initial coefficient value of the tap coefficient D1 used for iterative update of the decision feedback equalizer can be set to -1, and the coefficient compensation value can be set to 0. Reference Figure 2 At this time, in the initial state of the digital equalization system, the adaptive difference S between the tap coefficient F1 and the tap coefficient D1 is 1. With the continuous iterative update of the tap coefficient, the adaptive difference S also decreases continuously, and finally the adaptive difference S stabilizes within the preset range. Figure 2 The horizontal axis represents the adaptive period, and the vertical axis represents the adaptive difference.
[0029] In some embodiments, the first range is greater than or equal to 1×10-4 , the second range is less than 1×10 -4 and greater than or equal to 1×10 -6 , the third range is less than 1×10 -6 In this way, the maximum value of the adaptive weight is determined by the signal power, and the adaptive weight that the equalizer can select in different working states is limited based on the maximum value. Since the adaptive weights in different working states are range values, the compatibility of the equalizer with different signal processing can be effectively improved.
[0030] Reference Figure 2 In a specific scenario, the first threshold is 0.5 and the second threshold is 0.1. When the absolute value of the difference S is greater than or equal to 0.5, the digital equalization system is relatively unstable, and the first adaptive weight in the first range should be selected. When the absolute value of the difference S is less than 0.5 and greater than or equal to 0.1, the digital equalization system gradually tends to be stable, and the second adaptive weight in the second range should be selected. When the absolute value of the difference S is less than 0.1, the digital equalization system has tended to be stable, and the third adaptive weight in the third range should be selected. In some embodiments, the first adaptive weight can be set to 2 -4 (i.e. 1 / 16≈0.0625), 2 -4 This is a relatively large step size, suitable for scenarios where the signal changes quickly or requires a quick response. The second adaptive weight can be set to 2 -6 (i.e. 1 / 64≈0.015625), 2 -6 It is a medium-sized step size, which is usually suitable for most application scenarios and balances convergence speed and stability. The third adaptive weight is set to 2 -8 (i.e. 1 / 256≈0.00390625), 2 -8 is a smaller step size, suitable for scenarios where the signal changes slowly or requires very stable behavior. In this case, the algorithm converges more slowly, but the system is less likely to diverge. Based on this, the adaptive weight can be set to 2 as long as the absolute value of the difference is less than 0.1. -8 .
[0031] In some embodiments, the first equalizer is a forward feedback equalizer, and the second equalizer is a decision feedback equalizer. In addition, making the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge based on the adaptive weights respectively includes: making the first equalizer iteratively update the first tap coefficient based on the adaptive weights and the first digital signal input to the first equalizer, and making the second equalizer iteratively update the second tap coefficient based on the adaptive weights and the second digital signal input to the second equalizer. In this way, since the first equalizer is used to process the forward interference of the signal and the second equalizer is used to process the backward interference of the signal, adaptive convergence is performed based on the same adaptive weights, and the convergence speed and convergence accuracy of the two equalizers can be synchronously optimized. At the same time, adaptive convergence is performed based on different digital signals, and the tap coefficients of the first equalizer and the second equalizer can be independently adapted to improve the processing effect of different interferences.
[0032] In some embodiments, until the first tap coefficient and the second tap coefficient tend to be stable includes: if the difference is within a preset range, then stop updating the adaptive weight, wherein the preset range is a numerical interval symmetric about 0. In this way, ideally, when the two equalizers are in a stable state, the difference in tap coefficients between the two should be equal to 0, but because the tap coefficients are in a convergence state, that is, the tap coefficients are stable but fluctuate, it is impossible to achieve absolute 0, so the preset range is set to a numerical interval symmetric about 0, so that the adaptive weight can be stopped when the difference between the tap coefficients tends to 0.
[0033] Figure 3 is a schematic diagram showing a change in iterative updating of tap coefficients according to an embodiment of the present disclosure. Figure 3 In the figure, the horizontal coordinate represents the signal processing period, the vertical coordinate represents the signal amplitude, and the black dot represents the value of the tap coefficient. When the black dot is located on the signal waveform, it means that the tap coefficient converges, making the digital equalization system region stable.
[0034] According to another aspect of the present invention, Figure 4 2 is a block diagram showing an apparatus 200 for adjusting adaptive weights for an equalizer according to an embodiment of the present invention. Figure 4, the apparatus 200 includes an extraction module 202 and a weight configuration module 204. The extraction module 202 is configured to extract a first tap coefficient of a first equalizer and a second tap coefficient of a second equalizer. The weight configuration module 204 is configured to dynamically configure an adaptive weight according to a comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weight until the first tap coefficient and the second tap coefficient tend to be stable.
[0035] According to yet another aspect of the present invention, Figure 5 is a diagram showing an architecture of a balancing system 300 according to an embodiment of the present invention. Figure 5 The system 300 includes a first equalizer 302, a second equalizer 304, and the above-mentioned device 200 for adjusting adaptive weights for an equalizer. The first equalizer 302 and the second equalizer 304 are both electrically coupled to the device 200 for adjusting adaptive weights for an equalizer. In some embodiments, the system 300 further includes a decision maker 306, the outputs of the first equalizer 302 and the second equalizer 304 are electrically coupled to the input of the decision maker 306, and the output of the decision maker 306 is electrically coupled to the input of the second equalizer 304.
[0036] In summary, the present invention provides a method and device for adjusting adaptive weights for an equalizer, and an equalization system, which obtains tap coefficients of different equalizers, and determines the difference between two tap coefficients to determine the current working state of the adaptive equalizer, dynamically configures the adaptive weights according to the difference between the tap coefficients, and adaptively converges the tap coefficients based on the configured adaptive weights, so that the update step size of the tap coefficients can be dynamically adjusted according to the working state of the equalizer. Compared with the method of fixing adaptive weights, the adaptive weights of the present invention can automatically configure larger adaptive weights to speed up the convergence speed when the equalizer is in an unstable state; when the equalizer tends to converge, it automatically configures smaller adaptive weights to improve the adjustment accuracy and stability, so as to simultaneously optimize the convergence speed and steady-state error of the adaptive equalizer.
[0037] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.
Claims
1. A method for adjusting adaptive weights for an equalizer, characterized in that: include: extracting a first tap coefficient of the first equalizer and a second tap coefficient of the second equalizer; The adaptive weight is dynamically configured according to the comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weight until the first tap coefficient and the second tap coefficient tend to be stable.
2. The method according to claim 1, characterized in that Dynamically configuring the adaptive weight according to the comparison result between the first tap coefficient and the second tap coefficient includes: Obtaining a preset coefficient compensation value; and The difference between the first tap coefficient and the second tap coefficient is compensated according to the coefficient compensation value, so as to dynamically set the adaptive weight according to the compensated difference.
3. The method according to claim 1, characterized in that The first tap coefficient and the second tap coefficient are used to process the same periodic signal, wherein dynamically configuring the adaptive weight according to the comparison result between the first tap coefficient and the second tap coefficient comprises: The adaptive weight corresponding to the difference between the first tap coefficient and the second tap coefficient is selected according to the difference between the first tap coefficient and the second tap coefficient.
4. The method according to claim 3, characterized in that Selecting the adaptive weight corresponding to the difference between the first tap coefficient and the second tap coefficient according to the difference comprises: Adaptive weights within different ranges are selected according to the size of the difference.
5. The method according to claim 4, characterized in that Selecting adaptive weights in different ranges according to the difference includes: If the absolute value of the difference is greater than or equal to a first threshold, selecting an adaptive weight within a first range; If the absolute value of the difference is less than the first threshold and greater than or equal to the second threshold, selecting an adaptive weight within a second range; and If the absolute value of the difference is less than the second threshold, an adaptive weight within a third range is selected, The first threshold is greater than the second threshold, the first range is greater than the second range, and the second range is greater than the third range.
6. The method according to claim 5, characterized in that The first range is greater than or equal to 1×10 -4 , the second range is less than 1×10 -4 and greater than or equal to 1×10 -6 , the third range is less than 1×10 -6 .
7. The method according to claim 1, characterized in that The first equalizer is a forward feedback equalizer, and the second equalizer is a decision feedback equalizer, wherein making the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge based on the adaptive weights respectively comprises: enabling the first equalizer to iteratively update the first tap coefficient based on the adaptive weight and a first digital signal input to the first equalizer; and The second equalizer is enabled to iteratively update the second tap coefficient based on the adaptive weight and a second digital signal input to the second equalizer.
8. The method according to claim 1, characterized in that Until the first tap coefficient and the second tap coefficient tend to be stable includes: If the difference between the first tap coefficient and the second tap coefficient is within a preset range, then the updating of the adaptive weight is stopped, The preset range is a numerical interval that is symmetric about 0.
9. A device for adjusting adaptive weights for an equalizer, characterized in that: include: An extraction module configured to extract a first tap coefficient of the first equalizer and a second tap coefficient of the second equalizer; as well as The weight configuration module is configured to dynamically configure the adaptive weight according to the comparison result between the first tap coefficient and the second tap coefficient, so that the first tap coefficient of the first equalizer and the second tap coefficient of the second equalizer converge respectively based on the adaptive weight until the first tap coefficient and the second tap coefficient tend to be stable.
10. A balancing system, characterized in that: include: The device according to claim 9; as well as A first equalizer and a second equalizer are electrically coupled to the device.