CN-122027405-A - TI-DAC amplitude-frequency response pre-equalization method and system based on BO-BiCGStab

Abstract

A TI-DAC amplitude-frequency response pre-equalization method and system based on BO-BiCGStab belong to the technical field of digital signal processing and solve the problems that TI-DAC is affected by factors such as zero-order hold roll-off characteristics, inconsistent sub-channel amplitude-frequency response, non-ideal analog links and the like, so that the amplitude-frequency response of an output signal passband is uneven and the signal quality is reduced. Firstly modeling the TI-DAC passband amplitude-frequency response and analyzing the influencing factors, and converting the precoder frequency response into a linear equation set form by using a cosine matrix form. And then adopting a stable bi-conjugate gradient algorithm to carry out iterative solution on the linear equation set, introducing a frequency weighting mechanism in a modeling stage, and carrying out self-adaptive configuration on the frequency weight by using Bayesian optimization to obtain the optimal coefficient of the pre-equalizer. The invention is suitable for scenes such as high-speed broadband communication and the like.

Inventors

• LIU LIANSHENG
• GU GUANGYU
• PENG YU

Assignees

• 哈尔滨工业大学

Dates

Publication Date

20260512

Application Date

20260210

Claims (10)

1. 1. The TI-DAC amplitude-frequency response pre-equalization method based on BO-BiCGStab is characterized by comprising the following steps of: S1, collecting output amplitude-frequency response data in an uncompensated state of a TI-DAC, preprocessing the output amplitude-frequency response data, and extracting effective frequency point amplitude values in a target passband to obtain amplitude-frequency response to be compensated; S2, constructing a pre-equalization target frequency response by taking the condition that the output amplitude response in the target passband is close to the ideal flat response as a criterion; S3, establishing a pre-equalizer model, selecting a linear phase FIR filter, expanding the frequency response of the pre-equalizer through a cosine basis function, converting a frequency domain fitting problem into a weighted least square linear equation set with a regular term, and adopting BiCGStab algorithm to iteratively solve the linear equation set to obtain initial coefficients of the pre-equalizer; And S4, taking the frequency weight of the frequency band corresponding to the target passband as a super parameter, defining an evaluation index, automatically searching an optimal weight combination through Bayesian optimization, optimizing the initial coefficient of the pre-equalizer based on the optimal weight combination, obtaining the optimal coefficient of the pre-equalizer, and completing the TI-DAC amplitude-frequency response pre-equalization.
2. 2. The BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of claim 1, wherein the pre-processing includes removing outlier data, screening for frequency points within a target passband.
3. 3. The BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of claim 1 wherein the pre-equalization target frequency response is constructed by minimizing amplitude errors in the target passband such that the pre-equalized output amplitude-frequency response approximates the input signal frequency response.
4. 4. The BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of claim 1 wherein the linear phase FIR filter is a Type-I FIR filter with coefficients that satisfy a symmetric characteristic.
5. 5. The BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of claim 1, wherein the weighted least squares linear system of equations with regularization term is expressed as: , Wherein, the Representing the cosine base matrix of the image, , As a vector of weight parameters, In order for the parameters to be regularized, Representing the pre-equalizer coefficient vector, Representing the target frequency response vector.
6. 6. The BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of claim 1 wherein the process of solving the set of linear equations by the BiCGStab algorithm includes setting initial values of pre-equalizer coefficients, calculating initial residuals, iteratively updating coefficient vectors by construction operators until the residuals meet a threshold condition or a maximum number of iterations is reached, outputting pre-equalizer initial coefficients.
7. 7. The BO-BiCGStab-based TI-DAC amplitude-frequency response pre-equalization method according to claim 1, wherein the bayesian optimization adopts a gaussian process to construct a probability proxy model of an evaluation index, and a prediction mean value and uncertainty are obtained.
8. 8. BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization system, characterized in that it is implemented based on the BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method according to any of claims 1-7, comprising: the data acquisition and processing module is used for acquiring output amplitude-frequency response data in an uncompensated state of the TI-DAC, preprocessing the output amplitude-frequency response data, extracting effective frequency point amplitude values in a target passband and obtaining amplitude-frequency response to be compensated; The target frequency response construction module is used for constructing a pre-equalization target frequency response by taking the approach of the output amplitude response in the target passband to the ideal flat response as a criterion; The initial coefficient calculation module is used for establishing a pre-equalizer model, selecting a linear phase FIR filter, expanding the frequency response of the pre-equalizer through a cosine basis function, converting the frequency domain fitting problem into a weighted least square linear equation set with a regular term, and adopting BiCGStab algorithm to iteratively solve the linear equation set to obtain initial coefficients of the pre-equalizer; The optimal coefficient determining module is used for taking the frequency weight of the frequency band corresponding to the target passband as a super parameter, defining an evaluation index, automatically searching an optimal weight combination through Bayesian optimization, optimizing the initial coefficient of the pre-equalizer based on the optimal weight combination, obtaining the optimal coefficient of the pre-equalizer, and completing TI-DAC amplitude-frequency response pre-equalization.
9. 9. A computer device, characterized in that it comprises a memory and a processor, the memory having stored therein a computer program, which when executed by the processor performs the BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method according to any one of claims 1-7.
10. 10. A computer readable storage medium for storing a computer program that performs the BO-BiCGStab based TI-DAC amplitude-frequency response pre-equalization method of any of claims 1-7.

Description

TI-DAC amplitude-frequency response pre-equalization method and system based on BO-BiCGStab Technical Field The invention belongs to the technical field of digital signal processing, and particularly relates to a TI-DAC amplitude-frequency response pre-equalization method based on BO-BiCGStab. Background The time interleaving DAC (TIME INTERLEAVED DAC, TI-DAC) technology can effectively break through the limitation of single-channel sampling rate by parallel interleaving sampling of multi-channel sub-DACs, and realize multiplication of the sampling rate of a system. The TI-DAC structure is relatively simple, the system expandability is strong, and the linearity is high, so that the TI-DAC structure becomes one of the mainstream implementation schemes for improving the sampling rate of the DAC currently. However, TI-DACs are still inevitably affected by various non-ideal factors, and the amplitude-frequency response flatness of the output signals thereof is difficult to ensure. On the one hand, the inherent zero-order hold (ZOH) roll-off characteristic of the DAC severely attenuates the amplitude-frequency response flatness within the passband, thereby degrading signal quality. And the ZOH output of each sub NRZ-DAC can be overlapped with the previous holding sample, and partial amplitude overlapping of the output signal is caused in the interleaving and combining process, so that additional frequency response distortion is introduced. On the other hand, in TI-DACs, amplitude-frequency response inconsistencies between sub-channels, frequency selective effects during interleaving reconstruction, and non-ideal device characteristics of analog links can further lead to amplitude-frequency fluctuations and distortions within the passband. Disclosure of Invention The invention provides a TI-DAC amplitude-frequency response pre-equalization method and system based on BO-BiCGStab. The method aims to solve the problems of uneven amplitude-frequency response and reduced signal quality of an output signal passband caused by the influence of factors such as zero-order maintaining roll-off characteristic, inconsistent amplitude-frequency response of a sub-channel, non-ideal analog link and the like of the TI-DAC. In a first aspect, the present invention aims to provide a BO-BiCGStab-based TI-DAC amplitude-frequency response pre-equalization method, comprising the steps of: S1, collecting output amplitude-frequency response data in an uncompensated state of a TI-DAC, preprocessing the output amplitude-frequency response data, and extracting effective frequency point amplitude values in a target passband to obtain amplitude-frequency response to be compensated; S2, constructing a pre-equalization target frequency response by taking the condition that the output amplitude response in the target passband is close to the ideal flat response as a criterion; S3, establishing a pre-equalizer model, selecting a linear phase FIR filter, expanding the frequency response of the pre-equalizer through a cosine basis function, converting a frequency domain fitting problem into a weighted least square linear equation set with a regular term, and adopting BiCGStab algorithm to iteratively solve the linear equation set to obtain initial coefficients of the pre-equalizer; And S4, taking the frequency weight of the frequency band corresponding to the target passband as a super parameter, defining an evaluation index, automatically searching an optimal weight combination through Bayesian optimization, optimizing the initial coefficient of the pre-equalizer based on the optimal weight combination, obtaining the optimal coefficient of the pre-equalizer, and completing the TI-DAC amplitude-frequency response pre-equalization. Further, a preferable scheme is provided, wherein the preprocessing comprises the steps of removing abnormal data and screening frequency points in a target passband. Further, a preferred scheme is provided, wherein the pre-equalization target frequency response is constructed by minimizing amplitude errors in a target passband, so that the pre-equalized output amplitude-frequency response approximates the input signal frequency response. Still further, a preferred solution is provided wherein the linear phase FIR filter is a Type-I FIR filter with coefficients that satisfy a symmetric characteristic. Still further, a preferred solution is provided wherein the system of weighted least squares linear equations with regularized terms is expressed as: , Wherein, the Representing the cosine base matrix of the image,,As a vector of weight parameters,In order for the parameters to be regularized,Representing the pre-equalizer coefficient vector,Representing the target frequency response vector. Still further, the process of solving the linear equation set by the BiCGStab algorithm comprises setting initial values of the pre-equalizer coefficients, calculating initial residuals, iteratively updating coefficient vectors by const