Foreground mismatch calibration method applied to noise shaping SAR ADC

Abstract

The invention aims to provide a foreground mismatch calibration method applied to a noise shaping SAR ADC (Synthetic Aperture Radar Analog to Digital Converter). The method comprises the following steps: setting parameters of a particle swarm; setting positions and speeds of particles; calculating the fitness of each particle; and repeatedly iterating until the particle position does not change, and outputting the position of the optimal fitness particle. According to the invention, a particle swarm optimization algorithm is introduced, and through meticulous design, a special calibration algorithm is designed in a targeted manner according to the characteristics of the noise shaping SAR ADC. Through random variation, initialization of specific position parameters and gain unification processing, the algorithm is prevented from falling into a local optimal solution, and then a global optimal solution is found.

Description

Technical Field

[0001] The present invention relates to the technical field of integrated circuits, and in particular to a foreground mismatch calibration method applied to a noise shaping SAR ADC. Background Art

[0002] The traditional solution to the impact of feedback capacitor mismatch is to use dynamic element matching (DEM) technology in the design. The most commonly used DEM technologies include single-level averaging (ILA), randomization technology (Randomization), and data weighted averaging (DWA) technology. Among them, DWA is the DEM technology with the best performance and the simplest implementation. Figure 2 As shown in the figure, taking the DWA algorithm as an example, the binary code input Y is converted into the thermometer code dt, and the input Y is passed through the accumulator to obtain the pointer signal dp. dp controls the barrel shifter to move dt by dp bits, and the output signal ds controls the selection of CDAC components. Since the probability of selecting each unit component is equal, the mismatch error Ed' is uniformly shaped into the frequency band to reduce harmonic distortion. DEM technology is widely used, but there are obvious defects. The DEM circuit module is added to the feedback loop, and the scale of the circuit module increases exponentially with the increase of the number of bits, which seriously affects the running speed of the circuit. Therefore, the number of bits is usually limited to 3 to 6 bits. At the same time, considering the complexity of the timing of the noise shaping SAR ADC circuit itself, the design difficulty of the DEM circuit module is greatly increased, and the calibration effect is very limited. In order to break through the limitations of traditional technology, it is of great significance to use new calibration technology to break the performance limitations of noise shaping SAR ADC.

[0003] In ADC design, calibration technology is not uncommon and is widely used in Nyquist ADC. It can be roughly divided into four types: analog foreground calibration, analog background calibration, digital foreground calibration and digital background calibration. Calibration using analog methods requires changes to the original analog circuit, which affects the performance of the analog circuit. In contrast, digital calibration has less impact on the performance of analog circuits and is more popular. In recent years, various advanced ADC calibration methods have emerged: such as pseudo-random injection algorithms, ADC calibration methods based on reference channels, ADC calibration methods using split structures, and ADC calibration methods using neural networks. These algorithms can basically only be used in Nyquist ADCs, such as pipelined ADCs. This is because the output of the oversampled ADC needs to be filtered by a downsampling filter (DSF). The filter output is accompanied by the conversion of the original N-bit code system to M (M>N) bit code values. The additional bit code M compared to N has destroyed the corresponding relationship between the original code value and the mismatch component, making it particularly difficult to calibrate the oversampled ADC in the digital domain, such as the noise shaping SAR ADC. Summary of the invention

[0004] The purpose of the present invention is to provide a foreground mismatch calibration method for noise shaping SAR ADC, which introduces a particle swarm optimization algorithm and is carefully designed to specifically design a special calibration algorithm according to the characteristics of the noise shaping SAR ADC. The algorithm is prevented from falling into a local optimal solution by random mutation, initialization of specific position parameters, and gain unification processing, thereby finding a global optimal solution.

[0005] A foreground mismatch calibration method for noise shaping SAR ADC, comprising:

[0006] Set the parameters of the particle swarm;

[0007] Set the position and speed of particles;

[0008] Calculate the fitness of each particle;

[0009] Repeat the iteration until the particle position no longer changes, and output the position of the particle with the best fitness.

[0010] Preferably, the parameters of setting the particle swarm include:

[0011] Initialize the basic parameters of the particle swarm algorithm, including: particle swarm size J, space dimension S, maximum number of optimization iterations M, initial inertia weight A 1 , Self-learning factor A 2 , Group learning factor A 3 , random mutation probability P, boundary conditions V of position and velocity max 、V min , X max , X min .

[0012] Preferably, setting the position and speed of the particle comprises:

[0013] Initialize the position and velocity of each particle, where the initial positions of all particles are specified as the binary weight coefficient vector C of the ADC output data 0 , and remove C 0 The first item of weight is 1;

[0014] For an n-bit ADC output data code, C 0 is an n-dimensional vector, C 0 Each dimension of n-1 The weight coefficient of the exponential increase is composed of C 0 =[1,2,4,8,...,2 n-1 ] T , remove C 0 After the first item with a weight of 1, C 0 '=[2,4,8,...,2n-1 ] T ;

[0015] C 0 'As the initial position X of all particles 0 , the spatial dimension of each particle in the particle swarm is n-1, the velocity spatial dimension of each particle is also n-1, and the velocity of each particle is randomly initialized.

[0016] Preferably, the step of calculating the fitness of each particle comprises:

[0017] The position of a single particle is restored to the weight coefficient C. The restored weight C is obtained by filling the first term 1 of the position of a single particle X, that is, X = [x 1 ,x 2 ,x 3 ,...,x n-1 ] T , then C=[1,x 1 ,x 2 ,...,x n-1 ] T , let x 0 =1, that is, weight C = [x 0 ,x 1 ,x 2 ,...,x n-1 ] T , Y=[y 0 ,y 1 ,y 2 ,...,y l-1 ] T , where l represents the total length of data stored in BUF1, k∈N and 0≤k≤l-1, After binary bit splitting, y Y' and C perform matrix multiplication to obtain Z, Z = Y'C = [z 0 ,z 1 ,z 2 ,...,z l-1 ] T in

[0018] Window function coefficient W = [w 0 ,w 1 ,w 2 ,...,w l-1 ] T , dot multiplication of Z and W gives R = Z.*W = [r 0 ,r 1 ,r 2 ,...,r l-1 ] T , where rk =z k w k , R is calculated through fast Fourier transform FFT to obtain the frequency domain samples F = FFT{R} = [f 0 ,f 1 ,f 2 ,...,f l-1 ] T , through the cutter Slice, select the sample to be calibrated F'=[f s ,f s+1 ,f s+2 ,...,f p-1 ] T , where s≥0, OSR is the downsampling factor. The specific value range of s and p is determined by the selected dual-tone calibration signal frequency. F' is a complex signal. The fitness is obtained by taking the modulus of the complex signal and then summing it.

[0019] Preferably, the step of repeatedly iterating until the particle position does not change and outputting the position of the particle with the best fitness comprises:

[0020] After the iteration is terminated, the particle position X with the best global fitness is output. gbest =[x gbest,1 ,x gbest,2 ,x gbest,3 ,...,x gbest,n-1 ] T As calibration parameters;

[0021] Let x gbest,0 =1, restore weight C gbest =[x gbest,0 ,x gbest,1 ,x gbest,2 ,...,x gbest,n-1 ] T ;

[0022] Subtract the initial weight C 0 =[1,2,4,8,...,2 n-1 ] T , and obtain the mismatch error estimate

[0023] Preferably, the step of repeatedly iterating until the particle position does not change and outputting the position of the particle with the best fitness further comprises:

[0024] If the termination condition is not met, continue to update the position and velocity of each particle:

[0025] Update the particle's velocity:

[0026]

[0027] Among them, A 1 represents the inertia weight, A 2 is the self-learning factor, A 3 is the group learning factor, B 1 and B 2 is a random number uniformly distributed between 0 and 1. is the position where the jth particle obtains its historical optimal fitness in the mth iteration. is the position where all particles obtain the historical optimal fitness in the mth iteration, and That is, the old and new velocities of the jth particle in the mth iteration. The first term is the velocity inertia, and the second term The third part of self-awareness It constitutes the social cognitive part;

[0028] Determine whether the boundary conditions are met item by item. If the boundary conditions are exceeded, reset the current item value to the boundary value. is an n-1 dimensional column vector, Boundary value V max 、V min Custom initialization for V max =[v max,1 ,v max,2 ,v max,3 ,...,v max,n-1 ] T , V min =[v min,1 ,v min,2 ,v min,3 ,...,v min,n-1 ] T ;

[0029] Update the position of the particle:

[0030]

[0031] in and That is, the old position and new position of the jth particle in the mth iteration. After the particle position is updated, it is determined item by item whether it meets the boundary conditions. If it exceeds the boundary, the boundary value is reset. is an n-1 dimensional column vector, Boundary value X max , X min Custom initialization for X max =[x max,1 ,x max,2 ,x max,3,...,x max,n-1 ] T , X min =[x min,1 ,x min,2 ,x min,3 ,...,x min,n-1 ] T .

[0032] Preferably, after repeating the iteration until the particle position does not change and outputting the position of the particle with the best fitness, the method further includes:

[0033] Update the parameters of the particles;

[0034] Randomly mutate particles;

[0035] The signal gain is normalized.

[0036] Preferably, the parameters of the updated particles include:

[0037] The updated parameters include: in They represent the historically optimal fitness and corresponding position of the jth particle in the mth iteration, and They represent the historically optimal fitness and corresponding positions of all particles in the mth iteration.

[0038] Preferably, the randomly mutating particles comprises:

[0039] Set the mutation probability of each particle to P. When a particle mutates, randomly select the i-th position parameter of the current particle and re-initialize it randomly within the boundary conditions.

[0040] At each iteration the following statements are executed with probability P:

[0041]

[0042] Randomly select a sequence number i from 1 to n-1

[0043] Assign value to x min,i to x max,i A random value between .

[0044] Preferably, the unifying process of the signal gain includes:

[0045] make The sum of all dimensions is equal to X 0 The sum of all dimensions will be Reassign to

[0046] The beneficial effects of the present invention are as follows: 1. The present invention adopts a dual-tone signal for excitation. Compared with other calibration schemes, the dual-tone signal adopted by the present invention does not require a signal source that exceeds the accuracy of the ADC, so it can be easily obtained in an actual laboratory environment. 2. The present invention adopts a frequency domain calibration technology. Compared with the time domain calibration technology, the frequency domain calibration can extract obvious mismatch characteristics, and can separate the signal and the noise for targeted calibration. Due to the introduction of mismatch, the noise shaping SAR ADC behaves as a nonlinear system. The dual-tone signal will introduce intermodulation distortion through the nonlinear system, so it has obvious mismatch characteristics in the frequency domain. According to the limited time domain sample points, the FFT is converted into frequency domain sample points, and then the frequency domain is intercepted to intercept samples containing intermodulation components, and these samples have obvious digital characteristics. These features are constructed into an optimization problem, and a particle swarm optimization algorithm (PSO) is adopted. The frequency domain sample points are used as training data points, the cumulative sum of the absolute values ​​of the frequency domain samples that are minimized and intercepted is taken as the optimization target, and the weight parameters are used as the parameters to be optimized of the model. After multiple iterations, the optimal weight estimation is obtained. 3. The present invention uses the optimal weight estimation obtained after training with the particle swarm optimization algorithm as the calibration parameter, and uses a new test signal for test verification to evaluate the calibration effect. The model data and the actual data are verified respectively to evaluate the practicality and reliability of the calibration scheme. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0049] Figure 1 A flow chart of a foreground mismatch calibration method for noise shaping SAR ADC according to the present invention;

[0050] Figure 2 A schematic diagram of a foreground mismatch calibration principle for noise shaping SAR ADC according to the present invention;

[0051] Figure 3 This is a model verification case diagram of the foreground mismatch calibration algorithm applied to the noise shaping SAR ADC of the present invention;

[0052] Figure 4 It is a training convergence process and real-time verification curve diagram of the calibration algorithm of the present invention in the simulation model;

[0053] Figure 5 A diagram showing the use of spectrum analysis for the simulation model test data of the present invention;

[0054] Figure 6 It is a training convergence process and real-time verification curve diagram of the calibration algorithm of the present invention in the tape-out sample;

[0055] Figure 7 A diagram showing the use of spectrum analysis for the tape-out sample test data of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0058] In addition, the descriptions of “first”, “second”, etc. in the present invention are only for descriptive purposes and cannot be used to describe the present invention.

[0059] It is understood to indicate or imply its relative importance or implicitly indicate the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but it must be based on the ability of ordinary technicians in the field to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0060] In ADC design, calibration technology is not uncommon and is widely used in Nyquist ADC. It can be roughly divided into four types: analog foreground calibration, analog background calibration, digital foreground calibration and digital background calibration. Calibration using analog methods requires changes to the original analog circuit, which affects the performance of the analog circuit. In contrast, digital calibration has less impact on the performance of analog circuits and is more popular. In recent years, various advanced ADC calibration methods have emerged: such as pseudo-random injection algorithms, ADC calibration methods based on reference channels, ADC calibration methods using split structures, and ADC calibration methods using neural networks. These algorithms can basically only be used in Nyquist ADCs, such as pipelined ADCs. This is because the output of the oversampled ADC needs to be filtered by a downsampling filter (DSF). The filter output is accompanied by the conversion of the original N-bit code system to M (M>N) bit code values. The additional bit code M compared to N has destroyed the corresponding relationship between the original code value and the mismatch component, making it particularly difficult to calibrate the oversampled ADC in the digital domain, such as the noise shaping SAR ADC.

[0061] The present invention adopts a dual-tone signal for excitation. Compared with other calibration schemes, the dual-tone signal adopted in the present invention does not require a signal source that exceeds the accuracy of ADC, so it can be easily obtained in an actual laboratory environment. The present invention adopts a frequency domain calibration technology. Compared with the time domain calibration technology, the frequency domain calibration can extract obvious mismatch characteristics, and can separate signals and noise for targeted calibration. Due to the introduction of mismatch, the noise shaping SAR ADC behaves as a nonlinear system, and the dual-tone signal will introduce intermodulation distortion through the nonlinear system, so it has obvious mismatch characteristics in the frequency domain. According to the limited time domain sample points, it is converted into frequency domain sample points by FFT, and then the frequency domain is intercepted, and samples containing intermodulation components are intercepted, and these samples have obvious digital characteristics. These features are constructed into an optimization problem, and a particle swarm optimization algorithm (PSO) is adopted, and the frequency domain sample points are used as training data points, and the cumulative sum of the frequency domain samples after minimizing the interception is taken as the optimization target, and the weight parameter is used as the parameter to be optimized of the model, and the optimal weight estimation is obtained after multiple iterations. The present invention uses the optimal weight estimation obtained after training with the particle swarm optimization algorithm as the calibration parameter, and uses a new test signal for test verification to evaluate the calibration effect. The model data and the actual data are verified respectively to evaluate the practicality and reliability of the calibration scheme.

[0062] Example 1

[0063] A foreground mismatch calibration method for noise shaping SAR ADC, refer to Figure 1 ,include:

[0064] S100, setting the parameters of the particle swarm;

[0065] S200, set the position and speed of the particle;

[0066] S300, calculate the fitness of each particle;

[0067] S400, repeat the iteration until the particle position does not change, and output the position of the particle with the best fitness.

[0068] The present invention aims to achieve the nonlinear problem of noise shaping SAR ADC caused by capacitor mismatch by using digital calibration technology without changing the original analog circuit structure of ADC, which needs to overcome several technical difficulties. First, a suitable excitation signal is selected to be input into the ADC so that the output digital code has digital features that are obviously related to the mismatch. Second, the digital output signal of the ADC is collected, and the mismatch information is accurately extracted from the digital signal through the calibration algorithm. Third, the extracted mismatch information is used to calibrate the ADC, and a new test signal is input into the ADC. The output code system is calculated to obtain a digital signal with higher accuracy.

[0069] In order to solve the first technical problem, the present invention adopts a dual-tone signal for excitation. Compared with other calibration schemes, the dual-tone signal adopted by the present invention does not require a signal source with accuracy exceeding ADC, so it can be easily obtained in an actual laboratory environment.

[0070] In order to solve the second technical problem, the present invention adopts frequency domain calibration technology, and compared with time domain calibration technology, frequency domain calibration can extract obvious mismatch characteristics, and can separate signal and noise for targeted calibration. Due to the introduction of mismatch, noise shaping SAR ADC behaves as a nonlinear system, and the two-tone signal will introduce intermodulation distortion through the nonlinear system, so it has obvious mismatch characteristics in the frequency domain. According to the limited time domain sample points, it is converted into frequency domain sample points by FFT, and then the frequency domain is intercepted, and the samples containing the intermodulation components are intercepted, and these samples have obvious digital characteristics. These features are constructed into an optimization problem, and the particle swarm optimization algorithm (PSO) is adopted, and the frequency domain sample points are used as training data points, and the cumulative sum after the absolute value of the frequency domain samples that minimize the interception is taken is used as the optimization target, and the weight parameter is used as the parameter to be optimized of the model, and the optimal weight estimation is obtained after multiple iterations.

[0071] In order to solve the third technical problem, the present invention uses the optimal weight estimation obtained after training with the particle swarm optimization algorithm as the calibration parameter, and uses a new test signal for test verification to evaluate the calibration effect. The model data and the physical data are verified respectively to evaluate the practicality and reliability of the calibration scheme.

[0072] Preferably, reference Figure 2 , S100, setting the parameters of the particle swarm includes:

[0073] Initialize the basic parameters of the particle swarm algorithm, including: particle swarm size J, space dimension S, maximum number of optimization iterations M, initial inertia weight A 1 , Self-learning factor A 2 , Group learning factor A 3 , random mutation probability P, boundary conditions V of position and velocity max 、V min , X max , X min .

[0074] The calibration process is divided into two stages, the training stage and the testing stage. In the training stage, dual-tone signals are used for training, which consist of two close single-tone signals U 1 and U 2 Synthetic U in , specifically implemented by an analog circuit without introducing intermodulation distortion, simulating the training signal U in The digital signal Y is obtained through the noise shaping SAR ADC model, and the binary code of Y is stored in the buffer (BUF1) using the data acquisition system. The data in BUF1 is used as the training data for the particle swarm optimization algorithm calibration system.

[0075] Preferably, S200, setting the position and speed of the particle includes:

[0076] Initialize the position and velocity of each particle, where the initial positions of all particles are specified as the binary weight coefficient vector C of the ADC output data 0 , and remove C 0 The first item of weight is 1;

[0077] For an n-bit ADC output data code, C 0 is an n-dimensional vector, C 0 Each dimension of n-1 The weight coefficient of the exponential increase is composed of C 0 =[1,2,4,8,...,2 n-1 ] T , remove C 0 After the first item with a weight of 1, C 0 '=[2,4,8,...,2 n-1 ] T ;

[0078] C 0 'As the initial position X of all particles 0 , the spatial dimension of each particle in the particle swarm is n-1, the velocity spatial dimension of each particle is also n-1, and the velocity of each particle is randomly initialized.

[0079] Preferably, S300, calculating the fitness of each particle includes:

[0080] The position of a single particle is restored to the weight coefficient C. The restored weight C is obtained by filling the first term 1 of the position of a single particle X, that is, X = [x 1 ,x 2 ,x 3 ,...,x n-1 ] T , then C=[1,x 1 ,x 2 ,...,x n-1 ] T , let x 0 =1, that is, weight C = [x 0 ,x 1 ,x 2 ,...,x n-1 ] T , Y=[y 0 ,y 1 ,y 2 ,...,y l-1 ] T , where l represents the total length of data stored in BUF1, k∈N and 0≤k≤l-1, After Y is split into binary bits, Y'= Y' and C perform matrix multiplication to obtain Z, Z = Y'C = [z 0 ,z 1 ,z 2 ,...,z l-1 ] T in

[0081] Window function coefficient W = [w 0 ,w 1 ,w 2 ,...,w l-1 ] T , dot multiplication of Z and W gives R = Z.*W = [r 0 ,r 1 ,r 2 ,...,r l-1 ] T , where r k =z k w k , R is calculated through fast Fourier transform FFT to obtain the frequency domain samples F = FFT{R} = [f 0 ,f 1 ,f 2 ,...,f l-1 ] T , through the cutter Slice, select the sample to be calibrated F'=[fs ,f s+1 ,f s+2 ,...,f p-1 ] T , where s≥0, OSR is the downsampling factor. The specific value range of s and p is determined by the selected dual-tone calibration signal frequency. F' is a complex signal. The fitness is obtained by taking the modulus of the complex signal and then summing it.

[0082] Preferably, in S400, iterating repeatedly until the particle position does not change, outputting the position of the particle with the best fitness includes:

[0083] After the iteration is terminated, the particle position X with the best global fitness is output. gbest =[x gbest,1 ,x gbest,2 ,x gbest,3 ,...,x gbest,n-1 ] T As calibration parameters;

[0084] Let x gbest,0 =1, restore weight C gbest =[x gbest,0 ,x gbest,1 ,x gbest,2 ,...,x gbest,n-1 ] T ;

[0085] Subtract the initial weight C 0 =[1,2,4,8,...,2 n-1 ] T , and obtain the mismatch error estimate

[0086] Preferably, repeating the iteration until the particle position does not change, and outputting the position of the particle with the best fitness also includes:

[0087] If the termination condition is not met, continue to update the position and velocity of each particle:

[0088] Update the particle's velocity:

[0089]

[0090] Among them, A 1 represents the inertia weight, A 2 is the self-learning factor, A 3 is the group learning factor, B 1 and B 2 is a random number uniformly distributed between 0 and 1. is the position where the jth particle obtains its historical optimal fitness in the mth iteration. is the position where all particles obtain the historical optimal fitness in the mth iteration, and That is, the old and new velocities of the jth particle in the mth iteration. The first term is the velocity inertia, and the second term The third part of self-awareness It constitutes the social cognitive part;

[0091] Determine whether the boundary conditions are met item by item. If the boundary conditions are exceeded, reset the current item value to the boundary value. is an n-1 dimensional column vector, Boundary value V max 、V min Custom initialization for V max =[v max,1 ,v max,2 ,v max,3 ,...,v max,n-1 ] T , V min =[v min,1 ,v min,2 ,v min,3 ,...,v min,n-1 ] T ;

[0092] The pseudo code for performing boundary judgment operations is:

[0093] At the beginning of the loop, i increases by 1 from 1 to n-1.

[0094] if:

[0095] but: Assign value to v max,i

[0096] if:

[0097] but: Assign value to v min,i

[0098] The loop ends.

[0099] Update the position of the particle:

[0100]

[0101] in and That is, the old position and new position of the jth particle in the mth iteration. After the particle position is updated, it is determined item by item whether it meets the boundary conditions. If it exceeds the boundary, the boundary value is reset. is an n-1 dimensional column vector, Boundary value X max , X min Custom initialization for X max =[x max,1 ,x max,2 ,x max,3 ,...,x max,n-1 ] T , X min =[x min,1 ,x min,2 ,x min,3 ,...,x min,n-1 ] T .

[0102] The pseudo code for performing boundary judgment operations is:

[0103] At the beginning of the loop, i increases by 1 from 1 to n-1.

[0104] if:

[0105] but: Assign value to x max,i

[0106] if:

[0107] but: Assign value to x min,i

[0108] The loop ends.

[0109] Preferably, S400, after repeating the iteration until the particle position does not change and outputting the position of the particle with the best fitness, further includes:

[0110] S510, updating the parameters of the particles;

[0111] S520, randomly mutating particles;

[0112] S530: performing uniform processing on the signal gains.

[0113] Preferably, S510, updating the parameters of the particle includes:

[0114] The updated parameters include: in and They represent the historically optimal fitness and corresponding position of the jth particle in the mth iteration, and They represent the historically optimal fitness and corresponding positions of all particles in the mth iteration.

[0115] Preferably, S520, performing random mutation on particles includes:

[0116] Set the mutation probability of each particle to P. When a particle mutates, randomly select the i-th position parameter of the current particle and re-initialize it randomly within the boundary conditions.

[0117] At each iteration the following statements are executed with probability P:

[0118]

[0119] Randomly select a sequence number i from 1 to n-1

[0120] Assign value to x min,i to x max,i A random value between .

[0121] In order to avoid the particle swarm falling into the local optimal solution as much as possible, a random mutation mechanism is introduced. The mutation probability of each particle is P. When a particle mutates, the i-th position parameter of the particle is randomly selected and re-initialized randomly within the boundary conditions.

[0122] Preferably, S530, performing unified processing on the signal gain includes:

[0123] In order to ensure that the signal gain remains unchanged before and after calibration, the gain unification processing module is used to make The sum of all dimensions is equal to X 0 The sum of all dimensions will be Reassign to

[0124] Repeat the above steps for the new particle positions. When the termination condition is met, the training phase ends and the testing phase begins, which is also the normal conversion phase. Figure 4 and Figure 6 As shown, the input signal U_in uses a single tone signal as the input source to evaluate the algorithm calibration performance. The calibration coefficient is stored in BUF2, and the performance parameters SNR (signal-to-noise ratio), SNDR (signal-to-noise and distortion ratio), SFDR (spurious-free dynamic range), THD (total harmonic distortion), ENOB (effective number of bits) and other parameters are calculated for evaluation.

[0125] Before implementing the calibration algorithm, build a verification platform, taking the SAR ADC quantization bit number of 9 bits as an example. Figure 3 shown.

[0126] U1, U2, and U3 are external single-tone signal sources, where U1 and U2 synthesize dual-tone signals through an addition circuit to stimulate ADC training data. The two dashed boxes are exactly the same noise-shaping SAR ADC simulation models with exactly the same mismatch. At the same time, non-ideal factors such as op amp noise, sampling noise, and op amp limited gain are added to collect digital signals Y1 and Y2. In the actual implementation process, Y1 and Y2 are data collected from the same noise-shaping SAR ADC at different time periods. The amount of training data collected is l = 1048576 samples. U1 and U2 are within half of the full amplitude of the ADC, and U3 is within the full amplitude of the ADC to ensure that the ADC will not be overloaded during the conversion process. H(z) is the loop filter. In this case, it is set to a second-order loop filter, and the expression is: When the noise shaping SAR ADC sampling rate Fs is set to 32MSPS, the oversampling rate OSR is 8, and the full-scale input range is 2V. The frequencies of U1, U2, and U3 are set to 42929 and 51787 are both prime numbers to meet the conditions of coherent sampling. The peak-to-peak value of the signal is set to Vpp = 0.95V. U3 is used as the test signal and the frequency is set to The signal peak value is set to Vpp = 1.8 V. Initialize the basic parameters of the particle swarm algorithm, set the swarm size J = 50, the space dimension S = 8, the maximum number of optimization iterations M = 200, the initial inertia weight A 1 =0.8, self-learning factor A 2 =0.6, group learning factor A 3 =0.4, random mutation probability P = 0.05, initial weight coefficient C 0 =[1,2,4,8,16,32,64,128,256] T , initial position X 0 =[2,4,8,16,32,64,128,256] T , initialization velocity boundary condition V max =0.1X 0 、V min =-0.1X 0 , X max =1.1X 0 , X min =0.9X 0 The slicer Slice is set to s=512, p=42000.

[0127] After 100 iterations, the data has basically converged, and the data after the 200th iteration is taken as the final calibration data. The simulation model can be used to easily extract the true value of the original mismatch and calculate its equivalent weight as a reference value for accurate comparison of the calibration effect. The results after 200 iterations are shown in Table 1. The error of the calibrated value relative to the reference value is within 0.01%. These three sets of weights are used for spectrum analysis of the test signal. The spectrum analysis results are shown in Table 1. Figure 5 As shown, it can be found that the calibration effect is very obvious, and the performance after calibration is almost the same as the reference performance, indicating that the solution after calibration is already the global optimal solution.

[0128] Table 1 Comparison of weights and reference values ​​of the calibration algorithm before and after 200 iterations of training in the simulation model

[0129]

[0130] The above simulation model verifies the accuracy and feasibility of the calibration algorithm. Next, the algorithm is verified by the test data of the actual noise shaping SAR ADC chip sample. The actual ADC chip used is a 9-bit second-order noise shaping SAR ADC, the sampling rate is also 32MSPS, and other algorithm parameters remain unchanged. The calibration results after 200 iterations are shown in Table 2. The performance changes before and after calibration are compared through spectrum analysis. The spectrum analysis results are shown in Table 2. Figure 7 As shown in the figure, SNDR is improved by 5.4dB, and ENOB is increased from the original 11.1bit to 12bit. The calibration effect is obvious.

[0131] Table 2 Weight changes of the calibration algorithm before and after 200 iterations of training in the tape-out sample

[0132]

[0133] The present invention adopts a dual-tone signal for excitation. Compared with other calibration schemes, the dual-tone signal adopted in the present invention does not require a signal source that exceeds the accuracy of ADC, so it can be easily obtained in an actual laboratory environment. The present invention adopts a frequency domain calibration technology. Compared with the time domain calibration technology, the frequency domain calibration can extract obvious mismatch characteristics, and can separate signals and noise for targeted calibration. Due to the introduction of mismatch, the noise shaping SAR ADC behaves as a nonlinear system, and the dual-tone signal will introduce intermodulation distortion through the nonlinear system, so it has obvious mismatch characteristics in the frequency domain. According to the limited time domain sample points, it is converted into frequency domain sample points by FFT, and then the frequency domain is intercepted, and samples containing intermodulation components are intercepted, and these samples have obvious digital characteristics. These features are constructed into an optimization problem, and a particle swarm optimization algorithm (PSO) is adopted, and the frequency domain sample points are used as training data points, and the cumulative sum of the frequency domain samples after minimizing the interception is taken as the optimization target, and the weight parameter is used as the parameter to be optimized of the model, and the optimal weight estimation is obtained after multiple iterations. The present invention uses the optimal weight estimation obtained after training with the particle swarm optimization algorithm as the calibration parameter, and uses a new test signal for test verification to evaluate the calibration effect. The model data and the actual data are verified respectively to evaluate the practicality and reliability of the calibration scheme.

[0134] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A foreground mismatch calibration method for noise shaping SAR ADC, characterized in that: include: Set the parameters of the particle swarm; Set the position and speed of particles; Calculate the fitness of each particle; Repeat the iteration until the particle position no longer changes, and output the position of the particle with the best fitness.

2. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: The parameters for setting the particle swarm include: Initialize the basic parameters of the particle swarm algorithm, including: particle swarm size J, space dimension S, maximum number of optimization iterations M, initial inertia weight A1, self-learning factor A2, group learning factor A3, random mutation probability P, position and velocity boundary conditions V max 、V min , X max , X min .

3. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: The setting of the particle position and velocity includes: Initialize the position and velocity of each particle, where the initial positions of all particles are specified as the binary weight coefficient vector C0 of the ADC output data, and the first weight 1 of C0 is removed; For an n-bit ADC output data code, C0 is an n-dimensional vector, and each dimension of C0 consists of 1, 2, 4, 8, ..., 2 n-1 The weight coefficients are exponentially increasing, that is, C0 = [1, 2, 4, 8, ..., 2 n-1 ] T , after removing the first item of C0 with weight 1, C0'=[2,4,8,...,2 n-1 ] T ; C0' is used as the initial position X0 of all particles. The spatial dimension of each particle in the particle swarm is n-1, and the spatial dimension of the velocity of each particle is also n-1. The velocity of each particle is randomly initialized.

4. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: The calculation of the fitness of each particle includes: The position of a single particle is restored to the weight coefficient C. The restored weight C is obtained by filling the first term 1 of the position of a single particle X, that is, X = [x1, x2, x3, ..., x n-1 ] T , then C=[1,x1,x2,...,x n-1 ] T , let x0=1, that is, weight C=[x0,x1,x2,...,x n-1 ] T ,Y=[y0,y1,y2,...,y l-1 ] T , where l represents the total length of data stored in BUF1, k∈N and 0≤k≤l-1, After Y is split into binary bits Y' and C perform matrix multiplication to obtain Z, Z = Y'C = [z0,z1,z2,...,z l-1 ] T in Window function coefficient W = [w0, w1, w2, ..., w l-1 ] T , Z and W are multiplied by points to get R = Z.*W = [r0,r1,r2,...,r l-1 ] T , where r k =z k w k , R is calculated through fast Fourier transform FFT to obtain frequency domain samples F = FFT{R} = [f0,f1,f2,...,f l-1 ] T , through the cutter Slice, select the sample to be calibrated F'=[f s ,f s+1 ,f s+2 ,...,f p-1 ] T , where s≥0, OSR is the downsampling factor. The specific value range of s and p is determined by the selected dual-tone calibration signal frequency. F' is a complex signal. The fitness is obtained by taking the modulus of the complex signal and then summing it.

5. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: Repeating the iteration until the particle position does not change and outputting the position of the particle with the best fitness includes: After the iteration is terminated, the particle position X with the best historical global fitness is output. gbest =[x gbest,1 ,x gbest,2 ,x gbest,3 ,...,x gbest,n-1 ] T As calibration parameters; Let x gbest,0 =1, restore weight C gbes t=[x gbest,0 ,x gbest,1 ,x gbest,2 ,...,x gbest,n-1 ] T ; Subtract the initial weight C0 = [1,2,4,8,...,2 n-1 ] T , and obtain the mismatch error estimate 6. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: Repeating the iteration until the particle position does not change and outputting the position of the particle with the best fitness also includes: If the termination condition is not met, continue to update the position and velocity of each particle: Update the particle's velocity: Among them, A1 represents the inertia weight, A2 is the self-learning factor, A3 is the group learning factor, B1 and B2 are random numbers evenly distributed between 0 and 1, is the position where the jth particle obtains its historical optimal fitness in the mth iteration. is the position where all particles obtain the historical optimal fitness in the mth iteration, and That is, the old and new velocities of the jth particle in the mth iteration. The first term is the velocity inertia, and the second term The third part of self-awareness It constitutes the social cognitive part; Determine whether the boundary conditions are met item by item. If the boundary conditions are exceeded, reset the current item value to the boundary value. is an n-1 dimensional column vector, Boundary value V max 、V min Custom initialization for V max =[v max,1 ,v max,2 ,v max,3 ,...,v max,n-1 ] T , V min =[v min,1 ,v min,2 ,v min,3 ,...,v min,n-1 ] T ; Update the position of the particle: in and That is, the old position and new position of the jth particle in the mth iteration. After the particle position is updated, it is determined item by item whether it meets the boundary conditions. If it exceeds the boundary, the boundary value is reset. is an n-1 dimensional column vector, Boundary value X max , X min Custom initialization for X max =[x max,1 ,x max,2 ,x max,3 ,...,x max,n-1 ] T , X min =[x min,1 ,x min,2 ,x min,3 ,...,x min,n-1 ] T .

7. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 1, characterized in that: After repeating the iteration until the particle position does not change and outputting the position of the particle with the best fitness, the method further includes: Update the parameters of the particles; Randomly mutate particles; The signal gain is normalized.

8. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 7, characterized in that: The parameters of the updated particles include: The updated parameters include: in and They represent the historically optimal fitness and corresponding position of the jth particle in the mth iteration, and They represent the historically optimal fitness and corresponding positions of all particles in the mth iteration.

9. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 7, characterized in that: The random mutation of particles comprises: Set the mutation probability of each particle to P. When a particle mutates, randomly select the i-th position parameter of the current particle and re-initialize it randomly within the boundary conditions. At each iteration the following statements are executed with probability P: Randomly select a sequence number i from 1 to n-1 Assign value to x min,i to x max,i A random value between .

10. The foreground mismatch calibration method for noise shaping SAR ADC according to claim 7, characterized in that: The unifying process of the signal gain comprises: make The sum of the values ​​of all dimensions is equal to the sum of the values ​​of all dimensions of X0. Reassign to

Images