Method for testing static parameters of ADC (Analog to Digital Converter) under loose condition

Abstract

The invention requests to protect an ADC static parameter test method used under a loose test condition. The method enables the static parameter test of an ADC to be carried out under a non-high-precision device. The method comprises the steps of signal identification and reconstruction, spectrum preprocessing, curve fitting and excitation noise removal. The signal identification and reconstruction part comprises the following steps: acquiring parameters of an ADC output sequence through a frequency spectrum three-interpolation and sine wave three-parameter fitting method, and reducing fitting errors; in the spectrum preprocessing part, the problem of spectrum leakage of data after FFT (Fast Fourier Transform) is inhibited through simple windowing processing and sampling point number conversion; the curve fitting part comprises the following steps: acquiring n-order harmonics through frequency spectrum transformation, fitting an ADC transmission curve, and reducing the requirements of a traditional test on coherent conditions and sample quantity points; the excitation noise removal part comprises the steps of offsetting noise of a signal source through amplitude ratio characteristics of the signal source, identifying nonlinearity of the ADC from static parameters containing external noise, and reducing condition limitation of testing on signal source precision.

Description

Technical Field

[0001] The present invention belongs to the field of integrated circuit testing, and particularly relates to a method for testing static parameters of an ADC, and more particularly to a method for testing static parameters of an ADC under loose test conditions. Background Art

[0002] With the rapid development of fields such as automotive electronics, medical treatment, radar communication, and industrial control, as well as the improvement of the working speed of digital circuits, the requirements of electronic systems for signal sensitivity are getting higher and higher. Therefore, the performance of analog-to-digital converters has become one of the key factors for evaluating signal acquisition circuits.

[0003] The static parameters of an ADC describe the difference between the actual transfer characteristic curve and the ideal transfer curve under the condition of low-speed or DC signal input, reflecting the accuracy, linearity, and stability of the conversion process, which is particularly important in scenarios such as industrial measurement and process control. By testing and analyzing the static parameters of an ADC, its conversion characteristics can be comprehensively understood, which is crucial for the design and evaluation of an ADC. With the continuous improvement of the ADC resolution, it is becoming increasingly difficult to implement traditional static parameter testing methods, and the hardware test platform is difficult to meet the test requirements of high-precision ADCs, and the test time brought by data acquisition increases rapidly.

[0004] The traditional code density histogram testing method is mainly based on mathematical statistics theory. The ADC randomly samples a sine wave signal, and calculates static parameters according to the bathtub-shaped code bins. This method requires collecting a large amount of data input to ensure that each code value appears multiple times. For a 12-bit ADC, under the conditions of a confidence level of 95% and a DNL error of 0.1 LSB (least significant bit), the number of sampling points required is 2,471,680. Since the test results of this method depend on the number of occurrences of digital codes, the test is sensitive to phenomena such as input signal amplitude change, noise, and jitter. The purity of the signal source will directly affect the shape and distribution of the histogram, and it is necessary to strictly control the test equipment in actual testing.

[0005] In order to improve the test accuracy, the cumulative code density histogram testing method is proposed. This method uses the step transition voltage difference between adjacent digital codes to obtain the actual step voltage, making the INL and DNL calculations independent of the amplitude of the sine wave input signal. However, this method, like the code density histogram Figure 1 also requires a large sample size, and when the ADC resolution increases, the number of samples will increase exponentially with a power of 2.

[0006] In order to solve the problem of the number of test points in traditional statistical testing methods, the Chebyshev testing method is proposed.

[0007] This algorithm uses Fourier series and Chebyshev polynomials to approximate the ADC transfer curve, and then uses the Fast Fourier Transform (FFT) to obtain the transfer curve coefficients, thereby obtaining the static parameters of the ADC. However, the algorithm needs to meet the test conditions of coherent sampling to accurately calculate using FFT. When the sequence is an incoherent sequence, spectral leakage will occur after Fourier transform, affecting the test accuracy. And in actual tests, it is difficult to achieve complete synchronization between the signal frequency and the clock frequency due to equipment accuracy limitations.

[0008] To solve the problem of spectral leakage in the Chebyshev method, a spectral parameter estimation method is proposed, which uses the least squares fitting method to obtain the Fourier polynomial coefficients in the time domain. However, this algorithm is affected by the amplitude accuracy and frequency accuracy of the signal source. If the parameter fitting error is large, it will reduce the test accuracy of INL.

[0009] In the existing invention, although the double convolution algorithm proposed in CN114966373A can effectively suppress the spectral leakage caused by incoherent sampling, the system has a large amount of computation. And for the test of high-precision ADCs, this method still requires a high-precision signal source, and the amplitude of the input signal needs to be strictly controlled to avoid objective factors such as jitter and crosstalk from causing the amplitude of the input signal to exceed the ADC conversion range and affecting the spectral parameters. Therefore, to solve the dependence of traditional tests on external high-precision equipment, the present invention starts from three aspects: signal source linearity, amplitude, and coherence, and realizes the test of the ADC performance parameters by non-coherent sequences beyond the range under a low-precision signal source through three steps: clipping identification and reconstruction, data preprocessing, and parameter restoration.

[0010] The inventor's prior application CN117792392A, a method for testing ADC dynamic parameters under loose test conditions, enables the testing of ADC dynamic parameters under non-high-precision equipment. This method includes: clipping sequence identification and reconstruction, data preprocessing, and ADC parameter restoration. This method calculates dynamic performance parameters such as SNR (signal-to-noise ratio), SINAD (signal-to-integration-and-distortion ratio), ENOB (effective number of bits), and THD (total harmonic distortion) for each segment of the ADC output sequence. The present invention proposes a method for testing ADC static parameters under loose conditions. This method includes: signal identification and reconstruction, spectral preprocessing, curve fitting, and excitation noise removal. This method improves the requirements for coherent sampling and high-precision signal sources in the traditional Chebyshev fitting method, and this algorithm also has good adaptability to input signals of different amplitudes. This method tests static performance parameters such as INL (integral nonlinearity) and DNL (differential nonlinearity). Summary of the Invention

[0011] The present invention aims to solve the problems of the above prior art, and proposes a method for testing ADC static parameters under loose conditions. The technical solution of the present invention is as follows:

[0012] A method for testing ADC static parameters under loose conditions, which includes: a signal recognition and reconstruction step, a spectrum preprocessing step, a curve fitting step, and an excitation noise removal step; wherein,

[0013] The signal recognition and reconstruction step includes: collecting the ADC output sequence, performing digital reconstruction to convert the output code value sequence into a discrete sequence, and obtaining the parameters of the ADC output sequence through spectral triple interpolation and sine wave three-parameter fitting method;

[0014] The spectrum preprocessing step includes: obtaining an approximately coherent number of sampling points M through windowing processing and sampling point number transformation int ;

[0015] The curve fitting step includes: enhancing the continuity of the first and last bits of the transformed data through the Hann window function, and then fitting the ADC curve through the Chebyshev fitting algorithm;

[0016] The excitation noise removal step includes: utilizing the noise distribution characteristics, reducing the influence of the signal source noise by subtracting the two collected data, removing the signal source noise error, and estimating the ADC-generated INL (Integral Nonlinearity) and DNL (Differential Nonlinearity).

[0017] Further, the signal recognition step and reconstruction specifically include:

[0018] Collect the ADC output sequence, perform digital reconstruction to convert the output code value sequence into a discrete sequence x(n). The part where the input signal exceeds the positive and negative full scales of the ADC shows clipping and bottoming distortion in the ADC output sequence. Among them, A, f, V os are respectively the amplitude, frequency, phase and DC offset of the input signal, h.d is the high-order nonlinear distortion, w n is the total noise, Δt = 1 / F s is the sampling point interval time, F s represents the ADC sampling frequency, V fs represents the ADC full-scale voltage value:

[0019]

[0020] By adding a classical cosine window to the ADC output sequence, the continuity of the first and last bits of the sequence is increased, the energy concentration of the main lobe of the spectrum is further improved, and the amplitude estimation error caused by spectral leakage is reduced. Among them, I is the number of window function terms, a i is the i-th coefficient of the window function, k ris the fundamental frequency index k of the signal r = f / (F S × N), where N represents the number of sampling points, and the discrete sequence k = 1, 2 ··· M; when M is large enough,

[0021] When the Hanning window I = 1 and a1 = 0.5, the obtained w(n) and X(k) are as follows:

[0022]

[0023] When the Hanning window I = 1 and a1 = 0.5, the obtained w(n) and X(k) are as follows:

[0024]

[0025] Calculate the initial value f of the signal frequency through the frequency-domain expression of the ADC output int , where L m represents the spectral index value corresponding to the amplitude component with the largest spectrum, and the discrete sequence k = 1, 2 ··· M corresponds to the spectral index numbers L1, L2 ··· L M as follows:

[0026]

[0027] Since during non-coherent sampling, the spectral index L r corresponding to the fundamental frequency is non-integer and falls between the spectral lines L m-1 and L m+1 , and the relationship between the spectral lines L m-1 , L m , L m+1 and L r is as shown below, where the index difference η between each spectral line is in (-0.5, 0.5):

[0028] L r - L m = η

[0029] L r - L m-1 = η - 1

[0030] L r - L m+1 = η + 1

[0031] Use the spectral line L m corresponding to the maximum amplitude component max(|X(k)|) of the spectrum and its two adjacent spectral lines L m-1 and L m+1 to correct the fundamental frequency spectral line L r :

[0032]

[0033] After directly setting the frequency correction formula to solve the index difference η between each spectral line, further obtain the subsequent parameters by fitting the three parameters of the sine wave:

[0034]

[0035] Complete through the correction formula After estimation, establish the minimum square difference E between the fitting sequence and the original sequence, and obtain the least squares fitting value r in matrix form, where is the fitting DC offset, is the fitting cosine amplitude, is the fitting sine amplitude:

[0036] r = (M T M) -1 (M T x)

[0037]

[0038] Solve the three-parameter fitting of the sine wave with the minimum residual to obtain the amplitude phase parameters:

[0039]

[0040] Reconstruct the ideal clipped sequence according to the fitting parameters:

[0041]

[0042] Furthermore, the spectrum preprocessing step includes:

[0043] Obtain the residual sequence including ADC self-stray, external excitation stray, and fitting error through the reconstructed ideal clipped sequence and the ADC output sequence:

[0044] r(n) = x(n) - x clip (n) = s adc (n) + s c (n) + s f (n)

[0045] where s adc is the stray component generated by the ADC's own design, including quantization noise, thermal noise, high-order nonlinearity and other stray components; s c represents the influence of external excitation on the ADC output sequence, including stray components brought by external devices such as signal source accuracy and clock jitter; s f is the fitting error generated by the reconstructed sequence;

[0046] Considering that the residual sequence r(n) of the clipped input signal has a periodic law, the sampling transformation formula transforms the number of sampling points to enhance the first continuity of the residual sequence:

[0047]

[0048] Among them, J int is the integer part of the sampling period number J, M int is the number of data points contained in an integer period.

[0049] Furthermore, the curve fitting step includes:

[0050] When the ADC test does not meet the marked test conditions, the ADC output sequence is expressed in the form of x(n):

[0051]

[0052] By combining the first Chebyshev polynomial C n (ξ)=cos(n·arccos(ξ)), establish the relationship between the ADC input analog voltage x and the ADC output digital code y(n), and obtain the ADC transfer curve g(x) under ideal test conditions:

[0053]

[0054] Using the A obtained by preprocessing the spectrum Y(k) n The ADC transfer curve g can be constructed s (x) When the input signal amplitude exceeds the full scale of the ADC, limiting the harmonic amplitude calculation range to even harmonics can reduce the impact of additional spurious signals introduced by clipping distortion on the ADC transfer curve fitting. At this time, the measured INL and DNL include not only the ADC's own spurious signals, but also the spurious components introduced by factors such as signal source accuracy, clock jitter, and fitting errors, and cannot directly reflect the ADC's static performance:

[0055]

[0056] g s (x) = g(x) + n q +n a +n f

[0057] Among them, n q is the quantization error, n a is the stray introduced by external conditions, n f is the fitting error;

[0058] When the sampling voltage is x and the output level is C, g s(x) can be represented by the actual conversion level T(n), the ideal conversion level T i (n), the code value C, and the INL non - linear error, where V LSB represents the voltage corresponding to the least significant bit in the ADC:

[0059]

[0060] T(0) is the conversion level from code value 0 to 1. When the offset error is not considered, T(0)=T i (0). For the same ADC, T(0) is the same value.

[0061] Furthermore, the excitation noise removal step includes:

[0062] By the difference between the fitting curves g s1 (x) and g s2 (x) of two groups of sampling sequences, the signal source noise, harmonics, fitting error, and external environment spurs are cancelled, and the non - linear error of the ADC itself is obtained; where C1 is the code value C output by the first - stage fitting curve g s1 (x), INL1(C) is the integral non - linear error of the first - stage sequence at the code value C, C2 is the code value C output by the second - stage fitting curve g s2 (x), and INL2(C) is the integral non - linear error of the second - stage sequence at the code value C:

[0063]

[0064] By quantifying the difference between g s1 (x) and g s2 (x) with the voltage corresponding to the least significant bit in the ADC:

[0065]

[0066] The INL error obtained after the INL errors caused by the signal source spurs and the inherent spurs are cancelled is the non - linear error of the ADC itself:

[0067]

[0068] DNL = INL(C + 1)-INL(C).

[0069] An electronic device, which includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the program, it implements the ADC static parameter test method applied to loose test conditions as described in any one of the above.

[0070] A non-transitory computer-readable storage medium stores a computer program thereon, and when the computer program is executed by a processor, it implements the ADC static parameter test method applied to loose test conditions as described in any one of the above.

[0071] A computer program product includes a computer program, and when the computer program is executed by a processor, it implements the ADC static parameter test method applied to loose test conditions as described in any one of the above.

[0072] The advantages and beneficial effects of the present invention are as follows:

[0073] To solve the existing problems, the present invention proposes a new test algorithm for ADC static parameters based on spectral interpolation and Chebyshev polynomials. Without meeting the requirements of signal source purity and coherent sampling, the ADC static parameters can be accurately tested with fewer points. While reducing the design cost of the ADC test prototype, it improves the test range of analog-to-digital hybrid test equipment on the market and reduces the test cost.

[0074] 1. By adopting the above method of the present invention, the following beneficial effects can be achieved: It can measure the static performance parameters of the ADC through signals of different amplitudes when the signal coherence and accuracy do not meet the standard test conditions, and only 8000 points are required for the test, which can effectively reduce the test cost and improve the test efficiency, laying a foundation for testing dynamic and static parameters using the same set of data.

[0075] 2. The spectral preprocessing algorithm and curve fitting algorithm used in the present invention improve the requirements of the traditional Chebyshev fitting method for coherent sampling and high-precision signal sources. Its performance is significantly better than the traditional Chebyshev fitting method, and it can effectively suppress the spectral leakage problem caused by the phase discontinuity at the beginning and end of the sequence; when there is obvious clipping distortion in the sampling sequence, the fitting performance of the proposed algorithm is significantly better than the spectral parameter estimation method.

[0076] 3. The excitation noise removal algorithm used in the present invention reduces the conditional limit of the test on the signal source accuracy. According to the characteristics of Gaussian noise, this algorithm cancels external spurs such as the signal source and clock source through the sampling sequence of the same amplitude, and can obtain the ADC nonlinear error under the condition lower than the ADC's own accuracy. Description of the Drawings

[0077] Figure 1 It is a block diagram of the hardware system built by the preferred embodiment provided by the present invention.

[0078] Figure 2 It is a graph of the algorithm fitting performance of the present invention under non-coherent sampling conditions or over-range sampling conditions.

[0079] Figure 3The INL excitation noise removal performance graph of the present invention and the robustness graph of the proposed static test algorithm. Detailed implementation manners

[0080] Next, the technical solutions in the embodiments of the present invention will be clearly and detailedly described in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only a part of the embodiments of the present invention.

[0081] The technical solution for the present invention to solve the above technical problems is:

[0082] The present invention is an ADC static parameter test method applied to loose test conditions, and the sample points are from an ADC test system.

[0083] As shown in the attached Figure 1 figure, the ADC static parameter test system consists of 1, a low-precision signal source 2, an ADC test board 3, an FPGA carrier board 4, a PC 5, a DDR3 storage module 6, an Ethernet module 7, and a serial port module.

[0084] Among them, the FPGA carrier board serves as the core processor, which is used to control and drive peripheral modules, collect the ADC output sequence, and also provide the required sampling clock and power supply for the ADC.

[0085] The connection method is as follows: the clock sources generated by the low-precision signal source 1 and the FPGA carrier board 3 are input to the ADC test board 2 under test. The ADC test board 2, the serial port module 7, and the DDR3 module 5 are all connected to the FPGA carrier board 3. The Ethernet module 7 is respectively connected to the FPGA carrier board 3 and the PC terminal 4.

[0086] Its working process is as follows: the computer PC terminal transmits information to the FPGA carrier board through the Ethernet to configure the number of sampling points and the sampling frequency. After the FPGA receives the data transmitted by the PC terminal, it drives the ADC to perform sampling work, collects the ADC output sequence through the serial port, and stores the ADC output sequence in the DDR3. After the sampling is completed, the FPGA carrier board then sends the ADC output sequence to the PC terminal through the Ethernet.

[0087] After the PC terminal receives the ADC output sequence sent by the FPGA carrier board, it performs data processing and calculates the ADC static parameters. The steps are as follows:

[0088] S1. Collect the ADC output sequence and perform digital reconstruction to convert the output code value sequence into a discrete sequence:

[0089]

[0090] Among them, Δt = 1 / F s is the sampling point interval time.

[0091] S2. By adding a classical cosine window to the ADC output sequence, the continuity at the beginning and end of the sequence is increased, the energy concentration of the main lobe of the spectrum is further improved, and the amplitude estimation error caused by spectrum leakage is reduced. Here, I is the number of terms of the window function, a i is the coefficient of the i-th term of the window function, k r is the fundamental frequency index of the signal k r = L r = f / (F s × N). When M is large enough,

[0092] When the Hanning window I = 1 and a1 = 0.5, the obtained w(n) and X(k) are as follows:

[0093]

[0094] When the Hanning window I = 1 and a1 = 0.5, the obtained w(n) and X(k) are as follows:

[0095]

[0096] S3. Calculate the initial value f of the signal frequency through the ADC output frequency-domain expression int . Here, L m represents the spectral index value corresponding to the amplitude component with the largest spectrum. The discrete sequence k = 1, 2 ··· M corresponds to the spectral index numbers L1, L2 ··· L M :

[0097]

[0098] Since during non-coherent sampling, the spectral index L r corresponding to the fundamental frequency is non-integer and falls between the spectral lines L m-1 and L m+1 , the index difference η between each spectral line ∈ (-0.5, 0.5):

[0099] L r - L m = η

[0100] L r - L m-1 = η - 1

[0101] L r - L m+1 = η + 1

[0102] S4. Use the spectral line L m corresponding to the amplitude component with the largest spectrum max(|X(k)|) and its two adjacent spectral lines L m-1 and L m+1 to correct the fundamental frequency spectral line L r :

[0103]

[0104] S5. Solve the inverse function η = f -1 (χ) to estimate the amplitude, frequency, and phase information:

[0105]

[0106] S6. After solving η, further obtain the subsequent parameters by fitting the three parameters of the sine wave:

[0107]

[0108] S7. After the estimation is completed through the correction formula, establish the least - square difference relationship between the fitting sequence and the original sequence, and obtain the least - squares fitting value r in matrix form: r = (M

[0109] r = (M T M) -1 (M T x)

[0110]

[0111] S8. Solve the three - parameter fitting of the sine wave with the minimum residual to obtain the amplitude phase parameters:

[0112]

[0113] S9. Reconstruct the ideal clipped sequence according to the fitting parameters:

[0114]

[0115] S10. Obtain the residual sequence containing the ADC's own spurs, external excitation spurs, and fitting errors through the reconstructed ideal clipped sequence and the ADC output sequence:

[0116] r(n) = x(n) - x clip (n) = s adc (n) + s c (n) + s f (n)

[0117] where s adc is the spur component generated by the ADC's own design, including quantization noise, thermal noise, high - order nonlinearity, and other spur components; s c represents the influence of external excitation on the ADC output sequence, including spur components brought by external devices such as signal source accuracy and clock jitter; s f is the fitting error generated by the reconstructed sequence.

[0118] S11. The sampling transformation formula transforms the number of sampling points to enhance the first continuity of the residual sequence:

[0119]

[0120] Among them, J int is the integer part of the sampling period number J, M int is the number of data points contained in an integer period.

[0121] S12, Simultaneous Chebyshev's first polynomial C n (ξ)=cos(n·arccos(ξ)), establish the relationship between ADC input analog voltage x and ADC output digital code y(n):

[0122]

[0123] S13, using the A obtained by the preprocessed spectrum Y(k) n The ADC transfer curve g can be constructed s (x).

[0124]

[0125] g s (x) = g(x) + n q +n a +n f

[0126] Where g(x) is the ADC transfer curve under ideal test conditions, n q is the quantization error, n a is the stray introduced by external conditions, n f is the fitting error.

[0127] S14, when the sampling voltage is x and the output level is C, g s (x) can be obtained by the actual conversion level T(n), the ideal conversion level T i (n), code value C and INL nonlinear error are expressed as:

[0128]

[0129] T(0) is the transition level from code value 0 to 1. When the offset error is not considered, T(0) = T i (0). For the same ADC, T(0) has the same value.

[0130] S15, fitting curve g through two sets of sampling sequences s1 (x) and g s2(x) difference, cancels signal source noise, harmonics, fitting error, and external environment spurs, to obtain the ADC's own non-linear error:

[0131]

[0132] S16. By quantifying g with the voltage corresponding to the least significant bit in the ADC s1 (x) and g s2 (x) difference:

[0133]

[0134] The INL obtained after canceling the INL error caused by signal source spurs and inherent spurs is the ADC's own non-linear error:

[0135]

[0136] DNL = INL(C + 1) - INL(C)

[0137] Furthermore, the ADC test block diagram built according to the above algorithm is as Figure 1 shown, including: signal source, ADC test board, carrier board, PC, DDR3 memory module, Ethernet module, serial port module.

[0138] Furthermore, the FPGA carrier board, as the core processor, is used to control and drive peripheral modules, collect the ADC output sequence, provide sampling clock and power for the ADC; the test system configures the sampling points and sampling frequency through Ethernet. After the FPGA receives the data transmitted from the PC side, it drives the ADC to perform sampling work, then stores the ADC output sequence collected by the serial port into the DDR3. After sampling, the FPGA sends the data to the PC side through Ethernet, and the PC side completes the algorithms described in claims 1 to 5.

[0139] Furthermore, under the described ADC test system, when the input signal frequency is 999.98 Hz, the sampling rate is 200 MHz, the signal purity is 120 dB, the sampling points are 2048 points, and the signal amplitudes are 5 V and 5.05 V, the parameter fitting performance of the proposed algorithm is as Figure 2-1 and Figure 2-2As shown in the figure. It can be seen from the figure that when there is no clipping distortion in the sampling sequence, the fitting error of the proposed algorithm for the non-coherent sequence parameter fitting performance is larger than that of the spectral parameter estimation method, but its performance is significantly better than the traditional Chebyshev fitting method, and it can effectively suppress the spectral leakage problem caused by the discontinuous phase at the beginning and end of the sequence; when there is obvious clipping distortion in the sampling sequence, the fitting performance of the proposed algorithm is significantly better than the spectral parameter estimation method. Although the fitting performance gap between the two algorithms will decrease with the increase of the clipping degree, the fitting method for obtaining sequence parameters by FFT is not easily affected by noise and is more suitable for low-precision signal sources.

[0140] When the signal source accuracy is changed to 80 dB, the test results of different static parameter test algorithms are shown in Table 1. Figure 3-1 Shows the performance of the proposed algorithm in removing the excitation noise. In the figure, inl s Directly solve through the approximate continuous g s (x), which can only reflect the overall nonlinear trend of INL.

[0141] Table 1 Test results of different static parameter algorithms at 80 dB

[0142]

[0143] To verify the overall effectiveness and robustness of the proposed algorithm, 500 random simulation experiments were carried out in this section to simulate the external device environment, with the signal accuracy distributed in the range of 80 dB to 110 dB, the signal amplitude distributed in the range of 4.4 V to 5.2 V, and the spectral leakage size in the range of 0 to 0.5. The error diagrams of the simulation results of INL and DNL compared with the ideal parameters of a 16-bit ADC are as Figure 3-2 shown. Among them, the INL and DNL errors are concentrated in the range of 0 to 1 LSB, and the DNL is concentrated in the range of 0.5 to 1 LSB. In actual tests, the test points greater than 8000 can effectively reduce the influence of external random noise and increase the test accuracy.

[0144] The systems, devices, modules, or units described in the above embodiments can be specifically implemented by computer chips or entities, or by products with certain functions.

[0145] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0146] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0147] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A method for testing ADC static parameters under loose conditions, characterized in that: include: The steps include signal recognition and reconstruction, spectrum preprocessing, curve fitting, and excitation noise removal. The signal identification and reconstruction step includes: collecting the ADC output sequence, performing digital reconstruction to convert the output code value sequence into a discrete sequence, and obtaining the parameters of the ADC output sequence by spectrum triple interpolation and sine wave three-parameter fitting method; The spectrum preprocessing step includes: obtaining approximately coherent sampling points M by windowing and sampling point conversion. int ; The curve fitting step includes: enhancing the first continuity of the transformed data by using a Hanning window function, and then fitting the ADC curve by using a Chebyshev fitting algorithm; The excitation noise removal step includes: utilizing the noise distribution characteristics, reducing the influence of the signal source noise by subtracting the acquired data twice, removing the signal source noise error, and estimating the ADC-generated INL integral nonlinearity and DNL differential nonlinearity parameters.

2. The ADC static parameter testing method applied to loose test conditions according to claim 1, characterized in that: The signal recognition step and reconstruction specifically include: The ADC output sequence is collected and digitally reconstructed to convert the output code value sequence into a discrete sequence x(n). The input signal exceeds the ADC positive and negative full scale, and the ADC output sequence shows clipping and clipping distortion. V os are the amplitude, frequency, phase and DC bias of the input signal, hd is the high-order nonlinear distortion, w n is the sum of the noise, Δt=1 / F s is the sampling interval, F s Indicates the ADC sampling frequency, V fs Indicates the ADC full-scale voltage value: By adding a classic cosine window to the ADC output sequence, the continuity of the first position of the sequence is increased, the energy concentration of the main lobe of the spectrum is further improved, and the amplitude estimation error caused by spectrum leakage is reduced. Here, I is the number of window function terms, and a is i is the coefficient of the i-th term of the window function, k r is the signal fundamental frequency index k r =f / (F S ×N), N represents the number of sampling points, and the discrete sequence k=1,2···M; when M is large enough, When the Hanning window I=1, a1=0.5, the obtained w(n) and X(k) are as follows: When the Hanning window I = 1, a1 = 0.5, the obtained w(n) and X(k) are as follows: Calculate the initial value of the signal frequency f through the ADC output frequency domain expression int , where L m Indicates the spectral line index value corresponding to the maximum amplitude component of the spectrum, the discrete sequence k = 1, 2...M and the spectrum index number L1, L2...L M correspond: Due to the incoherent sampling, the spectrum index L corresponding to the fundamental frequency r is a non-integer, falling in L m-1 and L m+1 Between the spectral lines, L m-1 ,L m ,L m+1 With L r The relationship between the spectral lines is as follows, where the index difference between the spectral lines is η∈(-0.5,0.5): L r -L m =the L r -L m-1 =η-1 L r -L m+1 =η+1 Use the maximum amplitude component max(|X(k)|) of the spectrum corresponding to the spectrum line L m and its left and right lines L m-1 and L m+1 For the fundamental frequency spectrum line L r To make corrections: Directly set the frequency correction formula to solve η and then use the three-parameter fitting of the sine wave to further obtain the subsequent parameters: Completed by modifying the formula After estimation, the minimum square difference E between the fitted sequence and the original sequence is established, and the least square fitting value r is obtained in matrix form, where To fit the DC bias, To fit the cosine amplitude, To fit the sine amplitude: r=(M T M) -1 (M T x) Solve the three-parameter sine wave fitting with the minimum residual to obtain the amplitude Phase parameter: Reconstruct the ideal clipping sequence based on the fitted parameters:

3. The ADC static parameter testing method applied to loose test conditions according to claim 2, characterized in that: The spectrum preprocessing step comprises: The residual sequence including ADC's own spurious signals, external excitation spurious signals, and fitting errors is obtained through the reconstructed ideal clipping sequence and ADC output sequence: r(n)=x(n)-x clip (n)=s adc (n)+s c (n)+s f (n) Among them, s adc It is the spurious component generated by the ADC's own design, including quantization noise, thermal noise, high-order nonlinearity and other spurious components; c Represents the impact of external excitation on the ADC output sequence, including spurious components caused by external devices such as signal source accuracy and clock jitter; f is the fitting error caused by the reconstructed sequence; Considering that the residual sequence r(n) of the clipped input signal has a periodic law, the sampling transformation formula transforms the number of sampling points to enhance the first continuity of the residual sequence: Among them, J int is the integer part of the sampling period number J, M int is the number of data points contained in an integer period.

4. The ADC static parameter testing method applied to loose test conditions according to claim 3, characterized in that: The curve fitting step comprises: When the ADC test does not meet the marked test conditions, the ADC output sequence is expressed in the form of x(n): By combining the first Chebyshev polynomial C n (ξ)=cos(n·arccos(ξ)), establish the relationship between the ADC input analog voltage x and the ADC output digital code y(n), and obtain the ADC transfer curve g(x) under ideal test conditions: Using the A obtained by preprocessing the spectrum Y(k) n The ADC transfer curve g can be constructed s (x) When the input signal amplitude exceeds the full scale of the ADC, limiting the harmonic amplitude calculation range to even harmonics can reduce the impact of additional spurious signals introduced by clipping distortion on the ADC transfer curve fitting. At this time, the measured INL and DNL include not only the ADC's own spurious signals, but also the spurious components introduced by factors such as signal source accuracy, clock jitter, and fitting errors, and cannot directly reflect the ADC's static performance: g s (x)=g(x)+n q +n a +n f Among them, n q is the quantization error, n a is the stray introduced by external conditions, n f is the fitting error; When the sampling voltage is x and the output level is C, g s (x) can be obtained by the actual conversion level T(n), the ideal conversion level T i (n), code value C and INL nonlinear error, where V LSB Indicates the voltage corresponding to the least significant bit in the ADC: T(0) is the transition level from code value 0 to 1. When the offset error is not considered, T(0) = T i (0), for the same ADC, T(0) is the same value.

5. The ADC static parameter testing method applied to loose test conditions according to claim 1, characterized in that: The excitation noise removal step comprises: The fitting curve g of the two sets of sampling sequences s1 (x) and g s2 (x), offset the signal source noise and harmonics, fitting error and external environment spurious, and obtain the nonlinear error of the ADC itself; where C1 is the first segment of the fitting curve g s1 (x) Output code value C, INL1(C) is the integral nonlinear error of the first sequence at code value C, C2 is the fitting curve g of the second sequence s2 (x) Output code value C, INL2(C) is the integral nonlinear error of the second sequence at code value C: By using the voltage V corresponding to the least significant bit in the ADC LSB Quantization g s1 (x) and g s2 Difference of (x): The INL error caused by the signal source spurious and inherent spurious is offset, and the resulting INL is the ADC's own nonlinear error: DNL = INL(C+1) - INL(C).

6. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the ADC static parameter testing method applied to loose test conditions as claimed in any one of claims 1 to 5 is implemented.

7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the ADC static parameter testing method applied to relaxed test conditions as claimed in any one of claims 1 to 5 is implemented.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the ADC static parameter testing method applied to relaxed test conditions as claimed in any one of claims 1 to 5 is implemented.

Images