Time interleaved sampling system calibration method based on periodic time-varying filter and computer device

Abstract

The present invention relates to the field of time-interleaved sampling signal correction, and specifically to a time-interleaved sampling system calibration method and computer device based on a periodic time-varying filter, which realizes the error estimation and correction of broadband time-interleaved sampling of arbitrary channel frequency response. The sampling inconsistency error is greatly reduced. The technical scheme includes: first inputting sinusoidal swept frequency signals of different frequencies, then performing correlation analysis on the signals spliced ​​by the broadband time-interleaved system, using Fourier transform to obtain the amplitude and phase of each channel acquisition signal at different frequencies, setting channel 1 as an ideal channel, calculating the relative gain error coefficient and the relative delay error coefficient, estimating the relative frequency response and the frequency domain coefficient of the periodic time-varying filter according to the formula, and finally performing frequency domain filtering correction. The present invention is suitable for correcting the errors of broadband time-interleaved sampling systems.

Description

Technical Field

[0001] The invention relates to the field of time interleaved sampling signal correction, and in particular to a time interleaved sampling system calibration method based on a periodic time-varying filter and a computer device. Background Art

[0002] Oscilloscopes are important instruments for analyzing high-speed signals. They play an important role in optical performance testing, ultrafast phenomenon research, and high-speed optical communication network testing. At present, due to the limitations of ADC (Analog-to-Digital Converter) integration process, the highest indicators of single-chip ADCs in mature commercial products are only 10.4GSa / s sampling rate, 8GHz bandwidth, and 12-bit effective bits, which is difficult to meet the growing design requirements of high sampling rate and high bandwidth of acquisition systems. The data acquisition method based on the time-alternating TI sampling architecture is an effective means to increase the sampling rate of the system. This technology uses M parallel sub-ADCs for alternating sampling. The sampling rate of each sub-ADC is , at the digital end, M sampling data streams are combined into one sampling data stream in chronological order. At this time, the sampling rate of the sampling data stream becomes The sampling architecture based on time interleaving breaks through the limitation of low sampling rate of single ADC and realizes high-speed digital acquisition.

[0003] However, the inconsistent physical parameters of each acquisition channel of the time-interleaved sampling system lead to sampling mismatch errors in the sampling results, which greatly reduces the performance of the system. In order to improve the performance of the system, it is necessary to eliminate the errors of the acquisition system to achieve accurate reconstruction of the original signal. The traditional error model adopts a fixed adaptive error model, which is based on the assumption that the gain error coefficient, time delay error coefficient and offset error coefficient do not change with the change of the signal input frequency. However, for the broadband acquisition system, there are differences in the frequency response of each acquisition channel and the sampling time of different ADCs is offset from the ideal sampling time. Therefore, the fixed adaptive error model is not suitable for error estimation and correction of broadband acquisition systems. In order to solve the error correction problem of broadband time-interleaved sampling systems, related methods such as random alternating sampling and bandwidth mismatch models have been proposed one after another. These methods can solve the sampling inconsistency error correction problem of special time-interleaved systems, but cannot meet the error estimation and correction requirements of broadband time-interleaved acquisition systems with arbitrary number of channels and arbitrary channel frequency response. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a time interleaved sampling system calibration method and computer device based on a periodic time-varying filter, which realizes the error estimation and correction of broadband time interleaved sampling of arbitrary channel frequency response.

[0005] The present invention adopts the following technical solutions to achieve the above-mentioned purpose. In a first aspect, the present invention provides a time interleaved sampling system calibration method based on a periodic time-varying filter, comprising:

[0006] S1, sine signal frequency sweep;

[0007] Perform sinusoidal signal frequency sweep, collect the signal after the time interleaved sampling system is combined, the signal source inputs sinusoidal signals of different frequencies, and divides the frequency into N The corresponding frequency is , the sweep frequency step interval is , N Represents an integer greater than 0, represents the set of corresponding frequency points, Respectively, the frequency is divided into N The frequency points corresponding to the equal parts;

[0008] S2, frequency response estimation;

[0009] Calculation of measured frequency response of each channel:

[0010] The amplitude and phase of the collected signals of each channel at different frequencies are obtained by fast Fourier transform. If the time-interleaved sampling system has M channels, the acquisition frequency of the mth channel is The output signal of the sinusoidal signal after being sampled by a single ADC is ,right conduct K Fourier transform of the point to get its spectrum ,set up K The frequency domain of the signal after Fourier transform is:

[0011]

[0012] in Respectively indicate that the signal passes K The results after point Fourier transform at the 1st, 2nd, ..., K-1th points;

[0013] If the length of the single frequency signal is L , then ,but K The number of signal periods P contained in the Fourier transform of the point is , if P is an integer, then the signal frequency The frequency resolution Integer multiples of , no spectrum leakage will occur at this time, M represents the total number of channels, which is an integer greater than 0, m represents the specific number of channels, which is an integer greater than 0, and P represents the number of signal cycles. Indicates the sampling rate;

[0014] set up The The amplitude-frequency value and phase-frequency value at the point are the frequency The measured amplitude and phase values ​​of ,like for The simulated angular frequency corresponding to the position is The DC bias, amplitude and phase of , and ;

[0015] Calculation of relative frequency response of each channel:

[0016] Assume that channel 1 is an ideal channel, calculate the relative gain error coefficient and relative delay error coefficient of other channels relative to channel 1, and the simulated angular frequency of channel m is: The relative gain error coefficient is expressed as:

[0017] , represents the relative gain error coefficient, Indicates the DC bias of channel 1;

[0018] And the relative delay error coefficient:

[0019] , represents the relative delay error coefficient, represents the phase of channel 1, where represents the sampling period, ;

[0020] Then the simulated angular frequency of the mth channel is The relative frequency response at is expressed as:

[0021] ;

[0022] The relative frequency response of the sweep frequency of the mth channel is

[0023] ,in , is an imaginary unit, Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at

[0024] S3, frequency domain filtering correction of each channel;

[0025] Correction parameter calculation:

[0026] The error correction structure based on the periodic time-varying filter is adopted, and the relative frequency response of channel 1 is taken as the ideal relative frequency response. ,Right now , Indicates the relative frequency response of channel 1;

[0027] According to the channel frequency response decomposition properties of the time-interleaved sampling model, the frequency domain coefficients of the mth channel based on the periodic time-varying filter are It is expressed as:

[0028] ;

[0029] Full-band periodic time-varying filter frequency response data acquisition:

[0030] The frequency domain coefficients of the periodic time-varying filter Add several values ​​equal to 1 to each unilateral edge to obtain the full-band periodic time-varying filter frequency response data after completion ;

[0031] Assume that the full-band frequency sequence is , Indicates the full-band frequency sequence, using the full-band frequency sequence and the sweep interval frequency And its corresponding periodic time-varying filter frequency response data The data is resampled to obtain the frequency response data of the periodic time-varying filter at the specified frequency point;

[0032] The initial channel frequency response is interpolated using the cubic spline interpolation method. Fit to the full-band frequency sequence length to obtain the full-band periodic time-varying filter frequency domain coefficients , and finally perform filtering correction.

[0033] Furthermore, the filtering correction specifically includes:

[0034] The original data frame of each channel is subtracted from the corresponding mean value, and then the acquired signal is divided into multiple small segments, each of which is a data frame. The data frame is converted from the time domain to the frequency domain by performing an FFT with a point number of 1024 to obtain the pre-processed data, and then Perform point multiplication to complete frequency domain filtering, perform IFFT transformation on the filtered data frame to obtain the corrected data frames, then add the corresponding mean to the corrected data frame for compensation, and finally combine the compensated time domain signals of each channel to obtain the output signal .

[0035] Furthermore, after filtering correction is performed, step S3 further includes:

[0036] Perform coefficient iteration based on a periodic time-varying filter:

[0037] The first calculated As the primary correction factor , then order As input Repeat steps S2 to S3, and the correction coefficient obtained in the cth iteration is .

[0038] In a second aspect, the present invention provides a computer device, including a memory, wherein the memory stores program instructions, and when the program instructions are executed, the time interleaved sampling system calibration method based on periodic time-varying filter as described above is executed.

[0039] The beneficial effects of the present invention are:

[0040] The present invention adopts the strategy of first estimating and then correcting, that is, first using the data measured by the standard signal source to estimate the error, including gain error, bias error and time error. In particular, the accuracy of the error estimation determines the quality of the correction effect to a large extent. On this basis, the correction parameters of the system, namely the frequency domain coefficients of the periodic time-varying filter, are calculated, and then the system uses the obtained correction parameters to correct the sampling results during normal input, thereby realizing accurate reconstruction of the input signal. Specifically, first input sinusoidal swept frequency signals of different frequency points, and then perform correlation analysis on the signals spliced ​​by the broadband time interleaving system, use Fourier transform to obtain the amplitude and phase of the collected signals of each channel at different frequency points, set channel 1 as the ideal channel, calculate the relative gain error coefficient and the relative delay error coefficient, and estimate the relative frequency response and the frequency domain coefficients of the periodic time-varying filter according to the formula, and finally perform frequency domain filtering correction. The error estimation and correction of broadband time interleaved sampling of the frequency response of any channel is realized. The sampling inconsistency error is greatly reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 is a flow chart of a time interleaved sampling system calibration method based on a periodic time-varying filter provided by an embodiment of the present invention;

[0042] Figure 2 Schematic diagram of relative gain error coefficients and delay error coefficients of channel 1 and channel 2 provided in an embodiment of the present invention;

[0043] Figure 3 Schematic diagram of relative amplitude-frequency response and relative phase-frequency response of channel 1 and channel 2 in the entire 18 GHz bandwidth provided by an embodiment of the present invention;

[0044] Figure 4 It is a schematic diagram of a spectrum before compensation, a spectrum after one iteration, and a spectrum after ten iterations provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0046] The present invention proposes a time interleaved sampling system calibration method based on a periodic time-varying filter, such as Figure 1 As shown, the specific steps include:

[0047] S1, sine signal sweep frequency input;

[0048] Sine frequency sweep is performed to collect the signal after being spliced ​​by the broadband time interleaving system. Specifically, the signal source inputs sine frequency sweep signals with different frequencies, and the frequency is divided into N The corresponding frequency is , the sweep frequency step interval is , N Represents an integer greater than 0, represents the set of corresponding frequency points, Respectively, the frequency is divided into N The frequency points corresponding to the equal parts.

[0049] S2, frequency response estimation;

[0050] Calculation of measured frequency response of each channel:

[0051] The amplitude and phase of the collected signals of each channel at different frequencies are obtained by fast Fourier transform. If the time interleaved sampling system has M channels in total, the acquisition frequency for the mth channel is The output signal of the sinusoidal signal after being sampled by a single ADC is ,right conduct K Fourier transform of the point to get its spectrum ,set up K The frequency domain of the signal after Fourier transform is:

[0052]

[0053] in Respectively indicate that the signal passes K The results after point Fourier transform at the 1st, 2nd, ..., K-1th points;

[0054] If the length of the single frequency signal is L , then ,but K The number of signal periods P contained in the Fourier transform of the point is , if P is an integer, then the signal frequency The frequency resolution Integer multiples of , no spectrum leakage will occur at this time, M represents the total number of channels, which is an integer greater than 0, m represents the specific number of channels, which is an integer greater than 0, and P represents the number of signal cycles. Indicates the sampling rate;

[0055] set up The The amplitude-frequency value and phase-frequency value at the point are the frequency The measured amplitude and phase values ​​of ,like for The simulated angular frequency corresponding to the position is The DC bias, amplitude and phase of , and ;

[0056] Calculation of relative frequency response of each channel:

[0057] Assume that channel 1 is an ideal channel, calculate the relative error coefficient and relative delay error coefficient of other channels relative to channel 1, and the simulated angular frequency of channel m is The relative gain error coefficient can be expressed as:

[0058] , represents the relative gain error coefficient, Indicates the DC bias of channel 1;

[0059] And the relative delay error coefficient:

[0060] , represents the relative delay error coefficient, represents the phase of channel 1, where represents the sampling period, ;

[0061] Then the relative frequency response of the mth channel can be expressed as:

[0062]

[0063] In S1, the sweep frequency is There are N in total, so the relative frequency response of the sweep frequency of the mth channel is

[0064] .in , is an imaginary unit, Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at .

[0065] like Figure 2 The figure shows the relative gain error coefficient and delay error coefficient of the two channels. It can be seen from the figure that the relative gain error coefficient and delay error coefficient of 1.25GHz change dramatically relative to the adjacent frequency points. This is because the signal frequency of 1.25GHz is equal to half of the single ADC sampling frequency, which will cause the relative frequency response estimation error. In order to avoid this problem, the frequency response value corresponding to this frequency point is equal to the average of the frequency response values ​​of its two adjacent frequency points.

[0066] TI calibration method based on periodic time-varying filter:

[0067] Since the amplitude frequency and phase frequency of the relative frequency response fluctuate greatly between adjacent frequencies, it is difficult to implement them using a low-order time domain filter. Therefore, frequency domain filtering is used for error correction.

[0068] Correction parameter calculation:

[0069] By using an error correction structure based on a periodic time-varying filter, the relative frequency response of channel 1 can be taken as the ideal relative frequency response. ,Right now The calculation formula is:

[0070] , Indicates the relative frequency response of channel 1;

[0071] According to the channel frequency response decomposition properties of the time-interleaved sampling model, the frequency domain coefficients of the mth channel based on the periodic time-varying filter are It can be expressed as:

[0072]

[0073] Full-band periodic time-varying filter frequency response data acquisition:

[0074] The above steps only obtain the frequency domain coefficients of the periodic time-varying filter within the passband , and needs to be extended to the full frequency band. Add several values ​​equal to 1 to each unilateral edge to obtain the full-band periodic time-varying filter frequency response data after completion .

[0075] Assume that the full-band frequency sequence is , Indicates the full-band frequency sequence. Using the full-band frequency sequence and the sweep interval frequency And its corresponding periodic time-varying filter frequency response data The data is resampled to obtain the frequency response data of the periodic time-varying filter at the specified frequency point. The present invention uses the cubic spline interpolation method to interpolate the initial channel frequency response. Fit to the full-band frequency sequence length to obtain the full-band periodic time-varying filter frequency domain coefficients .

[0076] Error frequency domain filtering correction:

[0077] Since the DC bias of each data frame will affect the correction result, the correction also needs to remove the DC bias by subtracting the corresponding mean from the original data frame data. Then the acquired spliced ​​signal is divided into several small segments, each of which is called a data frame. The FFT point number is 1024 to convert the data frame from the time domain to the frequency domain to obtain the processed data. The pre-processed data frame is converted to the frequency domain and then Perform point multiplication to complete frequency domain filtering. Perform IFFT transformation on the filtered signal to obtain the corrected data frames, and add the corresponding mean value previously cached to the data frame to restore the DC component of the original signal. Finally, the compensated time domain signals of each channel are combined to obtain the corrected output signal of the TIADC system. .

[0078] Coefficient iteration based on periodic time-varying filter:

[0079] Since the original signal spectrum cannot be obtained in actual sampling, it is used Therefore, the data corrected by this coefficient is There is still a correction error. The correction error energy can be reduced by iteratively calculating the update coefficients. The coefficient iteration steps are as follows:

[0080] The first calculated As the primary correction factor , then order As input Repeat steps 2-3, and the correction coefficient obtained in the cth iteration is , and the iteration ends after 10 iterations.

[0081] The present invention solves the problem that the calibration algorithm of the traditional time-interleaved sampling system only calculates the error of the single-tone signal while the frequency response difference of the actual system is inconsistent through the above method. At the same time, it also solves the problem that the calibration algorithm of the time-interleaved sampling system is performed on a single-chip ADC and out-of-band compensation is difficult to implement.

[0082] The following is a detailed description of the technical effects of the present invention in combination with actual engineering tests:

[0083] 1. Test conditions and contents

[0084] The experimental equipment is a high-speed digital storage oscilloscope with a sampling rate of 80GSPS and a bandwidth of 20GHz, which is built based on a dual-channel broadband time-interleaved sampling system, and a signal generator that generates a sinusoidal signal with a frequency range of 20GHz, where the sampling rate of the sub-ADC is 2.5GSa / s. The signal source generates a sinusoidal sweep signal of equal amplitude and inputs it into the oscilloscope, with a frequency range of 100MHz~18GHz and a frequency interval of 100MHz. The oscilloscope collects the output signal of the system.

[0085] 2. Test result analysis

[0086] Reference Figure 3 , the relative amplitude-frequency response and relative phase-frequency response of channel 1 and channel 2 in the entire 18 GHz bandwidth are plotted. From the estimation results, it can be seen that the error coefficient of the amplitude-frequency response of the relative frequency response at adjacent frequency points changes relatively smoothly, while the error coefficient of the phase-frequency response at adjacent frequency points changes more dramatically, which is difficult to approximate with a low-order filter. Therefore, the present invention adopts frequency domain filtering to perform error correction.

[0087] Reference Figure 4 , the signal source inputs a single-tone signal of 18GHz, uses the calibration coefficients of one iteration and ten iterations, and performs frequency domain correction on the output signal passing through the acquisition system to obtain the compensated output signal. The spectrum diagram of the signal before and after correction is drawn. It can be seen from the spectrum diagram before and after correction that for the single-tone signal, there are more sampling inconsistency error spectra in the acquisition system. The error spectrum generated by sampling inconsistency before calibration is greater than the noise spectrum, and after frequency domain correction, the error spectrum is reduced to below the noise spectrum, and the sampling inconsistency error after correction is greatly reduced, indicating that the present invention can well realize sampling inconsistency error calibration.

[0088] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used in various other combinations, modifications and environments, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A time-interleaved sampling system calibration method based on a periodic time-varying filter, characterized in that: include: S1, sine signal frequency sweep; Perform sinusoidal signal frequency sweep, collect the signal after the time interleaved sampling system is combined, the signal source inputs sinusoidal signals of different frequencies, and divides the frequency into N The corresponding frequency is , the sweep frequency step interval is , N Represents an integer greater than 0, represents the set of corresponding frequency points, Respectively, the frequency is divided into N The frequency points corresponding to the equal parts; S2, frequency response estimation; Calculation of measured frequency response of each channel: The amplitude and phase of the collected signals of each channel at different frequencies are obtained by fast Fourier transform. If the time-interleaved sampling system has M channels, the acquisition frequency of the mth channel is The output signal of the sinusoidal signal after being sampled by a single ADC is ,right conduct K Fourier transform of the point to get its spectrum ,set up K The frequency domain of the signal after Fourier transform is: in Respectively indicate that the signal passes K The results after point Fourier transform at the 1st, 2nd, ..., K-1th points; If the length of the single frequency signal is L , then ,but K The number of signal periods P contained in the Fourier transform of the point is , if P is an integer, then the signal frequency The frequency resolution Integer multiples of , no spectrum leakage will occur at this time, M represents the total number of channels, which is an integer greater than 0, m represents the specific number of channels, which is an integer greater than 0, and P represents the number of signal cycles. Indicates the sampling rate; set up The The amplitude-frequency value and phase-frequency value at the point are the frequency The measured amplitude and phase values ​​of ,like for The simulated angular frequency corresponding to the position is The DC bias, amplitude and phase of , and ; Calculation of relative frequency response of each channel: Assume that channel 1 is an ideal channel, calculate the relative gain error coefficient and relative delay error coefficient of other channels relative to channel 1, and the simulated angular frequency of channel m is: The relative gain error coefficient is expressed as: , represents the relative gain error coefficient, Indicates the DC bias of channel 1; And the relative delay error coefficient: , represents the relative delay error coefficient, represents the phase of channel 1, where represents the sampling period, ; Then the simulated angular frequency of the mth channel is The relative frequency response is expressed as: ； The relative frequency response of the sweep frequency of the mth channel is ,in , is an imaginary unit, Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at Indicates the frequency The relative frequency response at S3, frequency domain filtering correction of each channel; Correction parameter calculation: The error correction structure based on the periodic time-varying filter is adopted, and the relative frequency response of channel 1 is taken as the ideal relative frequency response. ,Right now , Indicates the relative frequency response of channel 1; According to the channel frequency response decomposition properties of the time-interleaved sampling model, the frequency domain coefficients of the mth channel based on the periodic time-varying filter are It is expressed as: ； Full-band periodic time-varying filter frequency response data acquisition: The frequency domain coefficients of the periodic time-varying filter Add multiple values ​​equal to 1 to each unilateral edge to obtain the full-band periodic time-varying filter frequency response data after completion ; Assume that the full-band frequency sequence is , Indicates the full-band frequency sequence, using the full-band frequency sequence and the sweep interval frequency And its corresponding periodic time-varying filter frequency response data The data is resampled to obtain the frequency response data of the periodic time-varying filter at the specified frequency point; The initial channel frequency response is interpolated using the cubic spline interpolation method. Fit to the full-band frequency sequence length to obtain the full-band periodic time-varying filter frequency domain coefficients , and finally filter correction is performed; Filter correction specifically includes: The original data frame of each channel is subtracted from the corresponding mean value, and then the acquired signal is divided into multiple small segments, each of which is a data frame. The data frame is converted from the time domain to the frequency domain by performing an FFT with a point number of 1024 to obtain the pre-processed data, and then Perform point multiplication to complete frequency domain filtering, perform IFFT transformation on the filtered data frame to obtain the corrected data frames, then add the corresponding mean to the corrected data frame for compensation, and finally combine the compensated time domain signals of each channel to obtain the output signal .

2. The time interleaved sampling system calibration method based on periodic time-varying filter according to claim 1 is characterized in that: After filtering correction is performed, step S3 further includes: Perform coefficient iteration based on a periodic time-varying filter: The first calculated As a primary correction factor , then order As input Repeat steps S2 to S3, and the correction coefficient obtained in the cth iteration is .

3. A computer device, comprising a memory, wherein the memory stores program instructions, characterized in that: When the program instructions are executed, the time-interleaved sampling system calibration method based on a periodic time-varying filter as described in any one of claims 1 to 2 is executed.

Images