A high-precision measurement method for multi-channel high-speed ADC acquisition

Abstract

A kind of high-precision measurement method of multi-channel high-speed ADC acquisition, comprising: based on the measurement environment of baseband sampling and bandpass sampling and 1GHz sampling frequency, using 1KHz~800MHz sampling signal to test;The input sampling signal is divided into two signals by power divider, wherein, when measuring two channels of the same ADC chip of N ADC sampling circuits, two signals are respectively connected to the input end of two channels of the ADC chip;When testing two channels of different chips of ADC chip, two signals are respectively connected to the input end of one channel of two ADC chips;Test sampling, a piece of data is grabbed using vivado software, and different analysis methods are selected according to the grabbed data to calculate channel amplitude, channel DC offset calculation, channel delay calculation. Channel amplitude, channel DC offset and layout consistency measurement in multi-channel ADC test are realized.

Description

Technical Field

[0001] This application relates to the field of electrical technology, specifically to the application technology of analog or digital signal acquisition in software radio technology, and in particular to a high-precision measurement method for multi-channel high-speed ADC acquisition. Background Technology

[0002] In digital processing systems, a multi-channel high-speed ADC circuit is needed to sample data, and the acquired data needs to be digitally down-converted. However, due to the increased number of sampling circuit channels, and the influence of environmental factors such as PCB layout differences and clock jitter, channel inconsistencies can occur between channels within the same ADC chip. This is especially true when multiple ADC chips are operating simultaneously, resulting in two types of channel inconsistencies: one between the two acquisition circuits of the same ADC, and the other between the acquisition circuits of different ADC chips. These inconsistencies increase the error in subsequent signal processing and reduce the overall system accuracy. Therefore, sampling channel consistency is a very important issue.

[0003] Since most of the designs used in engineering projects employ multi-channel ADC chips, before testing, if the conversion rate, power consumption, data and control interface, power supply error, etc. of each ADC chip are not significantly different and meet the test requirements, then the sampling differences between each channel are related to several key specifications; among them, the more important indicators that have a greater impact on subsequent data processing are: channel amplitude, channel DC offset, and PCB layout and routing consistency. Summary of the Invention

[0004] This application provides a high-precision measurement method for multi-channel high-speed ADC acquisition, enabling the measurement of channel amplitude, channel DC offset, and layout and routing consistency in multi-channel ADC testing.

[0005] To achieve the above objectives, the present invention employs the following techniques:

[0006] A high-precision measurement method for multi-channel high-speed ADC acquisition includes the following steps:

[0007] The S100 was tested using a sampling signal ranging from 1kHz to 800MHz, based on a measurement environment using baseband sampling and bandpass sampling, and a sampling frequency of 1GHz.

[0008] S200. Prepare the N-channel ADC sampling circuit to be tested, which is located on N / 2 ADC chips, where N is a positive even number;

[0009] S300: The input sampling signal is split into two signals by a power divider. When measuring two channels of the same ADC chip in an N-channel ADC sampling circuit, the two signals are respectively connected to the two channel input terminals of the ADC chip. When testing two channels of different ADC chips, the two signals are respectively connected to one channel input terminal of each of the two ADC chips.

[0010] S400, Start test sampling, use Vivado software to capture multiple frames of sample data, and perform channel amplitude calculation, channel DC offset calculation, and channel delay calculation based on the captured data;

[0011] Among them, the channel amplitude calculation adopts the FFT analysis method to convert the signal from the time domain to the frequency domain and extract the amplitude information from the frequency domain. Let the amplitude values ​​obtained from the frequency domain of the two signals be A1 and A2, respectively. The amplitude consistency is calculated by the following formula: ΔA=20log(A2 / A1);

[0012] In the DC offset calculation of the channel, the data of the two signals are normalized and then the average value is calculated to obtain V1 and V2. The DC offset consistency is calculated by the following formula: ΔDC=V2-V1;

[0013] Channel delay is calculated using either the FFT method or the correlation method, where:

[0014] In the FFT method, the correspondence between time-domain delay and frequency-domain phase is first used, that is: x(t-τ)←→X(j2πf)exp(-2πfτ), to convert the sampled signal to frequency domain analysis, where x() is the time-domain signal and X() is the frequency-domain signal. Then, based on the spectrum of the sampled signal, the phase of the point where the maximum value can be read in the spectrum is found. If the positive spectrum phase φ+ is read, the delay can be obtained as: τ=φ / -wf, where w=2π.

[0015] In the relevant method, it is calculated according to the following formula:

[0016]

[0017] x 1( t )and x 2( t These are two signals with respect to time shift. t It is a periodic function.

[0018] The beneficial effects of this invention are as follows:

[0019] 1. The test results show that the consistency of channel amplitude, channel DC offset and layout routing in multi-channel ADC sampling has achieved a high level of accuracy in controlling the impact of latency.

[0020] 2. This measurement method is applicable to the analysis of various characteristic signals, including sine waves, triangular waves, square waves, etc. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the implementation methods of the present invention will be described in detail below. However, the embodiments described in the present invention are only some embodiments of the present invention, and not all embodiments.

[0022] This application provides a high-precision measurement method for multi-channel high-speed ADC acquisition, including the following steps:

[0023] S100. Considering that the background of this measurement is baseband sampling and bandpass sampling, and the sampling frequency in the test is 1GHz, the frequency range of the sampled signal taken in this embodiment of the application is from 1KHz to 800 MHz, so as to fully test the consistency of each ADC sampling circuit under different boundary conditions.

[0024] S200. During the test, a total of 8 ADC sampling circuits were prepared, which were respectively on 4 AD9680 chips. The AD96804 channel has 1Gsps. In order to measure more accurately, this embodiment of the application measured 4 sets of data from two channels of the same ADC chip and two channels of different ADC chips in the 8 sampling circuits for comparison and analysis, and compared their differences.

[0025] S300: The input sampling signal is split into two signals by a power divider. When measuring two channels of the same ADC chip in an N-channel ADC sampling circuit, the two signals are connected to the two channel input terminals of the ADC chip respectively. When testing two channels of different ADC chips, the two signals are connected to one channel input terminal of each of the two ADC chips respectively. This eliminates the influence of wires on consistency testing in the later processing.

[0026] S400, start test sampling, use Vivado software to capture a piece of data / multiple frames of data, and use MATLAB to calculate channel amplitude, channel DC offset, and channel delay based on the captured data.

[0027] (1) Channel amplitude calculation:

[0028] Channel amplitude is a very important indicator for ADC chips, and amplitude error will reduce SFDR. If there are significant differences in amplitude error among the channels in multiple ADC channel sampling circuits, it will have a very significant impact on back-end processing.

[0029] When the sampling rate is low and noise is present, it is impossible to extract the amplitude of a sinusoidal signal with high precision from the time domain. This is because the sampled signal is discrete, and it cannot be guaranteed that the maximum value of the input signal can be captured; therefore, accurate amplitude information cannot be extracted from the time domain. According to Pascal's theorem, the energy in the time domain and the energy in the frequency domain of a signal are the same. Therefore, this embodiment uses FFT analysis to convert the signal to the frequency domain and extract amplitude information from it. Let the amplitude values ​​obtained from the frequency domain for the two signals be A1 and A2, respectively. Then the amplitude consistency is: ΔA = 20log(A2 / A1).

[0030] Because the FFT result is a discretized spectrum, to obtain accurate amplitude information, it is necessary to ensure the correctness of the spectrum sampling, i.e., to avoid spectral leakage. This requires the analysis length of the signal to be an integer period; at the same time, the analysis lengths of the two channels must be the same to ensure that the FFT gain is the same. Specifically, an integer period refers to one clock cycle. For example, if the intermediate frequency signal is 10MHz and the sampling clock is 500MHz, then the number of sampling points in one clock cycle is 50. If a frame of data, excluding the frame header and frame tail, contains 1024 bytes, and 1 byte is 8 bits, then 1 bit is one point.

[0031] (2) Calculation of DC offset of channel:

[0032] The testing begins with the DC characteristics of the ADC, because there are various non-standard methods for testing the AC parameters of ADCs, and comparing two integrated ADC chips is easier based on DC characteristics. DC characteristics are generally more reflective of device issues than AC characteristics. Therefore, the first consideration for consistency is the consistency of DC offset.

[0033] In this embodiment, the consideration of DC offset consistency is to normalize the sampled dual-channel data and then calculate the average of the signals of the two channels to obtain V1 and V2 respectively, thus obtaining the DC offset consistency: ΔDC = V2 - V1.

[0034] (3) Impact of layout and cabling (channel delay calculation):

[0035] For ADC sampling circuits, the consistency of delay affects the correlation processing results, especially in undersampling techniques, where the delay generated by the circuit is more likely to affect subsequent stages. Therefore, this application embodiment focuses on the consistency of delays of multiple ADC sampling channels and tests it, employing two methods to analyze channel delays: FFT and correlation. These two methods are essentially the same; however, correlation analysis requires more time-shifting and correlation operations to improve analysis accuracy, resulting in a slower speed.

[0036] Among them, the FFT method has the advantage of fast analysis speed because the frequency domain characteristics of the sampled signal can be intuitively seen after the FFT is applied. Compared with the correlation method, the FFT method is faster and more intuitive.

[0037] In the FFT method, the embodiments of this application first utilize the correspondence between time-domain delay and frequency-domain phase, i.e.: x(t-τ)←→X(j2πf)exp(-2πfτ), to convert the sampled signal to frequency domain analysis; then, based on the spectrum of the sampled signal, find the phase of the maximum value point in the spectrum; if the positive spectrum phase φ+ is read, the delay amount can be obtained as: τ=φ / -wf, where w=2π.

[0038] In the correlation method, the correlation between two signals is a function of the time shift *t* between them. For autocorrelation processing, when *t*=0, the two signals essentially overlap, and the correlation value is maximum. As *t* increases, the correlation value decreases. Correlation functions are frequently used in signal processing to measure the similarity between two signals. Autocorrelation occurs when the two signals being correlated are the same signal. Autocorrelation is a measure of the similarity between a signal and its delayed version. When the delay time is zero, the autocorrelation result is the mean square value of the signal, and the autocorrelation value is maximum at this time. Furthermore, the autocorrelation result for white noise is zero, so the correlation method can effectively remove the influence of noise on the test results.

[0039] Since the sampled signals tested in this embodiment are two signals split by a power divider, ideally these two signals are essentially the same. Therefore, if there is no delay inconsistency between the two signals, the value obtained by performing correlation operations on these two channels should be the maximum. However, due to the existence of errors, the two signals will inevitably have a delay inconsistency, which can be calculated using this formula:

[0040]

[0041] x 1( t )and x 2( t These are two signals with respect to time shift. t It is a periodic function.

[0042] The consistency test results of the embodiments of this application are as follows:

[0043]

[0044] According to the test results, whether it is testing two channels of the same ADC chip or testing different ADC chips, the amplitude, offset and delay results obtained by post-test calculation and analysis are all within a very small range, which basically achieves consistent measurement of two channels or two chips.

[0045] The present application's embodiments involve the collection and analysis of a large amount of data to complete the testing of a multi-channel ADC. Taking into account both low-pass and band-pass sampling, consistency measurements of channel amplitude, channel DC offset, and layout / routing are performed. Two methods were used for delay consistency, and the test results all fell within the ranges listed in the table above. This indicates that the two methods are essentially the same; however, the correlation analysis method, in order to improve analysis accuracy, requires more time-shifting and correlation calculations, resulting in a slower speed.

[0046] The above description is only a preferred embodiment of this application and is not intended to limit this application. Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application.

Claims

1. A high-precision measurement method for multi-channel high-speed ADC acquisition, characterized in that, Including the following steps: The S100 was tested using a sampling signal ranging from 1kHz to 800MHz, based on a measurement environment using baseband sampling and bandpass sampling, and a sampling frequency of 1GHz. S200. Prepare the N-channel ADC sampling circuit to be tested, which is located on N / 2 ADC chips, where N is a positive even number; S300: The input sampling signal is split into two signals by a power divider. When measuring two channels of the same ADC chip in an N-channel ADC sampling circuit, the two signals are respectively connected to the two channel input terminals of the ADC chip. When testing two channels of different ADC chips, the two signals are respectively connected to one channel input terminal of each of the two ADC chips. S400, Start test sampling, use Vivado software to capture a segment of data, and perform channel amplitude calculation, channel DC offset calculation, and channel delay calculation based on the captured data; The channel amplitude calculation uses the FFT analysis method to convert the signal to the frequency domain and extract the amplitude information from the frequency domain. Let the amplitude values ​​obtained from the frequency domain of the two signals be A1 and A2, respectively. The amplitude consistency is calculated by the following formula: ΔA=20log(A2 / A1); In the DC offset calculation of the channel, the data of the two signals are normalized and then the average value is calculated to obtain V1 and V2. The DC offset consistency is calculated by the following formula: ΔDC=V2-V1; Channel delay is calculated using either the FFT method or the correlation method, where: In the FFT method, the correspondence between time-domain delay and frequency-domain phase is first used, that is: x(t-τ)←→X(j2πf)exp(-2πfτ), to convert the sampled signal to frequency domain analysis, where x() is the time-domain signal and X() is the frequency-domain signal. Then, based on the spectrum of the sampled signal, the phase of the point where the maximum value can be read in the spectrum is found. If the positive spectrum phase φ+ is read, the delay can be obtained as: τ=φ / -wf, where w=2π. In the relevant method, it is calculated according to the following formula: x 1( t )and x 2( t These are two signals with respect to time shift. t It is a periodic function.

2. The high-precision measurement method for multi-channel high-speed ADC acquisition according to claim 1, characterized in that, In amplitude calculation, the analysis length for both signals is the same integer period.

3. The high-precision measurement method for multi-channel high-speed ADC acquisition according to claim 1, characterized in that, Channel amplitude calculation, channel DC offset calculation, and channel delay calculation were performed on the captured data using MATLAB.

4. The high-precision measurement method for multi-channel high-speed ADC acquisition according to claim 1, characterized in that, The ADC chip is AD9680, with 1Gsps of channels.

5. The high-precision measurement method for multi-channel high-speed ADC acquisition according to any one of claims 1 to 4, characterized in that, It is suitable for measuring sine waves, triangle waves, or square waves.

Images