A distributed interferometry correlation processing method for broadband signal frequency domain extraction

Abstract

The present invention belongs to the field of radio interferometry, and discloses a broadband signal distributed interferometry correlation processing method based on frequency domain extraction. This method transforms the traditional centralized correlation processing mode of "correlation processing center" into a distributed correlation processing mode of "observation station-correlation processing center". First, the data is compensated for the delay model and the spectrum is calculated at the observation station. Then, the spectrum data of specific frequency points in the effective frequency band is extracted according to a certain extraction rate and transmitted to the correlation processing center. Finally, the spectrum data of the observation station extracted in the frequency domain is cross-correlated and the interferometric delay observation is obtained. This method loses a certain amount of measurement random accuracy. On the basis of effectively realizing the differential calibration of interferometric measurement, it can significantly reduce the data transmission requirements of broadband signals such as calibration radio sources, and improve the real-time performance of radio interferometry.

Description

Technical Field

[0001] The invention belongs to the field of radio interference measurement and relates to a broadband signal distributed interference measurement related processing method. Background Art

[0002] Radio interferometry generally adopts the alternating observation mode of radio source-detector-radio source. It is necessary to process the radio source and detector signals separately. Through two consecutive extragalactic radio source observations, the error calibration of link delay, equipment delay and other errors is realized to achieve high-precision angular position measurement of the detector. In the traditional interferometry mode, after the observation is completed, the observatory transmits the interferometric measurement raw data to the correlation processing center, and then performs correlation processing. Since the radio source signal is a noise-like signal, a higher sampling rate (such as 2MHz bandwidth and 2bit quantization for a single channel) is generally required in the observation to achieve high-precision delay estimation. Even in the narrowband correlation processing mode (such as local correlation processing), the spacecraft signal single channel is set to 50kHz / 100kHz bandwidth and 4bit / 8bit quantization, and the radio source signal is generally still set to 2MHz bandwidth, that is, compared with the detector signal, the radio source signal has a higher data transmission demand per unit time. In order to ensure the real-time performance of interferometry, a dedicated broadband transmission link is generally required from the observatory to the correlation processing center. Otherwise, correlation processing can only be performed after the completion of post-media transmission or low-rate link transmission, which seriously reduces the real-time performance of interferometry processing. Therefore, it is of great significance to study the correlation processing methods of broadband signals such as radio sources under certain transmission bandwidth constraints and reduce the observation data transmission requirements in the implementation of deep space interferometry in the future. Summary of the invention

[0003] In order to reduce the data transmission requirements for radio source observation, the present invention proposes an improved correlation processing method based on frequency domain extraction.

[0004] The technical solution of the present invention is as follows:

[0005] A method for distributed interferometric measurement of broadband signals based on frequency domain extraction is firstly used to perform delay model compensation and spectrum calculation on the data at the observation station, then extract the spectrum data of specific frequency points in the effective frequency band according to a certain extraction rate and transmit it to the relevant processing center, and finally perform cross-correlation processing and obtain the interferometric measurement delay observation quantity based on the frequency domain extracted spectrum data of the observation station.

[0006] Further, the method comprises the following steps:

[0007] (1) Calculate the radio source delay model at the observatory before observation;

[0008] (2) The observatory collects radio source data in broadband mode;

[0009] (3) The observatory performs delay compensation on the radio source observation data according to the radio source delay model and calculates the radio source data spectrum;

[0010] (4) Uniformly extract the radio source data spectrum within the effective bandwidth at a certain extraction rate, and transmit the extracted data to the relevant processing center;

[0011] (5) The correlation processing center uses the spectrum data extracted in the frequency domain to perform cross-correlation processing and bandwidth synthesis to obtain the radio source interferometry measurement delay observation quantity, thereby differentially calibrating the detector observation results.

[0012] Advantages of the present invention:

[0013] The present invention transforms the traditional centralized correlation processing mode of "correlation processing center" into a distributed correlation processing mode of "observation station-correlation processing center", significantly reduces the original data transmission requirements of interference measurement through frequency domain extraction, and effectively improves the real-time performance of radio interference measurement.

[0014] This method loses a certain amount of random measurement accuracy. On the basis of effectively realizing differential calibration of interferometric measurements, it can significantly reduce the data transmission requirements of broadband signals such as calibration radio sources and improve the real-time performance of radio interferometric measurements. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 Comparison of two correlation processing processes; (a) traditional correlation processing process, (b) correlation processing process based on frequency domain extraction;

[0016] Figure 2 Traditional correlation processing interference fringes;

[0017] Figure 3 Comparison of interference fringes in two related processing modes;

[0018] Figure 4 Interference fringes of two related processing modes under different delays; (a) delay = 0.5us, (b) delay = 1us;

[0019] Figure 5 Comparison of residual delay estimation accuracy of two correlation processing modes under different signal-to-noise ratios; (a) D = 10, (b) D = 100;

[0020] Figure 6 Comparison of related processed fringes under different decimation rates; (a) D = 10, NFFT = 4096, (b) D = 100, NFFT = 1024;

[0021] Figure 7 Comparison of delay estimation under different decimation rates for a single channel (D=10, NFFT=4096);

[0022] Figure 8 Cross-spectral phase comparison of single channel at different decimation rates (D=10, NFFT=4096);

[0023] Fig. 9 Cross-spectral phase comparison of single channel at different decimation rates (D=100, NFFT=4096);

[0024] Fig.10 Comparison of bandwidth comprehensive processing results (D=10, NFFT=2048); (a) cross-spectrum phase comparison, (b) delay estimation comparison;

[0025] Fig.11 Comparison of bandwidth comprehensive processing results; (a) D = 50, (b) D = 100. DETAILED DESCRIPTION

[0026] The present invention is further described in detail below with reference to the accompanying drawings.

[0027] The method of the present invention is based on the law that the cross-spectrum phase of the signal within the acquisition bandwidth is approximately linear with the frequency. After the observation data is compensated for the delay model at the measuring station, the spectrum data of a specific frequency point is extracted at a certain extraction rate and transmitted to the relevant processing center. Then, the cross-correlation processing is performed based on the spectrum data extracted in the frequency domain to obtain the interferometric delay observation. The following is a further explanation from a theoretical perspective.

[0028] Assume that station A receives a radio source signal s A (t) is q(t), and station B receives the same radio source signal s B (t) is q(t-τ), then the two station signals are correlated to obtain the cross-spectral signal:

[0029]

[0030] It can be seen from the above formula that the cross-spectrum phase is linearly related to the signal frequency, so the residual delay τ can be estimated by fitting the cross-spectrum phase. Furthermore, from the relationship between the cross-spectrum phase and the frequency, it can be seen that any frequency in the effective frequency band of data acquisition satisfies the above relationship, so the residual delay estimation can also be obtained by using only the cross-spectrum phase at some frequency points in the effective frequency band.

[0031] Two related processing processes are compared. Figure 1 As shown. Among them, N is the FFT length, M is the number of segments of the original data in the integration period, that is, it is assumed that the data correlation processing in one integration period is divided into M segments, each segment is N long; A and B are the frequency spectra of the original measurement data of the two measurement stations, and C is the cross-correlation spectrum of the data of the two measurement stations. is the average cross-correlation spectrum; L is the number of extracted frequency points in the new correlation processing method. Figure 1As shown in a, in the traditional correlation processing, the data of the two stations are segmented, FFT and correlated to obtain the cross spectrum, and the cross spectrum of the integration period is obtained by averaging the cross spectra of different data segments; Figure 1 As shown in FIG. 2 , in the new correlation processing, after the data of the two stations are segmented and FFT is performed, their respective spectra are first extracted according to fixed frequency points, and then the extracted data spectra are correlated to obtain the cross-spectrum and average cross-spectrum.

[0032] Without loss of generality, assuming that the original data length is N (i.e., only data of one FFT length), and the number of FFT points in the correlation processing is N, then the amount of data transmitted by a single station in the traditional correlation processing mode is N; assuming that some fixed frequency information is taken, the number of frequency points is L, then the amount of data transmitted by a single station is reduced to L / N in the traditional mode, and the decimation rate is approximately N / L.

[0033] In the processing of measured data, the radio source correlation processing has high requirements on the accuracy of delay model compensation. Taking into account that the phase of a fixed frequency point of the radio source data of a single station has irregular patterns over time, it is impossible to perform delay model compensation on the frequency and phase information of the single station data. Instead, it is only possible to use the traditional correlation processing mode to achieve accurate compensation of the delay model at each measuring station, obtain the spectrum information after delay model compensation, and then transmit it to the correlation processing center for correlation processing, so as to obtain residual interference fringes and residual delay estimation.

[0034] The present invention proposes a wideband signal correlation processing method based on frequency domain extraction, and adopts a distributed correlation processing mode of "observation station-correlation processing center" in the processing mode.

[0035] Specific implementation plan

[0036] (1) Before observation, the radio source delay model is calculated at the observatory using known techniques;

[0037] (2) The observatory collects radio source data in the broadband mode of traditional correlation processing (e.g., sampling rate 4 MHz, quantization bit number 2 bits);

[0038] (3) The observatory performs delay compensation on the radio source observation data according to the radio source delay model and calculates the radio source data spectrum;

[0039] (4) Uniformly extract the spectrum of radio source data within the effective bandwidth according to a certain extraction rate, and transmit the extracted spectrum data to the relevant processing center;

[0040] (5) The correlation processing center uses the spectrum data extracted in the frequency domain to perform cross-correlation processing and bandwidth synthesis to obtain the radio source interferometry measurement delay observation quantity, thereby differentially calibrating the detector observation results.

[0041] Example 1—Monte Carlo Simulation of an Algorithm

[0042] Assume that the sampling rate fs = 4MHz, the relative delay between station A and station B is 0.25us, the data duration is 1s, and NFFT = 1024; assume that only 1 / 100 of the spectrum data is transmitted in the frequency domain extraction correlation processing mode (that is, the single-station data transmission volume is reduced by 100 times). In the traditional correlation processing mode, the interference fringes are as follows Figure 2 It can be seen that the cross-spectrum phase has a linear relationship with the frequency within the effective bandwidth, and the residual delay can be estimated to be 0.249997us after fitting.

[0043] In the frequency domain extraction correlation processing mode, the data spectrum is first obtained at the measuring station, and the spectrum of the selected frequency is transmitted to the correlation processing center, and then the correlation processing is performed to obtain the cross-spectrum phase as shown in Figure 3 As shown in the figure, "Novel" indicates the frequency domain extraction correlation processing method, and "Traditional" indicates the traditional correlation processing method, the same below. It can be seen that the cross-spectral phases in the two modes almost overlap at the same frequency point. After fitting, the residual delay in the improved correlation processing mode is estimated to be 0.249781us, which is basically consistent with the traditional correlation processing mode.

[0044] Comparison of interference fringes in two related processing modes under different time delays Figure 4 It can be seen that when the relative delay is 0.5us and 1us, the results of the improved correlation processing mode are basically consistent with those of the traditional correlation processing mode, that is, the correlation processing results of the improved mode are not affected by the residual delay ambiguity.

[0045] Comparison of residual delay estimation accuracy of two correlation processing modes under different signal-to-noise ratios Figure 5 As shown in the figure, D represents the decimation factor, and D=10 means extracting 1 / 10 of the frequency information of the entire bandwidth for delay estimation. It can be seen that when SNR=-14dB, the deviation between the delay estimation results obtained by the two related processing methods and the true value is in the order of ns, where the smaller the decimation factor, the smaller the deviation; when SNR is greater than -10dB, the delay estimation deviation is in the sub-ns order of magnitude. When the decimation factor increases from 10 to 100, the delay estimation mean does not deteriorate significantly; however, the delay estimation error increases significantly. Figure 5 The results show the correctness of the frequency domain extraction correlation processing method.

[0046] Example 2—Measured Data Processing Verification (CE04ZJX 20180918)

[0047] The results were verified using radio source observation data from September 18, 2018.

[0048] Single channel correlation processing verification

[0049] The comparison of interference fringes of a single-channel radio source obtained by two correlation processing methods is shown in Figure 2. Figure 6 As shown. It can be seen that the cross-spectrum phase obtained by the frequency domain extraction correlation processing mode coincides with the traditional correlation processing mode at the frequency extraction point, and the extraction rate determines the amount of data for fringe fitting in the new correlation processing mode. Due to the influence of noise, the distribution of the cross-spectrum phase in the interference fringes shows a certain randomness. Since the frequency domain extraction correlation processing mode only uses the cross-spectrum phase at some frequency points, the single-channel delay estimation result is quite different from the traditional method, such as Figure 7 As shown in Figure 1, the random jitter of the delay estimation of the frequency domain extraction correlation processing method is larger. The average deviation of the delay estimation of the two correlation processing methods is about 10.41ns, and the random jitter is about 30.52ns.

[0050] Considering that bandwidth integration is required between interferometric measurement channels to obtain the final delay estimation, the single-channel cross-spectrum phase mean affects the bandwidth integration and the estimation of residual delay. Figure 8 The single-channel cross-spectrum phase mean values ​​obtained by two correlation processing modes when D = 10 and NFFT = 4096 are given. It can be seen that the trend of the frequency domain extraction correlation processing result is almost the same as that of the traditional correlation processing result, and the difference value between the two fluctuates around 0. According to statistics, the mean cross-spectrum phase difference under the two correlation processing modes is 0.01rad, and the random error is about 0.08rad.

[0051] Under the same conditions, when the extraction rate increases (D = 100), the mean phase of the interference fringe cross spectrum is as follows: Fig. 9 As shown. Figure 8 Compared with the two correlation processing results, the difference random jitter increases. The mean value of the cross-spectrum phase difference is 0.03 rad, and the random error is about 0.24 rad.

[0052] Bandwidth comprehensive processing verification

[0053] After the single-channel interferometric measurement is processed, the interferometric measurement delay observation can be extracted by comprehensive processing of all channel bandwidths. The comprehensive results of multi-channel bandwidths obtained by the two related processing methods are shown in Figure 2. Fig.10 As shown in Figure (a), the cross-spectral phases of the two results are basically consistent. Figure (b) shows that the delay estimation result of the frequency domain extraction correlation processing jitters around the result of the traditional method, the mean value of the deviation is about 0.2884ns, and the random accuracy of the deviation is about 1.6032ns. Therefore, it can be considered that the results of the new correlation processing method after bandwidth comprehensive processing are basically consistent with the results of the traditional method, which proves the effectiveness of the proposed method. Fig.11 In the figure, when the decimation rate increases to 50 and 100 respectively, the mean values ​​of the delay estimation deviations obtained by the two methods are 0.4682ns and 0.7994ns respectively, and the corresponding random precisions of the delay estimation are 4.3805ns and 4.1909ns respectively.

[0054] The results of comprehensive simulation and measured data processing. Considering the fact that the number of measured data samples is limited, which leads to weak statistical significance, it can be considered that (1) the broadband signal correlation processing method based on frequency domain extraction proposed in the present invention is correct; (2) the method proposed in the present invention, at the cost of a certain loss of measurement random accuracy, can significantly reduce the data transmission requirements of broadband signals such as calibration radio sources on the basis of effectively realizing differential calibration of interferometric measurement, and improve the real-time performance of radio interferometric measurement.

Claims

1. A distributed interferometry correlation processing method for broadband signal frequency domain extraction, characterized in that: The method comprises the following steps: (1) Calculate the radio source delay model at the observatory before observation; (2) The observatory collects radio source data in broadband mode; (3) The observatory performs delay compensation on the radio source observation data according to the radio source delay model and calculates the radio source data spectrum; (4) Uniformly extract the radio source data spectrum within the effective bandwidth at a certain extraction rate, and transmit the extracted data to the relevant processing center; (5) The correlation processing center uses the spectrum data extracted in the frequency domain to perform cross-correlation processing and bandwidth synthesis to obtain the radio source interferometry measurement delay observation quantity, thereby differentially calibrating the detector observation results.

Images