A real-time digital spectrum analyzer for electromagnetic environment monitoring

Abstract

The invention discloses a real-time digital spectrum analyzer for electromagnetic environment monitoring, which is characterized in that the radio frequency signal after the electromagnetic wave of the target area is frequency-converted to the baseband is processed by an ADC sampling module, an anti-aliasing filtering module, a windowing module and an FFT module to obtain multiple frequency spectrums and send them to a transposition module and a spectrum superposition module respectively; the spectrum superposition module accumulates the received spectrum to obtain the mean value of the spectrum within a set time length; the transposition module sends the transposed spectrum to a sorting module; the sorting module sends the sorted spectrum to a time occupancy calculation module; the time occupancy calculation module calculates the time occupancy of the spectrum within the set time length and the time occupancy of the spectrum within the total observation time; the data transmission module transmits the spectrum statistical data obtained from the spectrum superposition module, the sorting module and the time occupancy calculation module to the shared memory of the host end, so that the spectrum display module can display the electromagnetic wave spectrum of the target area in real time.

Description

Technical Field

[0001] The invention belongs to the field of digital signal processing and acquisition, and in particular relates to a real-time digital spectrum analyzer for electromagnetic environment monitoring. Background Art

[0002] The radio signals emitted by distant celestial bodies in the universe are very weak. FAST, which has extremely high sensitivity, has included the weak radiation source of Milliansky into its observation targets. This requires not only extremely sensitive radio telescopes, but also extremely strict requirements on the electromagnetic environment of the site, so it is necessary to carefully measure the electromagnetic environment of the site. Due to the widespread use of wireless communication equipment, electromagnetic interference is now more sudden. Traditional superheterodyne spectrometers do not capture data during the retrace process, so they may miss many signals when capturing burst signals and narrow pulses (such as radar signals).

[0003] To overcome the above shortcomings, a real-time digital spectrum analyzer is needed. The main processing steps of a digital spectrum analyzer include: ADC sampling, anti-aliasing filtering, windowing, FFT, and spectrum display. In order to seamlessly capture short-term burst signals, a real-time digital spectrum analyzer needs to overlap the data before performing FFT to offset the signal depression caused by the edge of the window function. Some real-time digital spectrum analyzers with higher real-time performance can achieve an overlap rate of more than 90%.

[0004] With the help of digital persistence technology, real-time digital spectrum analyzers compress fast-changing spectrums to make them easier for human eyes to observe; using frequency mask trigger (FMT), real-time digital spectrum analyzers can capture specific events of interest. However, due to the short length of transient signals and their limited energy, when the FFT length is long, the transient signal will be submerged in the noise; and when the FFT length is low, the spectrum resolution will be reduced. The general method to improve the signal-to-noise ratio is to accumulate the spectrum, but for transient signals with low time occupancy, it is not suitable for integration.

[0005] In order to meet the needs of spectrum monitoring in the professional field, the current design of commercial real-time digital spectrum analyzers pays more attention to the rapid identification of transient signals from a visual perspective. Such signals usually have a very short time occupancy, which may only last from a few microseconds to a few milliseconds. In order to ensure that these short-lived events can be captured, most commercial RTSAs use a low FFT point number, usually no more than 4K points, which limits the frequency resolution that the instrument can provide.

[0006] In fields such as radio astronomy and electromagnetic environment monitoring, researchers often need a higher level of spectral resolution to accurately distinguish closely adjacent signals and minimize the impact of potential interference on adjacent channels. Higher resolution means that the spectral structure can be analyzed more finely, which is essential for identifying signals in complex backgrounds. In addition, spectrum statistics are equally important for long-term monitoring applications. When performing RF interference mitigation, signals with low time occupancy are suitable for elimination in the time domain, and signals with high time occupancy are suitable for elimination in the spectrum. Time occupancy analysis provides a basis for this.

[0007] Existing digital real-time spectrum analyzers cannot achieve both high spectrum resolution and high signal-to-noise ratio when detecting transient signals. Summary of the invention

[0008] In view of the problems existing in the prior art, the purpose of the present invention is to provide a real-time digital spectrum analyzer for electromagnetic environment monitoring, which can monitor transient signals at high spectrum resolution and obtain time occupancy with high spectrum resolution, thereby providing richer statistical information for spectrum monitoring. This application implements the design of a real-time digital spectrum analyzer based on FPGA and GPU platforms in response to the development needs of the latest international electromagnetic environment monitoring technology, and its features are:

[0009] 1) In order to solve the problem of too low signal-to-noise ratio under high FFT points of traditional real-time spectrometer (integration of transient signals is meaningless), the present invention uses quantile information to smooth the spectrum. Figure 1 (a) to Figure 1 (h) are analyses of a signal received by the CHIME telescope (set duration t = 33ms, bandwidth = 500MHz, frequency converted to baseband), with an FFT point of 16K. It can be seen that quantiles have a good smoothing effect on noise, and the signal-to-noise ratio of the 99% quantile has been significantly improved.

[0010] 2) Provides calculation of time occupancy η. Supports dynamic time occupancy and the total observation time The analysis provides valuable reference information for spectrum management and RF interference mitigation. Note that the noise power at different quantiles is different, so when calculating η, different baselines need to be set for different quantiles Q. and different thresholds .

[0011] 3) Supports 450MHz real-time analysis bandwidth and uses 14bit data dynamic range to meet more observation needs.

[0012] 4) Supports setting the duration of each signal collection, and the setting range is 20ms~300ms.

[0013] The technical solution of this application is:

[0014] A real-time digital spectrum analyzer for electromagnetic environment monitoring, characterized in that it includes

[0015] The ADC sampling module 1 is used to sample the radio frequency signal after the electromagnetic wave of the target area is converted to the baseband, and transmit the sampled signal to the anti-aliasing filtering module 2;

[0016] The anti-aliasing filter module 2 is used to perform anti-aliasing filtering on the received sampling signal and then send it to the windowing module 3;

[0017] The windowing module 3 is used to perform one-dimensional convolution on the data output by the anti-aliasing filtering module 2, obtain data with a set overlap rate and send it to the FFT module 4;

[0018] FFT module 4, used for calculating the data output by windowing module 3, obtaining multiple frequency spectra and sending them to transposition module 6 and spectrum superposition module 5 respectively;

[0019] The spectrum superposition module 5 is used to accumulate the received spectrum, obtain the average value of the spectrum within a set time length and send it to the data transmission module 9;

[0020] The transposition module 6 is used to transpose the received spectrum so that the data to be sorted is continuous on the video memory, and send the transposed spectrum to the sorting module 7;

[0021] The sorting module 7 is used to perform segmented cardinality sorting on the transposed spectrum, send the sorted spectrum to the time occupancy calculation module 8, and send the peak value, the selected quantile Q, and the median of each frequency to the data transmission module 9;

[0022] The time occupancy calculation module 8 is used to process the sorted spectrum to obtain the time occupancy of the spectrum within the set time length. ;Will The temporal occupancy of the spectrum during the total observation time calculated last time The updated weighted average , then and Send to data transmission module 9 and retain the updated ; Time occupancy of the spectrum during the total observation time The initial value is the time occupancy of multiple historical spectra of set duration The average value of

[0023] Data transmission module 9: transmits spectrum statistical data obtained from spectrum superposition module 5, sorting module 7 and time occupancy calculation module 8 to the shared memory of the host side for use by data storage module 10 and spectrum display module 11;

[0024] A data storage module 10, used to store the spectrum statistics in the shared memory in a fits format;

[0025] The spectrum display module 11 is used to read the data in the shared memory and display the electromagnetic wave spectrum of the target area in real time;

[0026] Baseline and threshold calculation module 12, used to calculate different baselines for different quantiles Q and threshold And send it to the time occupancy calculation module 8.

[0027] Furthermore, the baseline and threshold calculation module 12 calculates different baselines for different quantiles Q and threshold The method is as follows: after setting the number of FFT points, collect a signal of a certain length and divide it into N segments of data, each segment of which is the set length; calculate the quantile Q of the spectrum amplitude for each segment of data, and for each quantile Q, calculate its mean value μ in the N segments of data Q and standard deviation σ Q ; Then from the mean μ Q Extract the mean baseline corresponding to the current quantile Q , for the standard deviation σ Q Extract the baseline to get the standard deviation baseline corresponding to the current quantile Q .

[0028] Furthermore, the selected quantile Q is the 90% quantile, and the threshold = .

[0029] Furthermore, the time occupancy is calculated by obtaining N frequency spectra within a period of time, and for each frequency point in the spectrum, if there are N r If the frequency spectrum exceeds the set threshold at this frequency point, the time occupancy of this frequency point is .

[0030] Furthermore, the FFT module 4 stores the multiple spectrums obtained by storing each spectrum as a row, and the total number of rows is the total number of spectrums of the set time length; the same frequency point is stored as a column, and the frequency points in the same column have the same frequency but different time.

[0031] Furthermore, the length of each row is the number of FFT points / 2.

[0032] Furthermore, the set duration is 20ms~300ms.

[0033] Furthermore, the FFT module 4 calculates the data output by the windowing module 3 using the CUFFT library function to obtain multiple frequency spectra.

[0034] Furthermore, the data storage module 10 uses the cfitsio library function to store the spectrum statistical data in the shared memory in a fits format, and the data format is float32.

[0035] Furthermore, the spectrum display module 11 uses an interface drawn by PyQt to read the data in the shared memory and displays the spectrum in real time.

[0036] The advantages of the present invention are as follows:

[0037] In view of the problem that transient signals cannot have both high RBW and high signal-to-noise ratio, the quantile method proposed in this invention can greatly improve the signal-to-noise ratio of transient signals under RBW. Furthermore, the time occupancy analysis based on the quantile method also has high RBW, which provides valuable reference information for radio telescope site selection, radio frequency interference (RFI) mitigation, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1(a) shows the mean of the spectrum within a set time length.

[0039] Figure 1(b) shows the peak value of the spectrum within the set time length.

[0040] Figure 1(c) shows the 99% quantile of the spectrum within the set time length.

[0041] Figure 1(d) shows the 90% quantile of the spectrum within the set time length.

[0042] Figure 1(e) shows the median of the spectrum within a set time length.

[0043] Figure 1(f) shows the 25% quantile of the spectrum within a set time length.

[0044] Figure 1(g) shows the temporal occupancy of the spectrum within a set time period.

[0045] Figure 1 (h) shows the spectrum of an arbitrary segment before sorting.

[0046] Figure 2 This is the overall structure of the real-time digital spectrum analyzer.

[0047] Figure 3 This is the storage method of the spectrum obtained by FFT module 4 in the video memory.

[0048] Figure numerals: 1-ADC sampling module, 2-anti-aliasing filtering module, 3-windowing module, 4-FFT module, 5-spectrum superposition module, 6-transposition module, 7-sorting module, 8-time occupancy calculation module, 9-data transmission module, 10-data storage module, 11-spectrum display module, 12-baseline and threshold calculation module. DETAILED DESCRIPTION

[0049] The present invention is further described in detail below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention but not to limit the scope of the present invention.

[0050] The architecture of a real-time digital spectrum analyzer for radio electromagnetic environment monitoring is as follows: Figure 2 As shown, the ADC sampling module 1, the data transmission module 9, the data storage module 10, and the spectrum display module 11 are located at the host end of the real-time digital spectrum analyzer, and the anti-aliasing filtering module 2, the windowing module 3, the FFT module 4, the spectrum superposition module 5, the transposition module 6, the sorting module 7, the time occupancy calculation module 8, and the baseline and threshold calculation module 12 are located in the GPU of the real-time digital spectrum analyzer. The specific functions of each module are:

[0051] ADC sampling module 1: Use the FPGA-based PCIe-ADC data acquisition card to sample the RF signal at 1Gsps@14bit after converting the electromagnetic wave of the target area to baseband. After passing through the ping-pong buffer, the data is transferred to the locked page memory of the PC through the PCIe interface, and then sent to the anti-aliasing filter module 2 through the PCIe interface;

[0052] Anti-aliasing filter module 2: Anti-aliasing uses a 16th-order or 32nd-order IIR low-pass filter. The filter coefficients are located in the constant memory, and the data is loaded into the shared memory of the GPU to achieve efficient IIR filtering. The data with high-frequency aliasing signals removed is sent to the windowing module 3;

[0053] Windowing module 3: Use the window function as the kernel function to perform one-dimensional convolution on the data obtained by the anti-aliasing filter module 2. By controlling the convolution step size, the overlap rate of the data is controlled, and then sent to the FFT module 4. Although windowing is a memory-intensive operation, it can hide the memory access latency by overlapping it with other computationally intensive operations in the CUDA stream;

[0054] FFT module 4: uses the efficient CUFFT library function to perform FFT (fast Fourier transform) on the signal output by the windowing module 3 to obtain multiple spectrums, which are sent to the transposition module 6 and the spectrum superposition module 5 respectively;

[0055] Spectrum superposition module 5: accumulates the spectrum obtained by FFT module 4, obtains the average value of the spectrum within a set time length, and sends it to data transmission module 9;

[0056] Transposition module 6: adjusts the storage position of the spectrum in the video memory so that the data to be sorted is stored continuously in the video memory. The transposed spectrum is sent to the sorting module 7. The spectrum obtained from the FFT module 4 is stored in the video memory in a similar way to the waterfall diagram, such as Figure 3As shown in the figure, the length of each row is the number of FFT points / 2, storing a spectrum; the total number of rows is the total number of spectrums within the set time length. Each column stores the same frequency point, with the same frequency but different time; each row of data is stored continuously in the video memory, while each column of data is not stored continuously in the video memory; when calculating the quantile of the spectrum, each frequency point needs to be sorted in the time direction, that is, each column of data needs to be sorted; however, when the data is continuous in the video memory, the sorting speed is faster, so this module transposes the spectrum.

[0057] Sorting module 7: Segmented cardinality sorting of the spectrum obtained by transposition module 6, that is, sorting the amplitude of each frequency point at different times. The sorted spectrum is sent to the time occupancy calculation module 8, and the peak value, 90% quantile, and median of each frequency are sent to the data transmission module. The frequency of each row of the sorted spectrum is the same, and each column stores the same quantile. Assuming there are M columns in total, the Mth column is the peak value, the M·0.9th column is the 90% quantile, and the M / 2th column is the median.

[0058] Time occupancy calculation module 8: processes the sorted spectrum obtained by sorting module 7 to obtain the time occupancy of the spectrum within a set time length and the total observed time spectrum . and Sent to data transmission module 9, and reserved The updated result is used by the time occupancy calculation module 8 in the next operation. The time occupancy calculation method is as follows: Assuming that N frequency spectra are obtained in a continuous period of time, for each frequency point, if there are N r If the frequency spectrum exceeds the threshold at this frequency point, the time occupancy of this frequency point . Take the signal of set time length t to calculate η, and get ; Multiple The average value is Actual calculation When With the previous Weighted average, get the new Different frequencies have different and .Threshold = ,in is the baseline for the mean of different quantiles, are baselines of standard deviations of different quantiles, which are derived from the baseline and threshold calculation module 12 , and the specific calculation is described in the baseline and threshold calculation module 12 .

[0059] Data transmission module 9: transmits spectrum statistical data obtained from spectrum superposition module 5, sorting module 7 and time occupancy calculation module 8 to the shared memory of the host side for use by data storage module 10 and spectrum display module 11;

[0060] Data storage module 10: Use the cfitsio library function to store the statistical data of the spectrum in the shared memory in the fits format, and the data format is float32.

[0061] Spectrum display module 11: Use the interface drawn by PyQt to read the data in the shared memory and display the spectrum in real time.

[0062] Baseline and threshold calculation module 12: Before calculating the time occupancy η, it is necessary to calculate different baselines for different quantiles Q and threshold Specifically, after setting the number of FFT points, collect a continuous signal (5-10s long) and divide it into N segments, each of which is the set time length. Calculate the quantile Q of the spectrum amplitude for each segment of data, and for each Q, calculate its mean μ in the N segments of data Q and standard deviation σ Q Then, using the arPLS method, μ Q Extract the baseline , for σ Q Extract the baseline ,Will and Sent to the time occupancy calculation module 8. Each time the FFT points are adjusted, this module needs to be used again to calculate and .

[0063] The backend data processing is controlled by three threads. Thread 1 controls module 1, thread 2 controls anti-aliasing filter module 2, windowing module 3, FFT module 4, spectrum overlay module 5, transposition module 6, sorting module 7, time occupancy calculation module 8 and spectrum display module 11, and thread 3 controls data storage module 10 to form a data stream for efficient data processing. The data processing in the GPU uses two CUDA streams to overlap data transmission and memory-intensive calculations. After each set length of RF signal is processed, the spectrum is displayed and parameter control is performed using an interface drawn by PyQt.

[0064] Table 1 below gives a summary of the technical parameters of the radio electromagnetic environment real-time spectrum analyzer.

[0065] Table 1 is the parameter table of the real-time spectrum analyzer for radio electromagnetic environment

[0066] Number of spectrum channels 4K~128K optional Overlap rate 50% or 75% Real-time bandwidth analysis 450MHz Window Function Gaussian window, Hann window, Hamming window, flat top window, Kaiser window, Blackman window Anti-aliasing filtering 16 or 32 order IIR filter Input range ±2V Dynamic Range 84dB(14bit)

[0067] Although the specific embodiments of the present invention are disclosed for the purpose of illustration, the purpose is to help understand the content of the present invention and implement it accordingly, those skilled in the art will understand that various substitutions, changes and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the best embodiment, and the scope of the present invention is subject to the scope defined in the claims.

Claims

1. A real-time digital spectrum analyzer for electromagnetic environment monitoring, characterized in that: include An ADC sampling module (1) is used to sample the radio frequency signal after the electromagnetic wave of the target area is converted to the baseband, and transmit the sampled signal to the anti-aliasing filtering module (2); An anti-aliasing filter module (2) is used to perform anti-aliasing filtering on the received sampling signal and then send the signal to the windowing module (3); A windowing module (3) is used to perform one-dimensional convolution on the data output by the anti-aliasing filter module (2), obtain data with a set overlap rate and send it to the FFT module (4); An FFT module (4) is used to calculate the data output by the windowing module (3) to obtain multiple frequency spectra and send them to the transposition module (6) and the spectrum superposition module (5) respectively; A spectrum superposition module (5) is used to accumulate the received spectrum, obtain the average value of the spectrum within a set time period and send it to the data transmission module (9); A transposition module (6) is used to transpose the received spectrum so that the data to be sorted is continuous on the video memory, and send the transposed spectrum to the sorting module (7); A sorting module (7) is used to perform segmented radix sorting on the transposed spectrum, send the sorted spectrum to the time occupancy calculation module (8), and send the peak value, the selected quantile Q, and the median of each frequency to the data transmission module (9); The time occupancy calculation module (8) is used to process the sorted spectrum to obtain the time occupancy of the spectrum within a set time length. ;Will The temporal occupancy of the spectrum during the total observation time calculated last time The updated weighted average , then and Send to the data transmission module (9) and retain the updated ; Time occupancy of the spectrum during the total observation time The initial value is the time occupancy of multiple historical spectra of set duration The average value of the time occupancy; wherein the calculation method of the time occupancy is: obtain N spectrums within a period of time, for each frequency point in the spectrum, if there are N in the N spectrum r If the frequency spectrum exceeds the set threshold at this frequency point, the time occupancy of this frequency point is ; Data transmission module (9): transmits spectrum statistical data obtained from spectrum superposition module (5), sorting module (7) and time occupancy calculation module (8) to the shared memory of the host side for use by data storage module (10) and spectrum display module (11); A data storage module (10), used for storing the spectrum statistical data in the shared memory in a fits format; A spectrum display module (11), used for reading the data in the shared memory and displaying the electromagnetic wave spectrum of the target area in real time; Baseline and threshold calculation module (12) for calculating different baselines for different quantiles Q and threshold And send it to the time occupancy calculation module (8).

2. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The baseline and threshold calculation module (12) calculates different baselines for different quantiles Q and threshold The method is as follows: after setting the number of FFT points, collect a signal of a certain length and divide it into N segments of data, each segment of which is the set length; calculate the quantile Q of the spectrum amplitude for each segment of data, and for each quantile Q, calculate its mean value μ in the N segments of data Q and standard deviation σ Q ; Then from the mean μ Q Extract the mean baseline corresponding to the current quantile Q , for the standard deviation σ Q Extract the baseline to get the standard deviation baseline corresponding to the current quantile Q .

3. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 2, characterized in that: The selected quantile Q is the 90% quantile, and the threshold = .

4. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The storage method of the multiple frequency spectra obtained by the FFT module (4) is as follows: each frequency spectrum is stored as a row, and the total number of rows is the total number of frequency spectra of the set time length; the same frequency point is stored as a column, and the frequency points in the same column have the same frequency but different time.

5. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 4, characterized in that: The length of each row is the number of FFT points / 2.

6. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The set time length is 20ms~300ms.

7. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The FFT module (4) uses the CUFFT library function to calculate the data output by the windowing module (3) to obtain multiple frequency spectra.

8. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The data storage module (10) uses the cfitsio library function to store the spectrum statistical data in the shared memory in a fits format, where the data format is float32.

9. The real-time digital spectrum analyzer for electromagnetic environment monitoring according to claim 1, characterized in that: The spectrum display module (11) uses an interface drawn by PyQt to read the data in the shared memory and displays the spectrum in real time.

Images