A pulsar incoherent real-time dispersion elimination system and method with pipeline structure
Through the combination of pipeline structure and Barlett periodic graph method, the shortcomings of the existing incoherent absent dispersion algorithm in signal processing speed, memory usage and time resolution are solved, and efficient pulsar signal processing is realized to adapt to different dispersion coefficients. Improves real-time and storage efficiency.
Patent Information
- Application Number
- CN202111192571.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-13
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-10-13
AI Technical Summary
The existing incoherent absent dispersion algorithms have shortcomings in signal processing speed, memory usage and time resolution, especially for pulsar signals with short periods and weak signals, which are difficult to achieve ideal results.
The pulsar incoherent real-time dispersion system using pipeline structure includes a preprocessing part and a dispersion processing part, which is stored through analog-to-digital conversion unit, power spectrum estimation unit, input buffer, data overlay unit, pulsar data buffer unit, and dispersion coefficient storage. Unit and output buffering are used to perform power spectrum estimation and data superposition using the Barlett periodic graph method, and dynamically configure dispersion coefficient storage to reduce inverse Fourier transform summing operations, improve processing speed and reduce storage space requirements.
It realizes efficient real-time processing performance, reduces storage space consumption, improves the processing speed of dispersion data, avoids the inverse Fourier transform summing operation, adapts to different dispersion coefficients, flexibly adjusts the dispersion coefficient storage, and improves the time resolution and frequency resolution.
Smart Images

Figure CN114021065B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a pulsar signal incoherent de-dispersion system and method, belonging to the technical field of pulsar signal observation and search. Background Art
[0002] A pulsar is a high-speed rotating neutron star with a very strong magnetic field. There is an angle between the rotation axis and the magnetic axis of the pulsar. Radiation rays similar to lighthouse beams are generated near the two magnetic poles. When the rays pass through the earth, the antenna will receive a periodic pulse signal with a period consistent with the rotation period of the pulsar, and a small rate of change in the period. During the propagation of pulsar signals, affected by the free electrons in the interstellar medium, signals of different frequencies produce different delay effects. The lower frequency signal components will arrive at the receiving end later than the higher frequency signal components, which will cause the pulse energy to disperse and deform the pulsar superposition profile, resulting in a poor signal-to-noise ratio. When the dispersion is severe, the pulse profile cannot even be seen.
[0003] The current de-dispersion methods mainly include coherent de-dispersion and incoherent de-dispersion. Coherent de-dispersion is based on the inverse transfer function model of the interstellar medium, and eliminates signal dispersion through frequency domain multiplication. Incoherent de-dispersion is a discontinuous de-dispersion based on the signal phase shift of each frequency band in the time domain. The current incoherent de-dispersion algorithms include direct de-dispersion, tree-type de-dispersion and sub-band de-dispersion. The direct dispersion elimination algorithm directly performs time shift operations on the signals at each frequency point according to the dispersion delay formula of each frequency point to obtain the data after dispersion elimination. The disadvantage is that the amount of calculation is relatively large. Tree dispersion elimination is mainly used in pulsar search. When multiple dispersion elimination coefficients (DM) need to be tried, a butterfly operation structure similar to the fast Fourier transform is used to reuse repeated operations in different DM dispersion elimination processes to reduce the amount of calculation. This method is only for the DM search process and does not actually improve the signal processing speed in the case of a single DM. Subband dispersion elimination is to divide the pulsar signal into multiple sub-bands for processing. Each sub-band is considered to have the same dispersion delay, which reduces the amount of calculation for dispersion elimination at the expense of time resolution. The usual incoherent dispersion elimination includes the following steps: (1) Multiple filtering operations can be performed on the pulsar baseband data converted into digital signals to reduce the spectrum leakage of the discrete Fourier transform; (2) Fourier transform; (3) Divide the frequency channel and shift the data; (4) Inverse Fourier transform and sum.
[0004] The existing de-dispersion algorithm stores a large amount of pulsar raw data and then processes it through the CPU or GPU. This method requires a large amount of memory space. The direct de-dispersion algorithm is difficult to achieve good real-time performance due to its large amount of calculation. Although the sub-band de-dispersion algorithm reduces memory usage, there will be a relatively large loss in time resolution. For pulsars with short periods and weak signals, de-dispersion is difficult to achieve ideal results. Summary of the invention
[0005] The purpose of the present invention is to provide a pulsar incoherent real-time de-dispersion system and method with a pipeline structure to solve the problems of signal processing speed, memory usage and resolution loss of existing incoherent de-dispersion algorithms.
[0006] To achieve the above object, the present invention adopts the following technical solution:
[0007] A pulsar incoherent real-time de-dispersion system with a pipeline structure includes a pre-processing part and a de-dispersion processing part, wherein:
[0008] The preprocessing part includes an analog-to-digital conversion unit and a power spectrum estimation unit. The analog-to-digital conversion unit is used to convert the input pulsar baseband signal into a digitally sampled pulsar signal and input it to the power spectrum estimation unit. The power spectrum estimation unit is used to obtain the power spectrum of the pulsar signal and transmit it to the de-dispersion processing part.
[0009] The de-dispersion processing part includes an input buffer, a data superposition unit, a pulsar data buffer unit, a dispersion coefficient storage unit and an output buffer. The input buffer is used to output the power spectrum signal output serially in the order of frequency channels to the data superposition unit. The data superposition unit is used to process the serially input power spectrum signal in a pipeline manner, superimpose it with the signal buffered in the pulsar data buffer unit, obtain a time domain power signal, and transmit it to the output buffer. The output buffer is used to buffer the burst-type de-dispersion time domain power signal; the dispersion coefficient storage unit is used to give the position where the current pulsar data needs to be written into the pulsar data buffer unit, and transmit it to the pulsar data buffer unit.
[0010] A pulsar incoherent real-time de-dispersion method based on the system comprises the following steps:
[0011] Step S1, the RF module outputs a pulsar baseband signal, which is input into an analog-to-digital conversion unit for sampling and digitization processing, and outputs a digitally sampled pulsar signal;
[0012] Step S2, the pulsar signal passes through a power spectrum estimation unit to obtain a power spectrum of the pulsar signal, which is transmitted to an input buffer as preprocessed data;
[0013] Step S3, the input buffer buffers the pre-processed data input in burst mode by the data pre-processing part, and outputs the power spectrum signal output serially in the order of frequency channels to the data superposition unit;
[0014] Step S4, the data superposition unit processes the serially input power spectrum signal in a pipeline manner, superimposes it with the signal buffered in the pulsar data buffer unit, obtains a time domain power signal, and transmits it to the output buffer;
[0015] Step S5, the dispersion coefficient storage unit gives the position where the current pulsar data in step S4 needs to be written into the pulsar data buffer unit;
[0016] Step S6, outputting a buffer of the burst-type time domain power signal after de-dispersion.
[0017] In step S1, the analog-to-digital conversion unit converts the pulsar digital signal with a high-speed sampling rate into multi-phase parallel data with a lower sampling rate.
[0018] In step S2, the power spectrum estimation unit is based on the Barlett periodogram method. While performing power spectrum estimation, it utilizes the data superposition characteristics of the Barlett periodogram to dynamically configure the number of superpositions to control the time resolution of the system and reduce the data processing amount of the subsequent modules.
[0019] In step S3, the input buffer completes the buffering of the input power spectrum signal to ensure the synchronization of the subsequent dispersion elimination processing; the power signal in the time domain needs to be obtained by superimposing the data of each frequency channel, and the input buffer outputs the power spectrum signal in series according to the frequency channel.
[0020] In step S3, the pulsar data buffer unit is a circular buffer, focusing on two types of addresses: (1) the current data address, which is incremented by 1 each time the power spectrum data of all frequency channels is received and then looped; and (2) the de-dispersion address, which is the current data address plus the dispersion delay number.
[0021] In step S4, the data superposition unit superimposes the data currently buffered in the pulsar data buffer unit with the newly received power spectrum data and writes them back to the pulsar data buffer unit; the data superposition and the power spectrum data reception are performed synchronously in a pipeline manner.
[0022] In step S5, the dispersion coefficient storage unit gives the dispersion delay number that needs to be corrected for the current frequency channel, and outputs it to the pulsar data buffer unit to obtain the required dispersion-eliminating address; the dispersion delay number is pre-calculated according to the pulsar dispersion delay calculation formula and written into the dispersion coefficient storage unit. When the dispersion coefficient changes, the stored dispersion delay is dynamically modified.
[0023] In step S6, the output buffer reads the superimposed time domain power signal in the pulsar data buffer unit according to the current data address, and outputs it synchronously.
[0024] Beneficial effects: Compared with the prior art solutions, the present invention has the following advantages:
[0025] 1. The characteristics of serial output of power spectrum data in the data preprocessing part are fully utilized, and the pipeline parallel structure is adopted to greatly improve the processing speed of the dispersion-free data and achieve good real-time processing performance.
[0026] 2. The storage space required for the de-dispersion part is: the dispersion coefficient storage unit with the frequency channel depth related to the frequency resolution requirement; the data buffer unit only related to the time resolution and the maximum dispersion coefficient. This greatly reduces the storage space consumption.
[0027] 3. According to different dispersion coefficients, the dispersion coefficient storage unit can be dynamically configured at any time for flexible adjustment.
[0028] 4. The power signal in the time domain is obtained directly through power spectrum estimation and power spectrum superposition, avoiding the inverse Fourier transform summation operation. BRIEF DESCRIPTION OF THE DRAWINGS
[0029] Figure 1 It is a schematic diagram of the pulsar incoherent real-time de-dispersion system introduced in the present invention;
[0030] Figure 2 This is a schematic diagram of the working principle of the pulsar data buffer unit. DETAILED DESCRIPTION
[0031] The present invention will be further described below in conjunction with the accompanying drawings.
[0032] like Figure 1 As shown, a pulsar incoherent real-time de-dispersion system of a pipeline structure of the present invention includes a pre-processing part and a de-dispersion processing part, wherein:
[0033] The preprocessing part includes an analog-to-digital conversion unit 101 and a power spectrum estimation unit 102. The analog-to-digital conversion unit 101 is used to convert the input pulsar baseband signal into a digitally sampled pulsar signal and input it to the power spectrum estimation unit 102. The power spectrum estimation unit 102 is used to obtain the power spectrum of the pulsar signal and transmit it to the de-dispersion processing part.
[0034] The de-dispersion processing part includes an input buffer 110, a data superposition unit 111, a pulsar data buffer unit 112, a dispersion coefficient storage unit 113 and an output buffer 114. The input buffer 110 is used to output the power spectrum signal output serially in the order of frequency channels to the data superposition unit 111. The data superposition unit 111 is used to process the serially input power spectrum signal in a pipeline manner, superimpose it with the signal buffered in the pulsar data buffer unit 112, obtain a time domain power signal, and transmit it to the output buffer 114. The output buffer 114 is used to buffer the burst-type de-dispersion time domain power signal; the dispersion coefficient storage unit 113 is used to give the position where the current pulsar data needs to be written into the pulsar data buffer unit, and transmit it to the pulsar data buffer unit 112.
[0035] Based on the above-mentioned pulsar incoherent real-time de-dispersion system with pipeline structure, the present invention also proposes a pulsar incoherent real-time de-dispersion method, comprising the following steps:
[0036] Step S1, the RF module outputs a pulsar baseband signal 100, which is input into an analog-to-digital conversion unit 101 for sampling and digitization processing, and outputs a digitally sampled pulsar signal;
[0037] Step S2, the pulsar signal passes through the power spectrum estimation unit 102 to obtain the power spectrum of the pulsar signal, which is transmitted to the input buffer 110 as pre-processed data;
[0038] Step S3, the input buffer 110 buffers the pre-processed data input in burst mode by the data pre-processing part, and outputs the power spectrum signal output serially in the order of frequency channels to the data superposition unit 111;
[0039] Step S4, the data superposition unit 111 processes the serially input power spectrum signal in a pipeline manner, superimposes it with the signal buffered in the pulsar data buffer unit 112, obtains a time domain power signal, and transmits it to the output buffer 114;
[0040] Step S5, the dispersion coefficient storage unit 113 gives the location of the pulsar data buffer unit where the current pulsar data in step S4 needs to be written;
[0041] Step S6: the output buffer 114 buffers the burst-type de-dispersed time domain power signal.
[0042] The present invention will be further described below with reference to the embodiments.
[0043] Example
[0044] This embodiment includes the following steps:
[0045] (1) The pulsar baseband signal 100 is an analog baseband signal processed by the front-stage RF module through low noise amplification, mixing, filtering, etc., and enters the analog-to-digital conversion unit 101 to obtain a pulsar baseband digital signal.
[0046] In this embodiment, the actual radio frequency signal corresponding to the pulsar baseband signal is a signal with a bandwidth of 2190 to 2300 MHz in the S band, and the signal tracks the pulsar J0835-4510.
[0047] In this embodiment, the sampling rate of the analog-to-digital conversion unit 101 is 2.4 GSps, and the number of bits is 8 bits. After the analog-to-digital conversion unit 101 converts and outputs 8 channels of 300 MSps pulsar digital signals, the total signal bandwidth is 1.2 GHz.
[0048] (2) In the power spectrum estimation part, the Barlett periodogram method is used. The formula for estimating the signal power spectrum is as follows:
[0049]
[0050] Among them, x is a time domain signal sequence with a total length of N = K × L points. The signal is divided into K segments, and each segment of the sequence with a length of L is subjected to L-point Fourier transform. j is an imaginary number, e is a natural constant, and ω is an angular frequency. Formula (1) involves two operations: Fourier transform of the signal and spectrum superposition and summation. Compared with the periodogram method, the Barlett periodogram method has a smaller deviation in estimating the signal power spectrum. At the same time, the number of spectrum superposition and summation K and the number of FFT points L can be flexibly configured, and correspondingly, the appropriate time resolution and frequency resolution can be selected to adapt to different hardware performance and post-processing data volume.
[0051] In this embodiment, the Barlett periodogram method is used to estimate the power spectrum of the signal, the Fourier transform is implemented by a multi-phase fast Fourier transform unit, the FFT point number is 2048 points, and the spectrum of 8 channels of 300MSps signals is obtained. After the spectrum superposition unit performs operations such as modulus superposition, the estimated power spectrum signal is obtained.
[0052] (3) The input buffer 110 can buffer the input power spectrum signal. Theoretically, the number of spectrum superposition and summation times K in step (2) can be 1, and no time resolution loss will occur. At the same time, the input buffer is not necessary. However, due to the performance limitation of the post-processing hardware, the size of K can be appropriately increased.
[0053] In this embodiment, FPGA is used to implement it. Due to the difficulty of timing convergence, the power spectrum signal works at f s =300MHz clock domain, the anti-dispersion processing part works at f s / 2=150MHz clock domain, at this time, K≥2 should be set. The power spectrum data is periodically and continuously output for 2048 frequency channels, and there will be no output for a period of time, so input buffering is required. FIFO can be used to ensure that the dispersion elimination processing part works at f s / K min Clock domain.
[0054] (4) The dispersion effect of pulsars is due to the fact that the high-frequency component in the pulsar signal propagates faster than the low-frequency component. The propagation time difference between the low-frequency signal f1 and the high-frequency signal f2 in the interstellar medium is:
[0055]
[0056] Where c is the speed of light in a vacuum; e is the charge of an electron; and D is the dispersion of the observed pulsar signal in cm. -3 ·pc; m is the mass of the electron.
[0057] In the pulsar dedispersion process, a frequency channel f chan Relative to the reference channel f ref The time delay can be calculated using the dispersion formula:
[0058]
[0059] Among them, K DM is the dispersion constant; D is the dispersion of the observed pulsar signal, in cm -3 pc; the frequency unit is MHz. The dispersion constant is:
[0060]
[0061] If the time resolution of the de-dispersion process is t res , then the number of data points that the high-frequency component needs to lag behind is:
[0062]
[0063] The dispersion coefficient storage unit 113 stores the number of data points that need to be delayed for each frequency channel.
[0064] In this embodiment, the number of data delay points is calculated by the host computer and can be dynamically written into the dispersion coefficient storage RAM.
[0065] (5) The data superposition unit 111, the pulsar data buffer unit 112 and the dispersion coefficient storage unit 113 jointly perform dispersion elimination processing. The working principle of the pulsar data buffer is as follows: Figure 2The final time domain power signal is obtained by adding the power spectrum signals of each frequency channel. When the number of frequency channels is N and there is no dispersion effect, the time domain power signal is:
[0066]
[0067] Among them, P(i) is the power spectrum signal.
[0068] However, due to the influence of dispersion, the actual time domain power signal is:
[0069]
[0070] correspond Figure 2 In the case of no dispersion, the data of frequency channel i is written to address cur, but due to the existence of dispersion, the data needs to be delayed by delay(i), so the destination write address is dest(i)=cur+delay(i).
[0071] Figure 2 The dedispersion process is described as follows:
[0072] delay(i) is given by the dispersion coefficient storage unit;
[0073] Calculate the destination write address dest(i). When dest(i)>max, dest(i)=dest(i)-max-1, that is, the loop starts from 0;
[0074] The power spectrum data P(i) is added to the data buffered in the pulsar data buffer unit and written to the address dest(i);
[0075] When the data of N frequency channels are all written into the pulsar data buffer unit, the writing of the power spectrum signal required for power signal calculation is completed;
[0076] At this time, the data stored at address cur is That is, the power signal at the current moment is output to the subsequent module;
[0077] The address cur=cur+1 is the same as the destination write address. When it exceeds max, the cur loop starts from 0.
[0078] It should be noted that delay(i) cannot exceed max, which will have a certain impact on the dispersion range that can be eliminated. The dispersion amount D range in formula (3) is jointly affected by the observation frequency range, time resolution and max.
[0079] The operations required for the above-mentioned dispersion elimination process can be summarized as: reading the dispersion coefficient, calculating the destination address, reading the buffered data, data superposition, and data writing back. It is actually implemented in a pipeline manner. For example, when calculating the destination address, the next dispersion coefficient can be read, making full use of the pipeline parallelism to improve the real-time performance of data processing.
[0080] In this embodiment, the number of frequency channels is 2048, the maximum is 65535, the time resolution is set to 34.1us, and the observation frequency range is 2190~2300MHz. At this time, the range of the dispersion that can be eliminated is D≤27670cm -3 ·pc
[0081] (6) The output buffer 114 buffers the power signal after de-dispersion.
[0082] In this embodiment, the output buffer is implemented in FIFO mode, the anti-dispersion module writes data into the output FIFO, and the host computer reads the power signal through the output FIFO.
[0083] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.
Claims
1. A pulsar incoherent real-time dispersion elimination system with a pipeline structure, characterized by: It includes a pre-processing part and a dispersion elimination processing part, wherein: The preprocessing part comprises an analog-to-digital conversion unit (101) and a power spectrum estimation unit (102), wherein the analog-to-digital conversion unit (101) is used to convert an input pulsar baseband signal into a digitally sampled pulsar signal and input the signal to the power spectrum estimation unit (102), and the power spectrum estimation unit (102) is used to obtain a power spectrum of the pulsar signal and transmit the power spectrum to the de-dispersion processing part; The de-dispersion processing part comprises an input buffer (110), a data superposition unit (111), a pulsar data buffer unit (112), a dispersion coefficient storage unit (113) and an output buffer (114). The input buffer (110) is used to output a power spectrum signal serially output in the order of frequency channels to the data superposition unit (111). The data superposition unit (111) is used to process the serially input power spectrum signal in a pipeline manner, superimpose it with the signal buffered in the pulsar data buffer unit (112), obtain a time domain power signal, and transmit it to the output buffer (114). The output buffer (114) is used to buffer the burst-type de-dispersion time domain power signal. The dispersion coefficient storage unit (113) is used to provide a position where the current pulsar data needs to be written into the pulsar data buffer unit, and transmit it to the pulsar data buffer unit (112).
2. A pulsar incoherent real-time de-dispersion method based on the system of claim 1, characterized in that: The following steps are involved: Step S1, the radio frequency module outputs a pulsar baseband signal (100), which is input into an analog-to-digital conversion unit (101) for sampling and digitization processing, and outputs a digitally sampled pulsar signal; Step S2, the pulsar signal passes through a power spectrum estimation unit (102) to obtain a power spectrum of the pulsar signal, which is transmitted to an input buffer (110) as pre-processed data; Step S3, the input buffer (110) buffers the pre-processed data input in burst mode by the data pre-processing part, and outputs the power spectrum signal output serially in the order of frequency channels to the data superposition unit (111); Step S4, the data superposition unit (111) processes the serially input power spectrum signal in a pipeline manner, superimposes it with the signal buffered in the pulsar data buffer unit (112), obtains a time domain power signal, and transmits it to the output buffer (114); Step S5, the dispersion coefficient storage unit (113) gives the position of the pulsar data buffer unit where the current pulsar data in step S4 needs to be written; Step S6, the output buffer (114) buffers the burst-type de-dispersed time domain power signal.
3. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In the step S1, the analog-to-digital conversion unit (101) converts the pulsar digital signal with a high-speed sampling rate into multi-phase parallel data with a lower sampling rate.
4. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In step S2, the power spectrum estimation unit (102) is based on the Barlett periodogram method. While performing power spectrum estimation, it utilizes the data superposition characteristics of the Barlett periodogram and controls the time resolution of the system by dynamically configuring the number of superposition times, thereby reducing the data processing amount of the subsequent modules.
5. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In step S3, the input buffer (110) completes the buffering of the input power spectrum signal to ensure the synchronization of the subsequent dispersion elimination processing; the power signal in the time domain needs to be obtained by superimposing the data of each frequency channel, and the input buffer (110) outputs the power spectrum signal in series according to the frequency channel.
6. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In step S3, the pulsar data buffer unit (112) is a circular buffer, focusing on two types of addresses: (1) the current data address, which is incremented by 1 each time the power spectrum data of all frequency channels are received and then circulated; and (2) the de-dispersion address, which is based on the current data address and the dispersion delay number is added.
7. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In step S4, the data superposition unit (111) superimposes the data currently buffered in the pulsar data buffer unit (112) with the newly received power spectrum data and writes them back to the pulsar data buffer unit (112); the data superposition and the power spectrum data reception are performed synchronously in a pipeline manner.
8. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In the step S5, the dispersion coefficient storage unit (113) gives the dispersion delay number that needs to be corrected for the current frequency channel and outputs it to the pulsar data buffer unit (112) to obtain the required dispersion elimination address; The dispersion delay number is pre-calculated according to the pulsar dispersion delay calculation formula and written into the dispersion coefficient storage unit (113). When the dispersion coefficient changes, the stored dispersion delay is dynamically modified.
9. The pulsar incoherent real-time de-dispersion method according to claim 2, characterized in that: In step S6, the output buffer (114) reads the superimposed time domain power signal in the pulsar data buffer unit (112) according to the current data address, and outputs it synchronously.
