A non-stationary channel simulation method and simulator based on double scattering clusters

Abstract

A non-stationary channel simulation method and simulator based on double scattering clusters, including a processing system platform, a programmable logic platform, and an integrated radio frequency transceiver platform, can simulate channels in a variety of special scenarios. According to the user configuration file input by the user, including the layout of the mobile network, the movement trajectory, scene parameters, etc., it can accurately and with low latency simulate the impact of a non-stationary dynamic fading channel based on double scattering clusters on the channel. Using a combined software and hardware embedded architecture, the calculation of channel model parameters is performed by the CPU, and the convolution operation in the channel model is implemented by hardware with high parallelism and low latency. The present invention accurately and with low latency simulates the condition of a non-stationary dynamic fading channel based on double scattering clusters according to the user configuration file input by the user, and is used for on-site performance testing and verification of wireless communication systems and communication devices.

Description

Technical Field

[0001] The present invention relates to the technical field of channel modeling, and in particular to a non-stationary channel simulation method and simulator based on double scattering clusters. Background Art

[0002] In special environments such as urban streets, inside buildings, and urban underground pipe networks, when signals propagate in such wireless environments, signal absorption or reflection may occur when encountering obstacles, which results in multiple propagation paths between the transmitter and the receiver, and the propagation distances and incident angles of the propagation paths are all different. In recent years, through a large number of tests on the new generation of communication environments such as 5G, the results show that the three-dimensional channel model based on geometry can more accurately represent the characteristics of wireless channels. This type of channel model combines the geometry and sparsity of the channel, abstracts the absorption and scattering of obstacles as scattering by scattering points, and then groups multiple scattering points into a scattering cluster. The paths of scattering by the scattering points within each cluster have similar time delays and angles of arrival. During the development and testing of wireless communication devices, they need to be tested in a real channel environment, but this testing method is costly and time-consuming. To reduce research costs and shorten the development cycle, it is an effective method to simulate the above channel characteristics through channel modeling and channel simulators.

[0003] The Chinese patent "Wireless Channel Time Delay and Fading Precise Simulation Device and Method" (application number CN202010799780.6) discloses a simulation method and structure of a wireless channel. The specific implementation steps are as follows: First, the user sets communication scenario parameters on the host computer, and then the parameter calculation unit calculates parameters according to the user input to obtain parameters such as time delay, Doppler, and fading. Then, it is transmitted to the time delay module through the PCIe bus. This module divides the time delay parameters into several precision time delays, and then outputs the signal after multi-precision delay to the multiplication unit. At the same time, the channel fading module generates a channel fading factor according to the fading parameters transmitted by the parameter calculation unit, and then outputs the fading factor to the multiplication unit. Finally, the result obtained by the multiplier is sent to the automatic gain module for adjustment and interpolated to output at the digital-to-analog conversion rate. The disadvantage of this design is that in today's 5G or 6G communication environments, the simulated channels are not accurate enough, and the near-field effect, spatio-temporal-frequency non-stationarity, and spatial consistency of the scattering clusters are not well realized.

[0004] The Chinese patent "Wireless fading channel simulation method and channel simulator with multiple independent signals in parallel" (Application No. CN202011301021.9) discloses a wireless fading channel simulation method and channel simulator with signals in parallel. The specific implementation steps are as follows: First, perform multipath delay on each path of the input signal, and adjust the phase and amplitude of each delayed signal. Then, by reading data of various spectral patterns stored in the RAM, generate colored noise that matches the sampling rate of the path signal after appropriate interpolation by FARROW. After that, multiply the signals of different paths by the colored noise with a matching rate; superimpose each path of the multiplied signal and merge them into one path of signal, and pass the merged signal through a large-scale fading unit, and finally output the signal after simulation. The deficiencies of this method are that there is a lack of input of angular signals, the accuracy of the delay module is not fine enough, and there is a lack of a high-speed design scheme, which may not be able to accurately simulate in the case of an extremely short channel coherence time. Summary of the Invention

[0005] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide a non-stationary channel simulation method and simulator based on double scattering clusters, which can simulate channels in a variety of special scenarios. According to the user configuration file input by the user, including the layout of the mobile network, the movement trajectory, the scene parameters, etc., accurately and with low delay simulate the influence of a non-stationary dynamic fading channel based on double scattering clusters on the channel. Using a combined software and hardware embedded architecture, the calculation of the channel model parameters is handed over to the processing system platform (PS) for calculation, and the convolution operation in the channel model is implemented by the programmable logic platform (PL), which is used for on-site testing and verification of the performance of wireless communication systems and communication devices.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0007] A non-stationary channel simulation method based on double scattering clusters, comprising the following steps:

[0008] Step 1, the user creates a network layout file on the PC, and the file contains the initial positions, movement speeds and trajectories of the transceiver ends and the scene parameters on the path trajectory; first generate multiple independent 3D space-related random variables with spatial consistency, and use these random variables to instantiate the scene parameters in the network layout file;

[0009] Step 2, initialize multiple groups of central scattered sub-paths for the transceiver ends, and the central scattered sub-path is defined as the propagation path scattered by the center of the double scattering cluster, and then assign power values to each central scattered sub-path; during this process, keep the initialized delay and angle values unchanged;

[0010] Step 3, correct the delay, angle and power of the central scattered sub-path according to the scene parameters instantiated in Step 1 to accurately reflect the scene parameters;

[0011] Step 4: Add the remaining scatterer paths to each pair of double-scattering clusters, and calculate the positions of all scatter points in each pair of double-scattering clusters by combining the delay and angle obtained in Step 3. When the mobile terminal moves to different channel sampling points, update the angle information and delay phase information according to the geometric relationship between the transceiver and the movement of the double-scattering clusters.

[0012] Step 5: Incorporate the influence of the remaining scene parameters, i.e., path loss and shadow fading, into each path. Since the channel coefficient values vary with time due to the movement, Steps 4 and 5 need to be performed at each sampling point, and a cosine function is used to simulate the smooth transition of the channel parameters in the transition area of the scene change. Finally, the channel impulse response at each sampling point is obtained.

[0013] Step 6: Generate the channel parameter matrix for each group of double-scattering cluster paths at each sampling point according to the results, and store it in the RAM of the PC in the form of a binary file. Use DMA to transfer the stored channel parameters to the FIFO in the FPGA to achieve information interaction from software to hardware. The signal to be simulated is input to the integrated radio frequency module through the radio frequency interface, sampled into digital baseband I / Q signals through a series of transformations, and the double-edge signal is converted into a single-edge signal easy for baseband processing through the IDDR module, and finally sent to the baseband channel simulation module. The baseband channel simulation module is divided into a delay module, a floating-point module, an attenuation module, a frequency and phase offset module, a white noise module, an integration module, and a floating-point module.

[0014] Step 7: The FIFO for caching parameters is connected to the baseband channel simulation module. The baseband signal enters the delay module, and the delay module performs multi-precision delay on each path signal according to the read value, and then inputs the delayed signal into the floating-point module. The floating-point module converts the fixed-point number into a floating-point number for subsequent high-precision mathematical operations.

[0015] Step 8: The signal output by the floating-point module passes through the frequency and phase offset module successively. The frequency and phase offset module adjusts the amplitude, frequency, and phase of the signal according to the stored parameters, and then inputs the adjusted multi-path signals into the integration module. The data represented by each cluster is summed with the noise generated by the white noise module, and one signal is output.

[0016] Step 9: Input the output single signal into the fixed-point number module. The fixed-point number module converts the floating-point data into a truncated fixed-point number, and finally up-converts it through the radio frequency module to restore and output the radio frequency signal.

[0017] A non-stationary channel simulator based on double-scattering clusters, comprising:

[0018] A processing system platform for dynamically generating channel coefficients of a non-stationary channel based on double scattering clusters; the processing system platform includes a user-configured channel parameter module, a channel coefficient calculation module, a DMA controller, and memory and memory control, and the channel coefficient calculation module is implemented by the logic of the above steps 1-step 5;

[0019] A programmable logic platform, including a baseband channel simulation module, a radio frequency register configuration module, and a channel parameter configuration module; the baseband channel simulation module is responsible for channel processing functions, performing a series of mathematical calculations on the baseband data received from the radio frequency end to implement functions such as signal attenuation, phase shift, frequency offset, and adding white noise, and the baseband channel simulation module is implemented by the logic of the above steps 7-step 9.

[0020] An integrated radio frequency transceiver platform for receiving and transmitting radio frequency signals of a communication system or device to be tested, and converting the radio frequency signals into baseband I / Q signals for processing by the FPGA programmable logic platform through processes such as amplification, mixing, and filtering, which is implemented by the logic of the above step 6.

[0021] The beneficial effects of the present invention are:

[0022] The main content of the present invention is divided into two major parts. The first part is the processing system platform, which uses an independently designed non-stationary channel model and simulation method based on double scattering clusters; the second part is the programmable logic platform, which realizes functions through logic circuits. According to the user configuration file input by the user, the present invention accurately and with low latency simulates the situation of a non-stationary dynamic fading channel based on double scattering clusters, and is used for field testing and verification of the performance of wireless communication systems and communication devices. It has the following advantages.

[0023] First: In terms of architecture, the present invention is a software-hardware combined embedded system. The channel coefficients represented by the channel model are calculated by the CPU in the processing system platform, breaking through the limitation of the difficulty of modifying system parameters in a pure hardware simulator, ensuring the flexibility of modeling, which makes the channel simulator highly configurable and can simulate a very wide range of channel types.

[0024] Second: Mathematical operations such as channel convolution in the channel simulator are implemented by the programmable logic platform in a decoupled manner, taking advantage of the high-speed parallel computing and high robustness of hardware, so that when the real-time signal to be simulated convolves with the channel coefficients, it has very low processing latency, which is more conducive to simulating high-dynamic channels with low coherence time and is used to solve the technical problems of existing channel simulators.

[0025] Third: In the construction of the channel model used in the present invention, a double-scattering cluster model of the near-field scattering clusters at the transceiver is modeled. In the problem of obtaining the positions of the scattering clusters based on the large-scale channel parameters, the present invention uses the idea of maximum likelihood estimation (MLE). Compared with the traditional method, new planning objectives and near-field constraints are added, making the simulation of communication environments with obvious near-field effects more accurate. For example, it can better describe the spatial non-stationarity of the channel along the antenna in large-scale MIMO.

[0026] Fourth: In today's 5G or 6G communication environments, the channel model is relatively accurate, and good implementations are achieved for near-field effects, spatio-temporal-frequency non-stationarity of scattering clusters, and spatial consistency of channel parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 It is a physical structure diagram of the channel simulator for the embodiment.

[0028] Figure 2 It is a schematic diagram of a typical double-scattering cluster non-stationary channel model.

[0029] Figure 3 It is a program flow chart inside the channel coefficient calculation module for the embodiment.

[0030] Figure 4 It is a structure diagram of the baseband channel simulation module for the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0031] The present invention will be described in detail below with reference to the drawings and embodiments.

[0032] As Figure 1 shown, a non-stationary channel simulator based on double-scattering clusters includes a processing system platform, a programmable logic platform, and an integrated radio frequency transceiver platform; the basic channel model it simulates is a non-stationary channel model based on double-scattering clusters, and its typical schematic diagram is as Figure 2 shown;

[0033] The processing system platform is used to dynamically generate the channel coefficients of a non-stationary channel based on double-scattering clusters; the processing system platform includes a user-configured channel parameter module, a channel coefficient calculation module, a DMA controller, and memory and memory control. Among them, the channel coefficient calculation module is implemented by the logic of steps 1 - 5, and its program flow chart is as Figure 3 shown;

[0034] A programmable logic platform, including a baseband channel simulation module, a radio frequency register configuration module, and a channel parameter configuration module; the baseband channel simulation module is responsible for channel processing functions, performing a series of mathematical calculations on the baseband data received from the radio frequency end to achieve functions such as signal attenuation, phase offset, frequency offset, and adding white noise. The baseband channel simulation module is implemented by the logic from step 7 to step 9; the baseband channel simulation unit in the programmable logic platform, whose structure diagram is as Figure 4 shown.

[0035] An integrated radio frequency transceiver platform, used to receive and transmit radio frequency signals of the communication system or communication device to be tested, and convert the radio frequency signals into baseband I / Q signals for the programmable logic platform to process through processes such as amplification, mixing, and filtering, which is implemented by the logic of step 6.

[0036] A non-stationary channel simulation method based on double scattering clusters, including the following steps:

[0037] Step 1, first, the user creates a network layout file on the PC. The file contains the initial positions, movement trajectories, movement speeds of the transceiver ends, parameters of each antenna, and scene parameters on the path trajectory, etc.; generate multiple independent spatially consistent 3D space-correlated random variables, and use these random variables to instantiate the scene parameters in the network layout file. The scene parameters that need to be instantiated include delay spread, angular spread, cluster angular spread, Rice factor, and shadow fading;

[0038] For the instantiation modeling of scene parameters, taking the delay spread as an example, its definition is a lognormal distribution:

[0039]

[0040] where DS(s) is the delay spread, s = {x t , y t , z t , x r , y r , z r} is the transceiver position vector, (x t , y t , z t ) and (x r , y r , z r ) are the position coordinates of the transmitter and receiver respectively; Z DS (s) is a spatially correlated normal distribution and is expressed as:

[0041] Z DS (s) = μ DS + μ d log 10 (d TR ) + X DS(s){σ DS +σ d log 10 (d TR )}

[0042] where d TR is the distance between the transmitter and the receiver, which can be obtained from the s vector; μ DS and σ DS are the normalized distances respectively. When d TR = 1m, they are the mean and standard deviation of the Gaussian variable; μ d and σ d are the distance scaling ratio coefficients of the mean and standard deviation respectively; is a standard normal distribution with six-dimensional spatial correlation. According to the reciprocity principle of transmission and reception, X DS can be expressed as:

[0043]

[0044] where is the spatially correlated random variable generated in step 1, and its distribution is a standard normal distribution. The spatial autocorrelation function ρ(Δd) of

[0045]

[0046] is modeled as: -1 where Δd is the Euclidean distance between two spatial points, and d1 is the decorrelation distance, defined as the spatial Euclidean distance when the correlation coefficient is e -1 ; when Δd≥d1, the spatial autocorrelation of the two random variables can be considered very low and they can be approximately independent. When Δd<d1, the correlation between two nearby points is described according to the proportionality coefficient k;

[0047] Step 2: Initialize the central scatterer paths in multiple double-scattering clusters for the transmitter and receiver. The initialized normalized parameters include the delay, horizontal departure angle, vertical departure angle, horizontal arrival angle, and horizontal departure angle of the central scatterer paths. The initialization method is similar to the instantiation of the scenario parameters in step 1, only the distribution models of different parameters need to be changed, and then a power value is assigned to each central scatterer path; the power P′ m of the m-th central scatterer path in the double-scattering cluster is assigned as:

[0048]

[0049] where τ m is the path delay, r τ is the delay distribution ratio factor, and DS(s) is the delay spread instantiated in step 1;

[0050] Step 3: Modify the delay, angle, and power of the central sub-path according to the instantiated scenario parameters in Step 1 to accurately reflect the scenario parameters. The scenario parameters are a set of samples with statistical characteristics obtained from measuring the actual environment, while the delay, angle, and power of the central sub-path are a set of small-scale random samples. The latter needs to be modified to make them consistent statistically.

[0051] In power modification, it is necessary to refer to the Rice factor of the scenario parameters. First, generate the power of the line-of-sight path P1′, and then normalize the power of M - 1 central scattered sub-paths:

[0052]

[0053]

[0054] In angle parameter modification, taking the horizontal departure angle as an example here, it is necessary to refer to the horizontal departure angle expansion of the scenario parameters. Since the angle is periodic, it is necessary to first obtain the initial angle mean and then calculate the central offset value of all angles. The calculation is as follows:

[0055]

[0056] where arg is to find the argument of a complex number, and j is the imaginary unit. is the central offset value of the horizontal departure angle, P m is the path power normalized in Step 3, and α′ m is the initial horizontal departure angle generated in Step 2; calculate the second moment of the offset angle, that is, the horizontal departure angle expansion is:

[0057]

[0058] According to the above calculation results, the initial horizontal departure angle α′ m is scaled to α m , and the scaling method is:

[0059]

[0060] where AS is the horizontal departure angle expansion instantiated in Step 1, and the modification of the delay parameter is similar to that of the angle.

[0061] Step 4: Add the characteristics of mobility (Doppler shift) and near-field spherical waves to the channel. Add remaining scattered sub-paths to each pair of double-scattering clusters. Since the signal frequency increases and the channel delay resolution improves, it is reasonably assumed that the closely scattered sub-paths originating from the same double-scattering cluster can be approximately resolved in the delay domain and arrive from different directions within a small range. Therefore, construct the remaining scattered sub-paths based on the central scattered sub-path. Taking the departure angle at the transmitter as an example, the horizontal departure angle α m,n , the vertical departure angle βm,n Expressed as:

[0062]

[0063]

[0064] for m>1 and n>1

[0065] The subscripts m and n represent the nth scatterer path in the double-scattering cluster numbered m; c α , c β are the horizontal and vertical cluster angular spreads respectively, are the offsets of the horizontal and vertical departure angles under the condition of unit angular spread respectively; the power P of the remaining scatterer paths m,n is allocated as:

[0066]

[0067] ρ m (α m,n , β m,n ) is the cluster angular power spectrum, P m is the power of the central scatterer path generated in step 3, and the delay τ of the mth central scatterer path generated in step 3 m,1 can be expressed as:

[0068]

[0069]

[0070] is the vector from the transmitting end to the nth scattering point in the mth scattering cluster of the first hop, is the vector from the first-hop scattering point to the last-hop near-field scattering point, is the vector from the last-hop near-field scattering point to the receiving end. When n = 1, it represents the central scatterer path; is the virtual link from the first-hop near-field scattering point to the last-hop scattering point, is the additional delay of the virtual link, following a unilateral exponential distribution;

[0071] Combining the delay and angle obtained in step 3, calculate a set of vectors to represent the positions of all scattering points in the double-scattering cluster. The position of the scattering point corresponding to the mth central scatterer path can be determined by solving a programming problem:

[0072]

[0073]

[0074] δ is the path loss parameter, D is the vector pointing from the transmitter to the receiver, (l 1max , l 2max ), (l 3max , l 4max ) are the distance ranges from the transmitter and the receiver to the nearest scatterer respectively, and specific values should be given according to the environmental characteristics;

[0075] Taking the near - field scatterer of the transmitter in the double - scatter cluster as an example, the position vector of the remaining scatterers is:

[0076]

[0077] α m,n , β m,n represent the horizontal and vertical departure angles of the remaining scattered sub - paths generated in step 4 respectively;

[0078] When the mobile terminal moves to different channel sampling points s, the initial scattered sub - path parameters calculated in the above steps are updated:

[0079]

[0080]

[0081]

[0082]

[0083] where are the motion vectors of the transmitter - receiver and the double - scatter cluster respectively;

[0084] Step 5, add path loss and shadow fading to each path; due to the time - varying caused by the numerical movement of the channel coefficient, steps 4 and 5 should be performed at each sampling point within the channel - correlation distance interval, and a cosine - raised function is used to simulate the smooth transition of the channel parameters in the transition area of the scene change. The impulse response h(s,τ) of the channel at time s can be expressed as:

[0085]

[0086] where P 1，1 is the power of the direct - line - of - sight path, P m,n is the power of the non - direct - line - of - sight double - scatter - cluster path, and the impulse response h LoS (s,τ) of the path is:

[0087]

[0088] where f c represents the carrier frequency, and the impulse response of the nth sub - path in the mth double - scatter cluster of the non - direct - line - of - sight path is expressed as:

[0089]

[0090]

[0091] where represents the random phase generated by path scattering, modeled as a uniform distribution U(0, 2π); β m,n (s) is defined as the large-scale effect of shadow fading and path loss, expressed as:

[0092]

[0093] where a, b, and c are scene characteristic coefficients, usually determined by measurement; d m,n,s represents the path length, and the shadow fading is modeled by the random variable z m,n,s and follows a log-normal distribution;

[0094] Step 6: Generate the channel parameter matrix for each pair of clusters according to the result and store it in the RAM of the PC in the form of a binary file; the channel parameters are stored in the form of the following binary data:

[0095] [A m,n,s B m,n,s τ m,n,s 3×m×n×s

[0096] where A m,n,s , B m,n,s are calculated from the channel impulse response, representing the overall amplitude and phase effects of the channel at the sampling point s, i.e., the channel coefficients; then, through direct memory access (DMA), high-speed data transfer between the memories of both the processing system (PS) and the programmable logic (PL) is provided; the initialization of this transfer action is completed by the APU in the processing system platform, and the transfer action itself is continuously completed by the DMA controller. This transfer method is directly controlled by the CPU, so there is no corresponding interrupt handling. A direct data transfer channel is opened for the memory by the hardware, greatly improving the data transfer efficiency of the overall system.

[0097] The signals to be simulated are input into each channel through the radio frequency interface, down-converted and sampled into digital baseband signals through a series of operations, the double-edge signals are converted into single-edge signals easy for baseband processing through IDDR, and finally sent to the baseband channel simulation module; the baseband channel simulation module is divided into a delay module, a floating-point module, an attenuation module, a frequency and phase offset module, a white noise module, an integration module, and a floating-point module;

[0098] Step 7: The cache parameter FIFO is connected to the baseband channel simulation module; the fixed-point data τ m,n,s ​Implement different delays for the analog signal s(t) to be simulated through Block ram; then input the delayed signal into the number system conversion module. Since the attenuation differences between various paths are relatively large, especially for the direct view and non-direct view paths, the differences between different paths of A m,n,s are very large. Therefore, the adjustable number system conversion module is used to convert s(t) and A m,n,s into floating-point numbers in a suitable format to meet most cases, reduce the problems of overflow or loss of precision, and eliminate the need to add a module for handling overflow problems;

[0099] Step 8, the signal output by the number system conversion module passes through the attenuation module and the frequency offset module in sequence. In addition to using A m,n,s data, the attenuation module also requires an additional gain coefficient A to offset the multi-stage hardware attenuation and power amplification caused by the input of the radio frequency signal to the baseband and the output from the baseband to the radio frequency. Then, the adjusted multi-path signals are input into the integration unit, and the data of each pair of clusters are summed with the noise generated by the noise module, and a single signal is output;

[0100] Gaussian white noise module; the generation principle is as follows: Use a linear feedback shift register LSFR to generate a set of numbers with a very large period, which can be approximately regarded as random numbers, that is, pseudo-random numbers. Since the number of occurrences of various values within one period of the generation principle is strictly equal, these numbers can be regarded as uniformly distributed samples. Then, through the Box-Muller transform, they are mapped into a sample of the Gaussian distribution. Finally, Gaussian white noise with different powers is output through the gain module;

[0101] Step 9, input the output single signal into the fixed-point number module. The fixed-point number module sums the data of each path and converts it into a suitable fixed-point number, and outputs it to the radio frequency module. By adopting a highly parallel computing method, the processing delay is greatly reduced.

Claims

1. A non-stationary channel simulation method based on double scattering clusters, characterized in that It includes the following steps: Step 1: The user creates a network layout file on the PC. The file contains the initial positions, moving speeds, trajectories of the transceiver, and the scene parameters on the path trajectory. First, generate multiple independent 3D space-related random variables with spatial consistency, and use these random variables to instantiate the scene parameters in the network layout file. Step 2: Initialize multiple groups of central scatterer paths for the transceiver. The central scatterer path is defined as the propagation path scattered by the center of the double-scattering cluster, and then assign power values to each central scatterer path. During the process, keep the initialized delay and angle values unchanged. Step 3: Correct the delay, angle, and power of the central scatterer path according to the scene parameters instantiated in Step 1 to accurately reflect the scene parameters. Step 4: Add remaining scatterer paths to each pair of double-scattering clusters, and combine the delay and angle obtained in Step 3 to calculate the positions of all scatter points in each pair of double-scattering clusters. When the mobile terminal moves to different channel sampling points, update the angle information and delay phase information according to the geometric relationship between the transceiver and the movement of the double-scattering cluster. The mobility (Doppler shift) and the characteristics of the near-field spherical wave are added to the channel: Residual scatterer paths are added to each pair of double-scattering clusters. Since the signal frequency increases, the channel delay resolution improves. It is reasonably assumed that the closely spaced scatterer paths originating from the same double-scattering cluster are approximately resolved in the delay domain and arrive from different directions within a small range. Therefore, the residual scatterer paths are constructed based on the central scatterer path. Taking the departure angle at the transmitter end as an example, the horizontal departure angle α m,n , the vertical departure angle β m,n are expressed as: The subscripts m and n represent the n-th scatterer path in the double-scattering cluster numbered m; c α and c β are the horizontal and vertical cluster angular spreads respectively, are the offsets of the horizontal and vertical departure angles under the condition of unit angular spread respectively; the power P of the remaining scatterer paths m,n is allocated as: ρ m (α m,n ,β m,n ) is the cluster angular power spectrum, and P m is the power of the central scatterer path generated in step 3. The delay τ m,1 of the m-th central scatterer path generated in step 3 is expressed as: is the vector from the transmitting end to the n-th scattering point in the m-th scattering cluster of the first hop, is the vector from the first-hop scattering point to the last-hop near-field scattering point, is the vector from the last-hop near-field scattering point to the receiving end. When n = 1, it represents the central scatterer path; is the virtual link from the first-hop near-field scattering point to the last-hop scattering point, is the additional delay of the virtual link, which follows a unilateral exponential distribution; Calculate a set of vectors in combination with the delay and angle obtained in step 3 to represent the positions of all scatterers in the double-scattering cluster, and determine the position of the scatterer corresponding to the m-th central scatterer path by solving a programming problem: δ is the path loss parameter, D is the vector pointing from the transmitter to the receiver, (l 1max , l 2max ), (l 3max , l 4max ) are the distance ranges from the transmitter and the receiver to the nearest scatterer respectively, and specific values should be given according to the environmental characteristics; Taking the near-field scattering points at the transmitting end in the double scattering clusters as an example, the position vectors of the remaining scattering points are as follows: α m,n and β m,n respectively represent the horizontal and vertical departure angles of the remaining scattered sub-rays generated by step 4; When the mobile terminal moves to different channel sampling points s, update the initial scatterer path parameters calculated in the above steps: wherein are the motion vectors of the transceiver and the double-scattering cluster, respectively Step 5: Add the influence of the remaining scene parameters to each path, that is, path loss and shadow fading. Since the channel coefficient values change with time due to movement, Steps 4 and 5 need to be performed at each sampling point, and a cosine function is used to simulate the smooth transition of the channel parameters in the transition area of the scene change. Finally, the channel impulse response of each sampling point is obtained. Step 6: Generate a channel parameter matrix for each group of double-scattering cluster paths for each sampling point according to the results, and store it in the RAM of the PC in the form of a binary file. Use DMA to move the stored channel parameters to the FIFO in the FPGA to achieve information interaction from software to hardware. The signal to be simulated is input into the integrated radio frequency module through the radio frequency interface, sampled into a digital baseband I / Q signal through a series of transformations, and the double-edge signal is converted into a single-edge signal easy to process by the baseband through the IDDR module, and finally sent to the baseband channel simulation module. The baseband channel simulation module is divided into a delay module, a floating-point module, an attenuation module, a frequency and phase offset module, a white noise module, an integration module, and a floating-point module. Step 7: The FIFO for caching parameters is connected to the baseband channel simulation module. The baseband signal enters the delay module, and the delay module performs multi-precision delay on each path of the signal according to the read value, and then inputs the delayed signal into the floating-point module. The floating-point module converts the fixed-point number into a floating-point number for subsequent high-precision mathematical operations. Step 8: The signal output by the floating-point module passes through the frequency and phase offset module successively. The frequency and phase offset module adjusts the amplitude, frequency, and phase of the signal according to the stored parameters, and then inputs the adjusted multi-path signals into the integration module to sum the data represented by each cluster and the noise generated by the white noise module, and outputs one signal. Step 9: Input the output single signal into the fixed-point module again. The fixed-point module converts the floating-point data into a truncated fixed-point number, and finally up-converts it through the radio frequency module to restore and output the radio frequency signal.

2. The method according to claim 1, characterized in that, It includes the following steps: Step 1: First, the user creates a network layout file on the PC. The file contains the initial positions, movement trajectories, movement speeds of the transceiver, the parameters of each antenna, and the scene parameters on the path trajectory. Then, multiple independent 3D space-correlated random variables with spatial consistency are generated, and these random variables are used to instantiate the scene parameters in the network layout file. The scene parameters to be instantiated include delay spread, angular spread, cluster angular spread, Rice factor, and shadow fading. For the instantiation modeling of scene parameters, taking the delay spread as an example, its definition is a lognormal distribution: where DS(s) is the delay spread, s = {x t , y t , z t , x r , y r , z r} is the position vector of the transceiver, (x t , y t , z t ) and (x r , y r , z r ) are the position coordinates of the transmitter and the receiver respectively; Z DS (s) is a spatially correlated normal distribution and is expressed as: Z DS (s) = μ DS + μ d log 10 (d TR ) + X DS (s){σ DS + σ d log 10 (d TR )} where d TR is the distance between the transceiver, which can be obtained from the s vector; μ DS and σ DS are the normalized distances respectively. When d TR = 1m, they are the mean and standard deviation of the Gaussian variable; μ d and σ d are the distance scaling ratio coefficients of the mean and standard deviation respectively; X DS ~N(0,1) is a standard normal distribution with six-dimensional spatial correlation. According to the reciprocity principle between transmission and reception, X DS is expressed as: where is the space-related random variable generated in step 1, and its distribution is a standard normal distribution, The spatial autocorrelation function ρ(Δd) of is modeled as: where Δd is the Euclidean distance between two points in space, and d1 is the solution correlation distance, defined as the Euclidean distance in space when the correlation coefficient is e -1 ; when Δd ≥ d1, it is considered that the spatial autocorrelation of two random variables is very low and they are approximately independent of each other. When Δd < d1, the correlation between two nearby points is described according to the proportionality coefficient k; Step 2: Initialize the central scatterer paths in multiple double-scattering clusters at the transceiver. The normalized parameters to be initialized include the delay, horizontal departure angle, vertical departure angle, horizontal arrival angle, and horizontal departure angle of the central scatterer paths. The initialization method is similar to the instantiation of the scenario parameters in Step 1, except that the distribution models of different parameters need to be changed, and then a power value is assigned to each central scatterer path. The power P' of the central scatterer path of the m-th double-scattering cluster m is assigned as: where τ m is the path delay, r τ is the delay distribution ratio factor, and DS(s) is the delay spread in step 1; Step 3: Modify the delay, angle, and power of the central sub-path according to the scene parameters in Step 1 to accurately reflect the scene parameters. The scene parameters are a set of samples with statistical characteristics obtained from measuring the actual environment, while the delay, angle, and power of the central sub-path are a set of small-scale random samples. The latter needs to be modified so that the two are consistent in statistical terms. In the power correction, the Rice factor of the scene parameters needs to be referred to. First, generate the direct-path power P1′, and then normalize the power of M - 1 central scattered sub-paths: In the angle parameter correction, taking the horizontal departure angle as an example, the horizontal departure angle spread of the scene parameters needs to be referred to. Since the angle is periodic, the initial angle mean needs to be obtained first, and then the central offset value of all angles is calculated as follows: where arg calculates the argument of a complex number, and j is the imaginary unit, is the horizontal departure angle center offset value, P m is the path power normalized in step 3, α′ m is the initial horizontal departure angle generated in step 2; calculate the second moment of the offset angle, i.e., the horizontal departure angle spread which is: Based on the above calculation results, the initial horizontal departure angle α′ m is scaled to α m , and the scaling method is as follows: where AS is the horizontal departure angle spread in Step 1, and the correction of the delay parameter is similar to that of the angle. Step 4, add the characteristics of mobility (Doppler shift) and near-field spherical waves to the channel: Add the remaining scattered sub-paths to each pair of double-scattering clusters. Since the signal frequency increases and the channel delay resolution improves, it is reasonably assumed that the closely scattered sub-paths originating from the same double-scattering cluster are approximately resolved in the delay domain and arrive from different directions within a small range. Therefore, construct the remaining scattered sub-paths based on the central scattered sub-path. Taking the departure angle at the transmitter as an example, the horizontal departure angle α m,n , the vertical departure angle β m,n are expressed as: The subscripts m and n represent the nth scatterer path in the double-scattering cluster numbered m; c α and c β are the horizontal and vertical cluster angular spreads respectively, are the offsets of the horizontal and vertical departure angles under the condition of unit angular spread respectively; the power P of the remaining scatterer paths m,n is allocated as: ρ m (α m,n ,β m,n ) is the cluster angular power spectrum, and P m is the power of the central scatterer path generated in step 3. The delay τ m,1 of the m-th central scatterer path generated in step 3 is expressed as: is the vector from the transmitting end to the nth scattering point in the mth scattering cluster of the first hop, is the vector from the first-hop scattering point to the last-hop near-field scattering point, is the vector from the last-hop near-field scattering point to the receiving end. When n = 1, it represents the central scatterer path; is the virtual link from the first-hop near-field scattering point to the last-hop scattering point, is the additional delay of the virtual link, following a unilateral exponential distribution; Calculate a set of vectors based on the delay and angle obtained in Step 3 to represent the positions of all scatterers in the double scattering cluster. Determine the position of the scatterer corresponding to the m-th central scatterer path by solving a programming problem: δ is the path loss parameter, D is the vector pointing from the transmitter to the receiver, (l 1max , l 2max ), (l 3max , l 4max ) are the distance ranges from the transmitter and the receiver to the nearest scatterer respectively, and specific values should be given according to the environmental characteristics; Taking the near-field scattering points at the transmitting end in the double scattering clusters as an example, the position vectors of the remaining scattering points are as follows: α m,n and β m,n respectively represent the horizontal and vertical departure angles of the remaining scattered sub-rays generated by step 4; When the mobile terminal moves to different channel sampling points s, update the initial scattered sub-path parameters calculated in the above steps: wherein are the motion vectors of the transceiver and the double-scattering cluster, respectively Step 5: Add path loss and shadow fading to each path. Since the channel coefficient values change with time due to the movement, Steps 4 and 5 need to be performed at each sampling point within the channel correlation distance interval, and a cosine function is used to simulate the smooth transition of the channel parameters in the transition area of the scene change. The impulse response h(s,τ) at time s of the channel is expressed as: where P 1，1 is the direct-path power, and P m,n is the power of the non-line-of-sight double-scattering cluster path. The impulse response h LoS (s,τ) is as follows: where f c represents the carrier frequency, and the impulse response of the n-th sub-path in the m-th double-scattering cluster of the non-line-of-sight path is expressed as: where represents the random phase generated by path scattering, modeled as a uniform distribution U(0, 2π); β m,n (s) is defined as the large-scale effect of shadow fading and path loss, expressed as: where a, b, c are scene characteristic coefficients, usually determined by measurement; d m,n,s represents the path length, and the shadow fading is modeled by a random variable z m,n,s which follows a lognormal distribution; Step 6: Generate the channel parameter matrix for each pair of clusters according to the results and store it in the RAM of the PC in the form of a binary file. The channel parameters are stored in the following binary data form: [A m,n,s B m,n,s τ m,n,s 3×m×n×s ​ Where A m,n,s and B m,n,s are calculated from the channel impulse response, representing the overall amplitude and phase influence of the channel at the sampling point s, i.e., the channel coefficient; then, through direct memory access (DMA), high-speed data transfer between the memories of both the processing system (PS) and the programmable logic (PL) is provided; the initialization of this transfer action is completed by the APU in the processing system platform, while the transfer action itself is continuously completed by the DMA controller. This transfer method is directly controlled by the CPU, so there is no corresponding interrupt handling. A direct data transfer channel is opened for the memory through hardware, greatly improving the data transfer efficiency of the overall system; The signal to be simulated is input into each channel through the RF interface, sampled into a digital baseband signal through a series of down-conversions, converted from a double-edge signal to a single-edge signal easy for baseband processing through IDDR, and finally sent to the baseband channel simulation module. The baseband channel simulation module is divided into a delay module, a floating-point module, an attenuation module, a frequency and phase offset module, a white noise module, an integration module, and a floating-point module. Step 7, the cache parameter FIFO is connected to the baseband channel simulation module; the quantized data τ m,n,s The different delays of the signal s(t) to be simulated are realized through a block random access memory (Block ram); then the delayed signal is input into the number system conversion module. Since the attenuation differences between the various paths are relatively large, the direct and non-line-of-sight paths result in a very large difference in A m,n,s between different paths; therefore, the s(t) and A m,n,s are converted into floating-point numbers in a suitable format, reducing the problems of overflow or loss of precision and eliminating the need to add a module for handling overflow problems; Step 8, the signals output by the number system conversion module pass through the attenuation module and the frequency offset module successively. In addition to using A m,n,s data, the attenuation module also requires an additional gain coefficient A to offset the multi-stage hardware attenuation and power amplifier caused by the input of the radio frequency signal to the baseband and the output from the baseband to the radio frequency. Then, the adjusted multi-channel signals are input into the integration unit, and the sum of each pair of cluster representative data and the noise generated by the noise module is calculated, and one signal is output; Gaussian white noise module; The generation principle is as follows: Use a linear shift feedback register LSFR to generate a set of numbers with a very large period. This set of numbers is regarded as random numbers, that is, pseudo-random numbers. Since the number of various values that appear within one period of this set of numbers with a very large period is strictly equal due to the generation principle, it is regarded as a uniformly distributed sample. Then, it is mapped into a sample of a Gaussian distribution through the Box-Muller transform, and finally, Gaussian white noise with different powers is output through the gain module. Step 9: Input one path of the output signal into the fixed-point module. The fixed-point module sums up the data of each path, converts it into an appropriate fixed-point number, and outputs it to the RF module. The high-parallel computing method is adopted, reducing the processing delay.

3. A non-stationary channel simulator based on double scattering clusters used in the method according to claim 1, characterized in that It includes: A processing system platform for dynamically generating channel coefficients of a non-stationary channel based on double scattering clusters. The processing system platform includes a user-configured channel parameter module, a channel coefficient calculation module, a DMA controller, and memory and memory control. The channel coefficient calculation module is implemented by the logic of the above Steps 1 - 5; A programmable logic platform, including a baseband channel simulation module, an RF register configuration module, and a channel parameter configuration module. The baseband channel simulation module is responsible for channel processing functions, performing a series of mathematical calculations on the baseband data received from the RF end to implement functions such as signal attenuation, phase offset, frequency offset, and adding white noise. The baseband channel simulation module is implemented by the logic of the above Steps 7 - 9; An integrated RF transceiver platform for receiving and transmitting RF signals of a communication system or device to be tested, converting the RF signals into baseband I / Q signals for processing by the FPGA programmable logic platform through processes such as amplification, mixing, and filtering, which is implemented by the logic of the above Step 6.

Images