Study On Key Techniques Of Software Process For GNSS Signal

With the development of computer technology,the computing ability of multi-core CPU has improved evidently and GPU is applied in engineering computing.Parallel computing improves as single processors develop to multi-processors.Real time GNSS signal simulator and signal receive process based on GPP(General Purpose Processor)become mature.At this time,most of GNSS software signal processing algorithms are serial algorithms,simply implanted to parallel computing circumstances,which doesn't exploit the computing ability of parallel devices fully.This dissertation did research on parallel signal generating?channel simulating ?acquisition and correlation of navigation signal.The main contents are listed below:(1)In the conventional high-dynamic DSSS signal DDS generating method,every single sample is computed depend on the former one.It couldn't be paralleled and isn't real-time.Considering this problem,one high-dynamic signal generating method based on instant phase polymerization interpolation is proposed.By introducing instant phase,which is computed by dynamic parameters with current time,the correlation of samples is removed,which enables signal parallel computing,improving the real-time ability.Experiments show that the algorithm running on GPU simulates 100 MSPS signal including signal simulation and transportation in 384 ms,the simulation speed is improved significantly compared with CPU implementation.(2)For solving the acquisition problem of long PN-code spread spectrum signal,a GPU-based partial code phase space globally direct acquisition algorithm is proposed.With FFT realizing partial correlation? code phase parallel searching,GPU accelerates FFT computation,which speeds up long PN-code acquisition.Experiments show that,using this algorithm to acquire signal with SNR higher than 42.7dBHz?time uncertainty range ±1s?frequency uncertainty range ±5kHz takes approximately 3s,the searching speed of the direct acquisition method based on GPU is improved significantly compared with CPU implementation.(3)To reduce data processing scale and memory requirement,adapting to different GPUs and being convenient for parallel task processing for multi-core processors,a GPU-based segment partially overlap long period code direct acquisition method is proposed.This method separates input signal data into segments,so GPU could process segmented data FFT in patches,speeding up long period code acquisition.(4)Regarding long period code frequency search,a segmented-correlation-phase-difference based fine frequency search method is proposed.This method utilizes two segmented correlation sequences to calculate phase differences to estimate signal doppler,and it is not related to specific windows.Simulation results show that this method could reach evidently better estimation accuracy than zero-padded DFT and window frequency feature based doppler estimation method.(5)As respect to DSSS signal correlation,vector is introduced to build a vector dot product based correlation model.Utilizing GPU to realize vector dot product leads to parallel correlation,which exploits GPU parallel computing ability.This method improves correlation real-time computing.(6)A parallel multi-correlation computing model is build based on matrix vector processing.Matrix vector product computes different-delay correlations,which improves multi-correlation computing real-time ability,realizing DSSS signal software multi-correlation real-time computing.(7)As respect to delay-doppler 2-dimension multi correlation real-time software realizing,a delay-doppler 2-level parallel computing acceleration algorithm is proposed.As respect to delay multi-correlation,GPU-based vector dot product correlation is adopted,while as respect to Doppler multi-correlation,partial correlation sequences FFT is adopted.By parallel accelerating in two dimensions,software correlation computing real-time ability is improved.(8)As respect to tens of channels correlation real-time computing in software receiver,a matrix vector product based multi-pilot-channel parallel correlation receiving model is built.This method improves multi-pilot-channel correlation computing real-time ability.Simulation results show that,25 channels with 150 correlators consumed 967?s in all for 1ms signal of 25 MHz sampling rate.Compared with the implementation on 32-thread CPU based on Math Kernel Library(MKL),the correlation computing speed of this method is improved significantly.This method could realize GNSS wide-band pilot channels software real-time receiving.All the research results have been applied in satellite device simulators or ground test systems.And the methods can also be used in navigation channel simulators?signal transmitting and receiving devices?signal quality monitoring and evaluating systems and GNSS-R devices software signal processing in real-time.