Design Of Digital Demodulation Software Based On CPU And GPU Heterogeneous Parallel S

With the development of digital communication systems,more attention has been paid to the demodulation analysis of vector signals.Compared with the hardware platform,the processing of vector digital signals through software avoids a series of problems such as high hardware development cost,long development cycle and poor platform versatility,and can also give full play to the advantages of software programming,such as convenient and fast,high degree of freedom and strong portability.As a traditional computing unit in a computer system,CPU has outstanding ability in controlling flow,but its computing power is relatively weak? The concurrent threads used by the GPU can process a large amount of parallel data,which has great advantages in single-and double-precision floating-point computing capabilities.Therefore,combining the characteristics of the two,this thesis adopts the CPU + GPU heterogeneous parallel architecture system to design the demodulation software of vector signals,and the demodulation algorithm mainly includes three main algorithm modules: timing synchronization,channel equalization and carrier recovery.The timing synchronization module introduces and analyzes the more commonly used timing error detection methods,determines the amplitude flat method that is not limited by the modulation format to complete the timing error estimation,and uses the fractional interpolation filtering method to compensate the timing error.Among them,parallel operations such as amplitude squared and decimal interpolation filtering are implemented by the GPU,and the CPU is responsible for controlling the decimal interpolation filtering parameters and the execution order of the timing module.The channel equalization module discusses the difference between adaptive equalization and blind equalization,compares and analyzes various equalization algorithms,and designs heterogeneous blind equalization software processing schemes for different modulation formats based on the idea of adaptive equalization.The scheme uses the CPU to calculate the equilibrium coefficient,and the GPU relies on the stable equilibrium coefficient to complete the parallel equalization correction,and combines the EVM parameters to determine whether the equilibrium coefficient needs to be updated,so as to realize the blind equalization software implementation of CPU + GPU heterogeneity.The carrier synchronization module introduces three carrier error judgment methods:direct judgment method,simplified constellation method and polarity judgment method,and designs carrier recovery algorithms suitable for multiple modulation formats based on digital phase-locked loop structure.The feasibility of parallel computing of carrier synchronization algorithm is analyzed and verified,and the heterogeneous computing deployment of CPU+GPU is realized,that is,the CPU is responsible for data distribution,and the GPU is responsible for realizing the structure of phase recognition and loop filter in the loop.This thesis introduces the test environment and test methods used in digital communication demodulation software,and focuses on comparing the acceleration effect of CPU+GPU heterogeneous platform compared with CPU serial computing on each module.And the software modules and overall performance tested,timing synchronization,channel equalization and carrier recovery module have good effects,when the signal through the digital demodulator,the EVM of each modulation format signal is less than10%,to meet the requirements of the index.