The Optimized Method and Algorithms in the CPU&GPU-Based GNSS Software Receiver
Wu Cailun,Qian Yi,Cui Xiaowei,Lu Mingquan
2009-01-01
Abstract:Over the past 10 years, several researchers have concentrated on the design of GNSS receivers whose real-time correlators, tracking loop, and navigation solution are all implementated by software on PC. GNSS Software receiver require few hardware and offer greater flexibility for algorithm updating, specially in the era of Galileo and Compass satellite navigation design phase.This paper presents the design of CPU- and GPU-based correlators. Some new architectures and methods are used in the correlator design. Many optimal methods about SIMD and GPU technology are described from the view point of complexity reduction and increasing the receiver running speed.The GNSS software is made up of the antenna, the RF front end, the ICS554 PMC, the PC. The RF mixes the incoming signal to 46.42MHz in L1 frequency and then samples at 100MHz. The ICS554 down convert the signals to the baseband signals and decimate the signals to 5MHz, which is a four channel, 14-bit A/D module in a PCI Mezzanine Card format which sample rate is 100MHz. The ICS554 card includes a user programmable FPGA which convert the IF signals to I and Q signals in the baseband signals. DELL Precision 690 which the CPU runs 3.2GHz, and memory is 4GB is chosen as the workshop. The signal samples are transferred to memory with PCI bus from the ICS554.The GNSS software receiver can complete signal acquisition, tracking and Ephemeris demodulation and positioning, using GPS L1, L2, L5, and Galileo E1, E5 signals. The receiver which works on the sample rate is 5M run up to 150 channels in real-time mode. The CPU-and GPU-based correlators are described in this paper. Some performance tests and results are given in this paper. At last, some conclusion and future tendency in software receiver are present in it.