Propagation Laboratory

Transcription

1 RUSK-Fujitsu Double-Directional Full Polarization Channel Sounder

2 Clustering Results of Estimated MIMO Double-directional Channel Data of a Small Macrocell (Lawrence Materum)

3

4 Introduction of Complex Radar Cross Section for Ray Tracing Simulation of Microcell Propagation Channels (Yukiko Kishiki) Introduction Propagation path loss at long distances decreases on the street model in conventional ray tracing simulator. The energy of specular reflection & diffraction are multiply counted because there are a lot of diffraction edges near the specular reflection. In this study Step.1) We survey the range where the geometrical optics can be applicable. Step.2)We introduce the complex radar cross section within the range that geometrical optics fail. Step.3)We integrate complex RCS to ray tracing. Distance between Tx & Rx: 10m 1km Tx Rx Rx Rx 27m Applicable Range Ray Tracing Step1 Step3 Step.1 Comparison between physical optics (PO) & stationary Model Scattering Specular Reflection & Diffraction Radar Cross Section Step2 phase method (SP) 1000m Tx Rx (*SP..asymptotic evaluation of PO to extract specular reflection and edge diffraction separately) Step.2 The radar cross section (RCS) is introduced for φi s 1 s 2 Conductor Xe 5m (a) The true diffraction: the range that GO fails. In this study, the complex RCS is introduced in order to calculate the scattering from multiple objects. PO SP's reflection (b) The diffraction of geometrics optics: SP's diffraction The complex RCS is calculated for 3-D rectangular plate with arbitrary surface impedance. Reflection and diffraction cannot be distinct within 1 st Fresnel zone. Step.3 Result Application of complex RCS to ray tracing calculation. Implemented scattering mechanisms : Specular reflection - Transmission - Diffraction - RCS scattering Conventional technique: the received level increases at long distances. Proposed technique: the received level decreases at long distances and is close to the measurement data. The RCS scattering paths are identified by these steps. 1) The 1 st Fresnel zone is made between the Tx and mirror point of Rx. 2) The diffraction and the specular reflection paths within the 1 st Fresnel zone are deleted. 3) Within the 1 st Fresnel zone, scatterings from the surfaces are coherently summed up. (similar to street model)

5 Spectrum Sensing based on Cyclostationarity for Cognitive Radio (Po Kimtho) Introduction Spectrum Sensing is one of the m ost im portant technique for Cognitive Radio (CR) in order to recognize the presence or absence of the prim ary users. In this poster, non-cooperative and cooperative spectrum sensing based on cyclostationarity are presented. Cyclostationary Detector The estim ation ofcyclic autocorrelation function (CAF)ofreceived signal where prim ary signal Gaussian noise To detect whether or not a prim ary signal is present,a row vector is com puted In order to test for the presence of signal at any of the cyclic frequencies of interest the hypothesis testing problem is form ulated as follows signalis absent signalis present where is an estim ation error.for large,it is known that is Gaussian distribution with probability function where is the estim ated covariance m atrix. Based on generalized likelihood ratio test (GLRT), the statistical cyclic test can be defined as follow Under : converges to a central Chi-square distribution Under : converges to Gaussian distribution with pdf For a given false alarm rate, the threshold can be determ ined. Then, the detection can be perform ed as follows signalis absent signalis present Non Cooperative Sensing vs. Cooperative Sensing Non cooperative sensing:each cognitive user com putes the statistical test at single or m ultiple cyclic frequencies. Cooperative sensing:m ultiple cognitive users share the statisticaltest with one another.

6 Cognitive Radio Testbed System and Demonstration for Spectrum Sensing (Sabita Maharjan and Minseok Kim) Multi-purpose Signal Processing Unit Introduction The fundamental principle of cognitive radio systems is the concept of spectrum sharing in order to dynamically utilize the precious spectrum. Spectrum sensing is needed to ensure that cognitive radios would not interfere with primary users. In this study, the testbed system to study the implementation of effective spectrum sensing techniques for cognitive radios for ISDB-T signal in Japan is presented. The digital signal processing unit was developed to integrate and evaluate multi-antenna signal processors. It consists of the signal processing board and controller board. The signal processing board incorporates 16 channels of ADCs and DACs, and high density FPGAs for signal processing. A general-purpose CPU is used as a system control unit and numerical computation coprocessor with floating-point arithmetic. 16-Channel ADC/DAC/FPGA board External I/O 16CH DACs SMB Objectives FPGAs To evaluate the algorithm and the system for weak signal sensing in real hardware. To comprehensively study the practical problems and hardware limitations from the experiment data in the testbed system 16CH ADCs Controller board (SH4 / NetBSD) Testbed System Configurations Ant. RF Filter Low Noise Amplifier Mixer IF/BB Filter AGC LNA Multiple CPEs VCO MAX3580 Ant. RF Filter Analog-toDigital Converter AD Automatic Gain Control Low Noise Amplifier Mixer AGC VCO MAX3580 To CPU Analog-toDigital Converter IF/BB Filter LNA FPGA Experimental setup DVB-T ISDB-T Direct Conversion Tuner AD Automatic Gain Control VHF/UHF DVB-T ISDB-T Direct Conversion Tuner (MAX3580) VHF-III 170MHz to 230MHz, UHF 470MHz to 878MHz Antenna Baseband Output (I) Baseband Output (Q) DSUB25 Parallel Connector MAX Channel ADC/DAC/FPGA board PC for analysis MALAB based Realtime Visualization VHF Input UHF Input Specifications of the multi-purpose signal processing unit ADC DAC AD9245 (Analog Device) 16 Resolution 14 bits Sampling rates 80 MSPS (Interleave sampling up to 320MSPS) Memory SDRAM 256 MB DAC904 ( Burrbrown) Channels 16 Resolution 14 bits Sampling 165 MSPS Memory SDRAM 64 MB STRATIX EP1S40 (ALTERA) FPGAs CPU Channels - 41,250 LEs (about 1000,000 Typical Gates) - 3,423,744 bits internal memory - 14 DSP blocks (14 parallel 36*36 multiplier available) Renesas SH4 200 MHz, 360 MIPS 1.4 GFLOPS OS NetBSD Power Spectrum Density Spectrum Correlation Density