Distributed Spectrum Occupancy Measurements in the MHz Band for LV PLC Networks

Transcription

1 Distributed Spectrum Occupancy Measurements in the MHz Band for LV PLC Networks 9th Workshop on Power Line Communications September 22, 2015 Prof. Dr.-Ing. habil. Klaus Dostert KIT University of the State of Baden-Wuerttemberg and National Laboratory of the Helmholtz Association

2 Motivation (Usable) spectrum is a limited ressource Idea: Dynamic Access Problem: Avoid interference Source: brunel.ac.uk Challenges Co-existence with wireless systems Primary User (PU) Detection Detect low noise frequency sections Determine spectrum occupancy (noise & PUs) = Dynamic Spectrum Access for Cognitive PLC systems 2/21

3 Service Survey civil and military Civil services Broadcasting services Military services Combined services (Various noise sources) broadcasting services f / MHz = Focus on broadcasting services 3/21

4 Outline 1. Motivation 2. Spectrum Estimation Method 3. Measurement Setup Automated Measurement System Campaign Description 4. Analysis Location Diversity Time Diversity Time-Frequency Analysis Wired vs. Wireless 5. Summary 4/21

5 Outline 1. Motivation 2. Spectrum Estimation Method 3. Measurement Setup Automated Measurement System Campaign Description 4. Analysis Location Diversity Time Diversity Time-Frequency Analysis Wired vs. Wireless 5. Summary 5/21

6 Spectrum Estimation Method Energy Detection (ED) as promising approach for PLC systems Oberservations of the Power Spectral Density (PSD) Welch s method Averaged periodograms Overlapped blocks of samples Consistent estimation Analysis of stationary random processes Short-time analysis based on the Short-Time Fourier Transform (STFT) Basic cyclostationary analysis = PSD as rough estimation of occupied spectrum (PUs & noise) 6/21

7 Spectrum Estimation Parameters PSD parameters STFT parameters Parameter Value Sample rate 25 MS/s (real-valued) Number of samples 2,500,000 (5 mains periods) FFT size 4096 ( f 6 khz) Window Hanning Overlap 50 % Parameter Value Sample rate 25 MS/s (real-valued) Number of samples 1,000,000 (2 mains periods) DFT size 5000 ( f = 5 khz) Window Hanning Overlap 50 % t 200 µs 7/21

8 Outline 1. Motivation 2. Spectrum Estimation Method 3. Measurement Setup Automated Measurement System Campaign Description 4. Analysis Location Diversity Time Diversity Time-Frequency Analysis Wired vs. Wireless 5. Summary 8/21

9 Automated Measurement System (AMS) Live-USB Image Based on SSH (W-)LAN, 3G/4G, VPN Central repository Central control unit 2 time modes (GPS, system time) Modular design (GNU Radio flowgraphs) Uniform, automated measurement procedure Systematic planning, offline design Reduced error probability Consistent filing of measurement parameters, results and raw data 9/21

10 Campaign Description Requirements Multiple locations Synchronous measurements Central control and monitoring Uniform data structure Wired/Wireless measurements Source: google.com Campaign details All measurements taken at June 11th & 12th in Karlsruhe, Germany 4 Locations: Transformer station, garage, cellar & apartment building 24 hours observation: Every 15 minutes 5 seconds of raw data at 25 MS/s Additionally 5 minutes of raw data every 6 hours at each location 10/21

11 Outline 1. Motivation 2. Spectrum Estimation Method 3. Measurement Setup Automated Measurement System Campaign Description 4. Analysis Location Diversity Time Diversity Time-Frequency Analysis Wired vs. Wireless 5. Summary 11/21

12 Location Diversity 80 apartment building garage cellar transformer station Scaled PSD [V 2 /Hz] Hz 2.0 MHz 4.0 MHz 6.0 MHz 8.0 MHz 10.0 MHz Frequency Relative frequency [%] Scaled PSD [V 2 /Hz] garage apartment building cellar transformer station 12/21

13 Location Diversity apartment building cellar transformer station Scaled PSD [V 2 /Hz] Hz khz khz khz khz 1.0 MHz Frequency = Common features, but highly location-dependent 13/21

14 Time Diversity Garage, Cellar PSD Histogram Garage Cellar 14/21

15 Time Diversity Transformer, Apartment PSD Histogram Transformer station Apartment building 15/21

16 STFT Analysis Garage, Transformer Garage 10.0 MHz 10.0 MHz MHz 105 Frequency 6.0 MHz MHz MHz Frequency 120 Scaled PSD [dbv2 /Hz] MHz MHz Scaled PSD [dbv2 /Hz] 8.0 MHz 2.0 MHz Hz 0.0Hz 5.0 ms 10.0 ms 15.0 ms 20.0 ms 25.0 ms 30.0 ms 35.0 ms 5.0 ms 10.0 ms 15.0 ms Time 20.0 ms 25.0 ms 30.0 ms 35.0 ms Time Transformer station 10.0 MHz MHz MHz MHz MHz MHz Frequency Frequency Scaled PSD [dbv2 /Hz] MHz MHz MHz Hz Hz 5.0 ms 10.0 ms 15.0 ms 20.0 ms 25.0 ms 30.0 ms 35.0 ms 5.0 ms 10.0 ms Time 06/12 03:15 16/21 Scaled PSD [dbv2 /Hz] 10.0 MHz 15.0 ms 20.0 ms 25.0 ms 30.0 ms 35.0 ms Time 06/12 15:15

17 Wired vs. Wireless PSD 100 Power Line Air 110 Scaled PSD [V 2 /Hz] Hz 2.0 MHz 4.0 MHz 6.0 MHz 8.0 MHz 10.0 MHz Frequency Relative frequency [%] Scaled PSD [V 2 /Hz] Power Line Air 17/21

18 Wired vs. Wireless Time Diversity STFT 10.0 MHz 10.0 MHz MHz 8.0 MHz MHz MHz MHz Hz 5.0 ms 10.0 ms 15.0 ms 20.0 ms 25.0 ms 30.0 ms 35.0 ms Scaled PSD [dbv2 /Hz] 4.0 MHz MHz Frequency Frequency Scaled PSD [dbv2 /Hz] MHz Hz 5.0 ms Time 10.0 ms 15.0 ms 20.0 ms 25.0 ms 30.0 ms 35.0 ms Time 06/12 03:15 06/12 15:15 PSD & Histogram 18/21

19 Outline 1. Motivation 2. Spectrum Estimation Method 3. Measurement Setup Automated Measurement System Campaign Description 4. Analysis Location Diversity Time Diversity Time-Frequency Analysis Wired vs. Wireless 5. Summary 19/21

20 Summary & Outlook Basic Observations towards spectrum sensing & occupancy estimation Novel, Software-Defined Radio based Automated Measurement System Spectrum occupancy is highly dependent on the location Distributed, cooperative spectrum sensing Spectrum occupancy is highly dependent on the (day) time Application of machine learning algorithms (Centralized) Wireless measurements to enhance PU detection accuracy Analysis still work in progress More sophisticated detection algorithms Dedicated detection of Primary User signals Occupancy evaluation divided into noise/pus Define (frequency-dependent) occupancy levels/spectrum masks Duty cycle models Distributed, adaptive sensing for DSA/Cognitve PLC systems 20/21

21 Questions? Thank you for your attention! 21/21