Identifying and Quantifying Spectrum Opportunities for 5G 3 May Dr. Melvin Ferreira

Transcription

1 Identifying and Quantifying Spectrum Opportunities for 5G 3 May 2017 Dr. Melvin Ferreira melvin.ferreira@nwu.ac.za

2 Outline Our view Work on TVWS spectrum availability Utility of TVWS for Broadband Wireless Access (BWA) Extending the geolocation database Automatic Modulation Classification

3 Our view Spectrum resources are limited Internationally trust in geolocation database approach FCC OFCOM Local momentum UHF primarily assigned for TV broadcasting ICASA currently tabled draft regulations for secondary access MHz proposed for secondary usage Current approaches will need to evolve with technology (5G) and adapt to challenges posed by temporal behaviour of users Improve database (adaptability) Augment with measurement based approaches Implies detection and identification of users in spectrum

4 Work on TVWS spectrum availability Determine amount of UHF TV white space available in SA MHz (48 Channels) Express as a number of channels in UHF bands Did it for 3 time periods: Analogue only / Dual illumination / Digital only A snapshot of WS channels available given input data & modelling assumptions Upper bound given policy ruleset, secondary use case & technical parameters of technology used, will determine ultimate utility

5 Work on TVWS spectrum availability Protection of the PU: Use ITU-R established guidelines Analogue: ITU-R BT.417 Digital: ITU-R BT.1368 Mobile DTT Fixed DTT

6 Work on TVWS spectrum availability TV transmitter database (Gazette) ERP Mast height AGL Services transmitted Input data for entire process Other protected services : SKA Topography data: SRTM v3, void filled ITU-R P.1546 propagation model

7 UHF Channels available: Dual illumination: co-channel rule

8 UHF Channels available: Post Digital migration: Adjacent channel rule

9 Utility of TVWS for Broadband Wireless Access (BWA) Quantify utility of TVWS for BWA Do this by planning a secondary network in TVWS using FCC policy ruleset Follow a modified cellular network planning approach (not all channels are available everywhere in TVWS) Determine the capacity available to population in the cell Interference from other secondary cells considered 36dBm TX power and 30m masts considered rural broadband use case Capacity increases possible with other configurations

10 Utility of TVWS for Broadband Wireless Access (BWA)

11 Utility of TVWS for Broadband Wireless Access (BWA)

12 Utility of TVWS for Broadband Wireless Access (BWA)

13 Utility of TVWS for Broadband Wireless Access (BWA)

14 Utility of TVWS for Broadband Wireless Access (BWA)

15 Utility of TVWS for Broadband Wireless Access (BWA)

16 Extending the geolocation database Computing is a big concern White space recalculations done every Week (FCC) 6 Months (Ofcom) Fine grained computations are involved Especially true for models using terrain data Prediction of field strengths present a performance bottleneck Computation options: GPGPU Programming Cloud Computing High Performance Computing Cluster

17 Extending the geolocation database Improve propagation prediction times Use the CSIR CHPC as computation node In-house MPI implementations of the following propagation models: Free space loss HATA Davidson Longley Rice ITM Area Longley Rice ITM P2P Ran tests to measure the speedup for calculating the propagation loss for the above propagation models on the CHPC cluster: Consider entire TV transmitter database s influence on a 30 grid of Gauteng Compute speed-up compared to baseline implementation on quad-core machine Increasing number of cores

18 Extending the geolocation database Speed-up results on the CHPC

19 Extending the geolocation database Results: TAU analysis of ITM P2P implementation to find bottleneck Time Exclusive msec Inclusive total msec #Calls Name 100, :15,784 1.TAU application 100,0 1,392 35:15,754 1 main() 85,7 30,684 30:13, get_globe_pfl() 84,3 11:12:851 29:43,234 4,3776E+07 GLOBE_elevation() 52,4 18:29,601 18:29,601 1,0579E+08 Get_GLOBE_data() 13,9 4:54,596 4:54,596 3 MPI_Barrier()

20 Extending the geolocation database Results before and after path extraction improvements Before After Transmitters CPUs Wall Time (hh:mm:ss) CPU Time (hh:mm:ss) 05:26:29 00:04: :06:02 17:17:00 Implications for recalculation time in geolocation database MPI implementation also useful for general field strength prediction

21 Automatic Modulation Classification (AMC) Identification of under-utilised spectrum Energy detection does not provide enough metadata about emission Identifying modulation schemes of a given communication system with a high probability of success. Determine the appropriate frequency band, and select the most common digital modulations present in that band. Model created to simulate the modulations in different communication channels (AWGN, Rayleigh, Fading) at different SNR's. Extract and evaluate features (Statistical not computationally complex), as opposed to likelihood-based techniques.

22 AMC System Model x[n] -> Input bit stream y[n] -> Modulated and transmitted signal z[n] -> Received signal corrupted by communication channel (AWGN/Rician/Rayleigh)

23 AMC Feature Evaluation Features considered (instantaneous amplitude & phase) Mean ( ) Variance & Higher-order moments ( 2,m ) Skewness (s) Kurtosis ( ) Fourth-order cumulants (Cum) Signal power key ( ) Final features selected 2 of the instantaneous phase of the instantaneous phase 12 of the normalised instantaneous amplitude Signal power key C 40 /C 42

24 AMC Classifier Performance

25 Work in progress Expanding policy ruleset to include additional analyses of other rules Further improving computational tractability of RF propagation modelling Support for additional propagation models Point verification of geolocation database predictions via ground truth spectrum measurements at key sites Automatic modulation classification Additional modulations & techniques Validation on practical measured emissions

26 Thank you Prof. Albert Helberg Dr. Melvin Ferreira