Secondary User Access for IoT Applications in the FM Radio band using FS-FBMC Kenny Barlee, University of Strathclyde (Scotland) 1/25
Overview Background + Motivation Transmitter Design Results as in paper Recent work 2/25
Presentation Key Terms Primary User (PU) = licensed radio spectrum user (e.g. cellular, TV, satellite) Secondary User (SU) = unlicensed radio spectrum user Dynamic Spectrum Access (DSA) = technique used by SUs to identify and gain access to available spectrum Software Defined Radio (SDR) = radio with a dynamic, software controlled front end that is a key enabler in DSA 3/25
Background and Motivation The radio spectrum is a finite (non exhaustive) resource High cost barriers associated with obtaining broadcast licenses Very competitive market licenses worth 50billion to the UK economy In coming years cities will be full of millions of sensors, all requiring low datarate connections Existing WiFi and free to use ISM bands are congested another solution required 4/25
[GAP] [GAP] [GAP] [GAP] [GAP] [GAP] [GAP] [GAP] [GAP] [GAP] Background and Motivation The radio spectrum is underutilized Gaps in Primary User (PU) spectrum can be used by the Secondary User (SU) with the help of Dynamic Spectrum Access (DSA) techniques Satellite TV Broadcast Cellular Aeronautical Nav Secondary Users (SUs) Primary Users (PUs) 5/25
Background and Motivation FM Band Band from 88 to 108 MHz, 100 individual 200 khz wide channels Often poorly utilized [1,2] Research shows in cities with populations around 1m, only 25% of the band is used, much less in rural areas Signals broadcast at these freqs have excellent propagation characteristics Able to diffract around objects such as hills and human-made structures, and can penetrate through buildings well Band is an excellent candidate for smart city IoT communication, e.g. traffic signal sequencing, smart street lighting etc. 6/25
Research Aims Design a SU radio capable of filling gaps in the FM Band for low data throughput applications (IoT) Radio to identify available channels by itself (then build channel mask, complete with guardbands) Radio to use an adaptive modulation scheme Radio must cause minimal interference to FM Station PUs (IMPORTANT!) 7/25
DSA Radio Design Modulation Schemes To mould around PUs, radio requires adaptive, Non Contiguous (NC) modulation scheme (i.e. can make signals with spectral holes) Out Of Band (OOB) leakage (power in disabled channels) must be minimal in order to protect PU + meet regulator interference rules Favourite NC schemes in literature are: [3,4] NC-OFDM (normal OFDM, with zeros transmitted on disabled subcarriers) FBMC (filterbank multicarrier, with zeros transmitted in disabled subchannels) Particular FBMC filter designed for DSA = PHYDYAS filter [5] 8/25
DSA Radio Design Modulation Schemes NC-OFDM uses rectangular pulse shaping, hence has high OOB leakage [6] FBMC uses specially designed pulse shaping filters, which minimize OOB leakage FBMC is the more attractive candidate due to spectral containment 9/25
DSA Radio Design Transmitter 2 possible FBMC transmitter architectures Frequency Spread (FS-FBMC) or Polyphase Network (PPN-FBMC) PHYDYAS FBMC requires Offset QAM symbols (FBMC/OQAM) FS-FBMC: symbols upsampled by K, filtered, input to IFFT, overlap/sum [7] 10/25
DSA Radio Design Transmitter Design is HDL-ready (samples rather than frames, valid lines, fixed point) FBMC parameters f S =20.48MHz, K =4, M =1024 x1024 40kHz wide overlapping channels created (10 per FM channel) 11/25
DSA Radio Design AutoMask Research has shown that the Matched Detector is the most reliable sensing technique (i.e. the official receiver for the type of signal being sensed, rather than generic Energy Detection) [8] Method adopted was to tune to each FM centre freq, FM demodulate, perform channel classification, then store the results in RAM Guard bands are added based on regulator minimum distance rules 12/25
DSA Radio Design AutoMask The AutoMask module was designed HDL-ready, and is optimised for FPGA targeting (e.g. with pipelining and polyphase serialized decimation filters) Mask creation is a function of a detection window Only takes 0.64 seconds to generate with a detection window of 2048 samples per FM channel 13/25
Testing SU Interference Levels Recordings of the FM Radio spectrum were obtained using a USRP B210 with a standard VHF/UHF omnidirectional antenna In central Glasgow, 22 stations were found The USRP was uncalibrated; hence samples received by the computer had relative power levels Channel model estimations were used to make informed adjustments Friis free space model Perez-Vega Zamanillo/ FCC F(50,50) [9] 14/25
Testing SU Interference Levels The FM Band recording was passed through the AutoMask module, and a mask was generated 275 OQAM subchannels (of 1024) were eligible for use (a function of the chosen guardband size), a total bandwidth of 5.5MHz SU signals with various transmit powers were generated using the PHYDYAS FBMC PHY and an equivalent NC-OFDM PHY These SU signals were overlaid on the FM Band recording, to simulate the RF transmission, and the interference they would cause to the PU 15/25
Testing SU Interference Levels Plotting the power spectra of the signals, clear to see that there is minimal OOB leakage with PHYDYAS FBMC when compared to NC-OFDM Each of the known PU FM stations were demodulated in turn, to allow the interference caused by the SU to be explored 16/25
Testing SU Interference Levels Quantitative Signal to Interference Ratio (SIR) was found for each SU Tx power N P FM = 1 N n 1 s FM n 2 N P SU = 1 N n 1 s FM+SU n 2 P FM SIR = 10 log 10 P FM P SU PHYDYAS radio shows 47dB improvement in PU SIR over NC-OFDM At 4W, PHYDYAS leakage x88 LOWER than PU power At 4W, NC-OFDM leakage x625 HIGHER than PU power 17/25
Testing SU Interference Levels Qualitative Audio listening tests were then performed to classify how bad the quality of each station was Each station, at each transmit power, for both PHYDYAS FBMC and NC-OFDM were evaluated Mean Opinion Score was used: MOS 1 = noise MOS 5 = perfect audio 18/25
DSA Radio Design ZynqSDR Tx Implementation Next step was to target the Transmitter + AutoMask to radio hardware Rapid prototyping made easy from Simulink with the Zynq Based Radio support package for ZynqSDR [10] 19/25
DSA Radio Design ZynqSDR Tx Implementation 20/25
DSA Radio Design Next Steps Perform tests in the University s RF shielded anechoic chamber to investigate how the SU interferes with standard FM Radio receivers Find an optimal guard band size (tradeoff between interference and data throughput) 21/25
DSA Radio Design Receiver A receiver PHY has been developed, which is able to infer unknown Tx masks Initial results show that transmitted data can be recovered correctly Fun quirk with the system the OQAM constellation contains 9 clusters of points! Received IQ (FBMC + FM Radio) After FFT After PHYDYAS Filter After Max Effect 22/25
Conclusions The spectrum (a finite resource) is often poorly utilized In coming years there will be sensors everywhere, requiring low datarate connections While the proposed DSA radio PHY is not mmwave, it is equally as valid an access technique for 5G communications The idea of DSA is gaining ground in 5G research (e.g. 5G Rural First project), and it is accepted that it will play a crucial role in enabling access to the radio spectrum for next gen communications 23/25
Conclusions The novel DSA radio PHY developed can enable SU access in the band traditionally used for FM Radio Initial tests suggest the radio can coexist with the PU, causing very little interference The PHYDYAS FBMC radio provides a 47dB improvement in PU interference over NC-OFDM The radio s smart abilities have been demonstrated, in that it can generate its own channel mask within 0.64 seconds of turn on 24/25
References [1] D. Otermat, C. Otero, I. Kostanic, Analysis of the FM Radio Spectrum for Internet of Things Opportunistic Access Via Cognitive Radio, in Proc. of WF-IoT 15, Milan, IT, pp. 166-171, Dec 2015 [2] D. Otermat, C. Otero, I. Kostanic, Analysis of the FM Radio Spectrum for Secondary Licensing of Low-Power Short-Range Cognitive Internet of Things Devices, in IEEE Access, Oct 2016 [3] B. Farhang-Boroujeny, R. Kempter, Multicarrier communication techniques for spectrum sensing and communication in cognitive radio, in IEEE Commun. Mag, vol. 46 no. 4, pp. 80-85, Apr 2008 [4] R. Gerzaguet et al., The 5G candidate waveform race: a comparison of complexity and performance, EURASIP Journal on Wireless Commun. and Networking, Jan 2017 [5] M. Bellanger. (2010, Jun). FBMC physical layer: a primer, PHYDYAS. [Online] Available: http://www.ict-phydyas.org/teamspace/internal-folder/fbmc-primer_06-2010.pdf [6] B. Farhang-Boroujeny, OFDM Versus Filter Bank Multicarrier, in IEEE Signal Process. Mag., vol. 28 no. 3, pp. 92-112, May 2011 [7] M. Bellanger, FS-FBMC: a flexible robust scheme for efficient multicarrier broadband wireless access, IEEE Globecom Workshops, Anaheim, USA, Dec 2012 [8] M. Hoyhtya, Spectrum Occupancy Measurements: A Survey and Use of Interference Maps, IEEE Communications Surveys and Tutorials, vol. 18 no. 4, pp. 2386-2414, April 2016 [9] C. Perez-Vega and J. Zamanillo. (2002, Jun). Path Loss Model for Broadcasting Applications and Outdoor Communications Systems in the VHF and UHF Bands. [Online]. Available: http://personales.unican.es/perezvr/pdf/fcc%20model02.pdf [10] MathWorks. (2018). Zynq SDR Support from Communications System Toolbox. [Online]. Available: https://uk.mathworks.com/hardware-support/zynq-sdr.html 25/25
Research co-funded by the MathWorks DCRG Grant on Dynamic Spectrum Access for 5G Communications 2016-2018 26/25