A Harmful Interference Model for White Space Radios Timothy X Brown Interdisciplinary Telecommunications Program Dept. of Electrical, Energy, and Computer Engineering University of Colorado at Boulder Visiting Professor, Carnegie Mellon University, Rwanda Presented at the ICTP School on Applications of Open Spectrum and White Spaces Technologies March 11, 2014 The Idea The spectrum is fully allocated. Urban measurements: > 75% never used > 90% unused on average Rural areas - even more Amplidue (dbm) Heavy Use Sparse Use Maximum Amplitudes Medium Use Heavy Use Cognitive Radio: Avoid Licensed users Communicate in white spaces Frequency (MHz) 1
Cognitive vs. Traditional Radios User Interaction Via Operating System Policy Input Cognitive Engine Geo locator Sensor Radio A CR does more than a traditional radio Cognitive Radio Review: Channel Avoidance Strategies Locate Unlicensed Device, check Database Detect Licensed Devices Licensed Device Beacons 4 2
Locate Licensed Device, Database Positioning can be sloppy As long as can bound error 100km OK in rural areas Potential conflict: Can not use L A U RE R I L A No conflict: Can use L B Proactive Database Licensed Users Database Unlicensed Users L B Internet? 5 Detect Licensed Devices Hidden Terminal Problem Licensed TX Hill Unlicensed TX Unlicensed RX Licensed RX Networked Detection Much more reliable Prob channel detected 1 0.8 0.6 0.4 0.2 D=0.9 D=0.5 D=0.1 Receiver Detection Interference is receiver phenomena Requires help from the receiver 0 0 5 10 15 20 25 Average Number of Neighbors 6 3
Licensed Device Beacons Per transmitter/receiver is laborious Area Beacons: Combine best of Database and Beacons Licensed Users Database Unlicensed Users 7 Avoidance Summary Method Appropriate Accurately: Requires: Cost Burden on: for: Avoids Exploits Constant High Low Lic. Lic. Positioning Standard BW Unlic. Interference Whitespace Monitoring Power Power Trans. Rec. Database X X X X X X Detecting Transmitters X X X X Transmitter Beacons X X X X X X X Area Beacons X X X X X X X X Unlicensed Signaling X X X X X X Receiver Beacons X X X X X X X X 8 4
The Problem Technically functional radio Detects primary users Correctly picks channel You can t turn just it on! The real world challenges of CR Licensed Service Providers These rules will degrade my service! Any unlicensed use can interfere FCC 04-186 NPRM Unlicensed Operation in the TV Broadcast Bands Regulator This will create more problems then it will solve CR Operators What are the risks with this technology CR Device Manufacturers How do we test our equipment? Should I invest R&D in this market? NSF, Brown, et al. (2004) Spectrum Management for Spectrum Plenty 5
A Harmful Interference Framework: The Idea Define interference up front: How it is measured By who What levels constitute harmful What to do if it is harmful Define so that: Licensed users have sufficient protection Unlicensed users/manufacturers have assurances Regulators solve problem once a priori 11 Measurement Scenario Evaluation Licensed Receiver Model CR Radios Remedy Cognitive Radio Transmitter Model 12 6
Unlicensed Transmitter Model Unbounded Deployment Bounded Deployment Aggregate Single Transmitter Per Device Per Usage 13 Licensed Receiver Model Expected Interference Widespread Extended Interference Aggregate Extended Interference Widespread Interference Single Receiver Observed Interference Conceivable Interference In practice More Detail on this Later 14 7
Evaluation Who evaluates Licensee Licensed Receiver User Unlicensed Device Manufacturer Unlicensed User Regulator How Evaluates What conditions What parameters 15 Remedy If harmful interference is found: Turnoff offending transmitter(s) Change Unlicensed Rules Operation parameters (Software Defined Radios) Definition of harmful interference Allowed/Prohibited Usages Change Licensed Rules Remedy should be incorporated into rules 16 8
Example 1: Simplest Model Licensee Evaluates Observed Interference Per Device Turnoff Device 17 Example 2: Broadcast TV Monitor Stations Expected Interference Unbounded Deployment Modify Rules 18 9
Example 2: Broadcast TV Monitor Stations Monitor TV outages at stations in coverage area TV Cable Line Case Signal Signal Power Normal Good Good Good Interference Outage Bad Good Good Cable Outage Good Bad Good Broadcast Outage Bad Bad Good Power Outage X X Bad 19 Broadcasters: These rules will degrade my service! What does TV suffer already? DBS is 99.8% extremely reliable TV is 90% at grade B contour Utility power is 99.9% Already many Wireless Microphones Can we achieve an expected 1 in 10,000 TV outage in an unlimited unlicensed deployment? Brown, An Analysis of Licensed Channel Avoidance Strategies for Unlicensed Devices, IEEE DySPAN 2005 10
Radio Interference Model F = r 2 min PCEGUL GLMNUL / A F = fraction of licensed devices suffering outage on average r min = worst case min separation between unlicensed and licensed devices P = power control factor (< 1) C = channel avoidance factor (< 1) G = antenna gain factors (< 1) M = model constants (~ 3) N UL = number of unlicensed devices A = area of system Sets Cognitive Requirement Helps Regulator Brown, A Harmful Interference Model for Unlicensed Device Operation in Licensed Service Bands, J. of Communication, 2006 Channel Avoidance Reliability Set 1000 unlicensed devices/km 2 For < 1 in 10,000 TV interference require Low Power CR: 90% reliable High Power CR: 99.99% reliable These are reasonable. But not to the broadcasters! Brown, Enabling and Evaluating Unlicensed Operation in the TV Broadcast Bands, reply comments to FCC Docket No. 04-186, 2005. Kolodzy et al., Technical Reply Comments, to Docket No. 04-18, 2005. Brown, A Model for Analyzing Unlicensed Device Operation in Licensed Broadcast Service Bands, reply to NTIA Docket No. 06051 21 29-61 29-01, 2006. 11
Harmful Interference Model Broadcasters: Conceivable Interference x w Widespread Graph % Users % Time Big Gap Users Excessive Graph Expected Interference Time x e Harmful Interference Taxonomy Expected Interference Widespread Excessive Excessive Widespread Aggregate Excessive Interference Widespread Interference Single Receiver Observed Interference Conceivable Interference In practice Provides a common framework for HI Berger, et al., IEEE Recommended Practice for the Analysis of In-Band and Adjacent Band Interference and Coexistence Between Radio Systems, IEEE Standard, P1900.2-2008. 12
Example 3: Wireless Microphones In TV bands Hard-to-detect 500,000 sold Cognitive Radios will harm them! 25 Example 3: Wireless Microphones No database exists No one is going to invest in beacons Detection is the only strategy What can the CR do? What can the Licensed devices do? 26 13
Single Interferer Model: Traditional Radio Interference WM TR WM range, R 0 27 Single Interferer Model: Traditional Radio Interference Probability of Interference (%) Traditional R 0 = 80m Interferer Distance (m) 28 14
Single Interferer Model: Cognitive Radio Interference WM CR Detection range, R det 29 Single Interferer Model: Cognitive Radio Interference Probability of Interference (%) Traditional CR R 0 = 80m R det = 176m Interferer Distance (m) 30 15
Single Interferer Model: Optimum Detection Range Detection Range (R det / Ro) 10dB better Detection = 10x lower interference n=4 n=2 Maximum Acceptable Interference (%) For < 1% interference, detection range 5 times Ro 31 Multiple CR Model: Multiple Cognitive Radios So far: Single Interferer Worst-case location WM system Now: Multiple interferers Spread randomly over area Function of density 16
Multiple CR Model: Cooperative Spectrum Sensing No Cooperation Partial Cooperation Full Cooperation WM system 33 Multiple CR Model: Detection Range of CR Devices Detection Range (R det / Ro) n=4 10dB better Detection = 100x lower interference Easier to get better performance Fully Cooperative Network Ro=80m Maximum Acceptable Interference (%) For interference < 1%, R det 4.5 Ro, similar to single interferer analysis Dhillon, Brown, Models for Analyzing Cognitive Radio Interference to Wireless Microphones in TV Bands, IEEE DySpAN 2008. 34 17
Single Interferer Model: What if the Licensed user changes? Probability of Interference (%) Traditional CR R 0 = 80m R det = 176m What if WM get only 0.8 R 0? Interferer Distance (m) It s a two-way street. 35 Conclusions Harmful Interference model should be considered up front Affords better protection to both incumbent and white spaces radio Many choices are possible. 36 18