White Space Networking in the TV Bands & Beyond Ranveer Chandra Microsoft Research Collaborators: Thomas Moscibroda, Victor Bahl, Bozidar Radunovic, Ivan Tashev, Paul Garnett, Paul Mitchell Rohan Murty (Harvard), George Nychis (CMU), Eeyore Wang (CMU), Aakanksha Chowdhery (Stanford) The Big Spectrum Crunch FCC Broadband Plan calls it the Impending Spectrum Crisis Limited amount of good spectrum, while demand increasing exponentially Growing Demand 24 HOURS 20X - 40X 50 BILLION
35X UPLOADED EVERY OVER THE NEXT CONNECTED DEVICES 2009 LEVELS 60 SECONDS FIVE YEARS BY 2020 BY 2014 Video
Uploads Streaming Video Increasing Wireless Demand Devices Proliferation* Mobile Data Traffic** Industry Forecasts of Mobile Data Traffic 50X 45X Traffic Relative to 2009 40X 35X
Cisco 30X Coda 25X Yankee Group 20X Average 15X 10X 5X See Ericsson Press Release, quoting its President and Chief Executive Officer Hans Vestberg, April 13, 2010, available at
* http://www.ericsson.com/thecompany/press/releases/2010/04/1403231 **. Federal Communications Commission, Staff Technical Paper, Mobile Broadband: The Benefits of Additional Spectrum, OBI Technical Paper No. 6 (Oct. 2010). 0X 2009 2010 2011 2012 2013 2014 The Big Spectrum Crunch
FCC Broadband Plan calls it the Impending Spectrum Crisis Limited amount of good spectrum, while demand increasing Globally, mobile data traffic is expected to double every year through 2013. Whether an iPhone, a Storm or a Gphone, the world is changing. Were just starting to scratch the surface of these issues that AT&T is facing., Cisco Systems, 2009 CTIA has requested for 800 MHz by 2015 Customers Angered as iPhones Overload AT&T Headline in New York Times , 2.Sept 2009 FCC promises to provide 500 MHz by that time The industry is quickly approaching the point where consumer demand for mobile broadband data will surpass the telecommunication companies abilities to handle the traffic. Something needs to happen soon De la Vega, chair of CTIA, 2009
Heaviest Users of Phone Data Will Pay More Headline in New York Times , 2.June 2010 Spectrum Allocation in the US 5 In contrast... Large portions of spectrum is unutilized 6 Dynamic Spectrum Access Power PU1 PU3 PU2
PU4 Frequency Determine available spectrum (white spaces) Transmit in available frequencies Detect if primary user appears Move to new frequencies Adapt bandwidth Adapted andfrompower levels Bob Brodersens presentation at Microsoft Research Summit 2008 Cognitive (Smart) Radios Frequency Signal Strength
Signal Strength 1. Dynamically identify currently unused portions of spectrum 2. Configure radio to operate in available spectrum band take smart decisions how to share the spectrum Frequency 8 Networking Challenges The KNOWS Project (Cogntive Radio Networking) How should nodes connect? Which spectrum-band should two cognitive radios use for transmission? 1. Frequency?
2. Channel Width? 3. Duration? How should they discover one another? Need analysis tools to reason about capacity & overall spectrum utilization Which protocols should we use? MSR KNOWS Program v1: Ad hoc networking in TV white spaces Capable of sensing TV signals, hardware functionality DySPAN 2007, MobiHoc 2007, LANMAN 2008 v2: Infrastructure based networking(WhiteFi)
Capable of sensing TV signals & microphones, deployed in lab SIGCOMM 2008, SIGCOMM 2009 (Best Paper) v3: Campus-wide WhiteFi network + geolocation Deployed on campus, and provide coverage in MS Shuttles 2010 (Top 3 paper), CoNEXT 2011 (Top 3 paper) v4: White spacesDySPAN beyond TV spectrum Spectrum measurements to identify additional white spaces In this talk DSA: Need & a primer Networking in the TV White Spaces Whats missing in the TV white space ruling Open research questions
DSA in other network bands 11 What are TV White Spaces? Wireless Mic TV 0 54-88 170-216 470 698 MHz50 TV Channels -60 ISM (WiFi) 2400 2500 5180 5300
White spaces 7000 MHz Each channel is 6 MHz wide dbm TV Stations in America -100 470 MHz Frequency 700 MHz White Spaces are Unoccupied TV Channels 12
v3 Goal: Campus WhiteFi Network Base Station (BS) Good throughput for all nodes Avoid interfering with incumbents 13 WHY NOT USE WI-FI AS IS? 14 White Spaces Spectrum Availability Differences from ISM(Wi-Fi) Fragmentation Fraction of Spectrum Segments
0.8 Urban 0.7 0.6 Suburban 0.5 Rural Variable channel widths 0.4 0.3 0.2 1 2 0.13 4 5 0
1 1 2 3 4 5 2 3 4 5 6 >6 # Contiguous Channels Each TV Channel is 6 MHz wide Spectrum
Use is Fragmented multiple channels for more bandwidth 15 White Spaces Spectrum Availability Differences from ISM(Wi-Fi) Fragmentation Variable channel widths Spatial Variation Cannot assume same channel free everywhere 1 2 3 4 5 1 2 3 4 5 TV Tower
Location impacts spectrum availability Spectrum exhibits spatial variation 16 White Spaces Spectrum Availability Differences from ISM(Wi-Fi) Fragmentation Variable channel widths Spatial Variation Cannot assume same channel free everywhere 1 2 3 4 5 1 2 3 4 5 Temporal Variation
Same Channel will not always be free Any connection can be disrupted any time Incumbents appear/disappear over time Must reconfigure after disconnection 17 Design Challenges Primary user detection Channel selection Recovering from disruptions Base station placement
Discovery Security 18 DETECTING PRIMARY USERS 19 KNOWS White Spaces Platform Windows PC TV/MIC detection Scanner (SDR) FFT Net Stack
FPGA UHF RX Daughterboard Whitespace Radio Connection Manager Atheros Device Driver Wi-Fi Card UHF Translator Variable Channel Width Support 20
Geo-location Service (http://whitespaces.msresearch.us) Use centralized service instead of sensing Returns list of available TV channels at given TV/MIC data location nPropagatio Modeling (FCC (FCC CDBS, CDBS, others) others) Location (Latitude, (Latitude, Longitude) Longitude)
Terrain Data (Globe, (Globe, SRTM) SRTM) Features Can configure various parameters, e.g. propagation models: L-R, Free Space, Egli detection threshold (-114 dBm by default) Protection for MICs by adding as primary user Accuracy: combines terrain sources for accurate results results validated across1500 miles in WA state Includes analysis of white space availability (forthcoming) Internationalization of TV tower data White-Fi: Geo-Location Database
Our geo-location database FCC mandated Pros & Cons Sensing: Pros: Leads to more availability of white spaces, allows disconnected operation Cons: Energy hungry, inaccurate, expensive Geo-location: Pros: easily extensible, simpler to implement Cons: miss out on white spaces, e.g. indoors CHANNEL SELECTION 24 Channel Assignment in Wi-Fi
1 6 11 1 6 11 Fixed Width Channels Optimize which channel to use 25 Spectrum Assignment in WhiteFi Spectrum Assignment Problem Goal
Maximize Throughput Include Spectrum at clients 1 2 3 4 5 Assign 1 2 3 4 5 Center Channel & Width Fragmentation Optimize for both, center channel and width Spatial Variation BS must use channel iff free at client 26
Accounting for Spatial Variation 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 =
1 2 3 4 5 27 Intuition Intuition Use widest possible channel BS But Limited by most busy channel 1 2 3 4 5 Carrier Sense Across All Channels All channels must be free BS(2 and 3 are free) = BS(2 is free) x BS(3 is free)
Tradeoff between wider channel widths and opportunity to transmit on each channel 28 3.5 3 2.5 2 1.5 1 0.5 0 M Cham -value T h ro u g h p u t ( M b p s ) Multi Channel Airtime Metric (MCham)
20 Mhz 5 MHz 10 MHz W BS ( c ) MChamn (F, W) = n 5 Mhz c( F ,W ) 0 10 20 30 40 Background1 traffic
3 4 delay 5 (ms) 2 - Packet 50 Pick (F, W) that maximizes (N * MChamBS + nMCham ) 2.5 n 1 20 Mhz 10 MHz (c) = Approx. opportunity node n will
(2) n BS(2)5 MHz Free BS Air Time on Channel 2 2 Contention BS(2) = Max (Free Air Time on 2, 1/Contention) 1.5 get to transmitchannel on channel c 1
0.5 0 0 5 10 15 20 25 30 35 40
45 50 Background traffic - Packet delay (ms) 29 Campus Wide WhiteFi Network FCC Experimental License (Granted: July 6, 2009) Goal: Deploy a white space network that provides corp. net access in Microsoft shuttles Centered at (47.6442N, 122.1330W)
Area of 1 square mile Perimeter of 4.37 miles WSD on 5-10 campus buildings Fixed BS operate at 4 W EIRP WSD inside shuttles at 100 mW EIRP 6-2
6-1 5-3 4-24-1 5-2 5-1 3-2 3-1 1-1 1-2 Range Experiments Raw received power at different Distances from the transmitter MSRs Redmond Campus Route taken by the shuttle (0.95 miles x 0.75 miles) ~4x range compared to 2.4 GHz (Wi-Fi) with same transmit power and receiver sensitivity White-Fi: Deployment Implemented and deployed the worlds first operational white space network on Microsoft Redmond campus (Oct. 16,
2009) White Space Network Setup Shuttle Deployment WS Antenna WS Antenna on MS Shuttle Data packets over UHF In this talk DSA: Need & a primer Networking in the TV White Spaces Whats missing in the TV white space ruling Open research questions DSA in other network bands 33
CoNEXT 2011 Coexisting with MICs? FCC & other regulators reserve entire channel for MICs Setup Observations Time: Even short packets (16 s) every 500 ms cause audible interference Power: No interference when received power was below squelch tones Frequency: #subcarriers to suppress depends on distance from MIC receiver How to reuse a TV channel without causing audible interference to MIC? 34
Coexistence among WS devices Results from our indoor WS testbed 4W 100mW Carrier Sense does not work! Our Solution: Weeble PHY: adaptive preamble detection at low SNR MAC: Recover CSMA using PHY detector 35 Indoor White Spaces F r a c ti o n o f lo c a ti o n s ( C D F ) Geo-location DB is conservative indoors LR-based models do not account for losses
through doors & walls Sensing is expensive! Can we install in-building geo-location servers to provide benefit of both? 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
Attenuation caused by door (dB) 36 LOOKING AHEAD: WHITE SPACES BEYOND TV BANDS With: Aakanksha Chowdhery (Stanford), Paul Garnett, Paul Mitchell 37 PCAST Report, July 2012 Directs govt. agencies to identify 1000 MHz and create the first shared use spectrum super highways Creation of test city & mobile test service to support development of DSA techniques Suggests possible frequencies suitable for DSA 38
What spectrum is good for DSA? Prior spectrum occupancy measurements: Limited time span (1 hour to 1 week) Uses fixed thresholds to determine occupancy Mostly single point measurements (or few static points) No easy way to translate occupancy to DSA! 39 Our Approach Fixed RFEye Measurements Mobile Spectrum Measurements FCC Spectrum Dashboard
Combined DSA metric Spectrum goodness for DSA at location 40 Initial Results Power Spectral Density Mean Spectrum Available Ongoing work: Incorporate availability in time, space and frequency into a DSA metric 41 Summary DSA has potential to unlock large portions of spectrum for unlicensed use
TV white spaces are a good first step New networking paradigm to build DSA networks WhiteFi is the first step to network devices Several exciting research problems need to be solved: coexistence, new DSA bands, sensing, and many more http://research.microsoft.com/knows 42 WhiteFi: Press WhiteFi: Regulatory Impact Radiocommunication Sector India Oct. 22, 2009 Singapore Apr. 8, 2010
Federal Communications Commission, USA (FCC), Apr. 28 & Aug. 14, 2010 China Jan. 11, 2010 Brazil (Feb. 2, 2010) Standards Fisher Communications Inc. Jan. 14, 2010 Industry Partners Jan. 5, 2010 White-Fi & Broadcast TV TV broadcasters opposed to white space networking
Hillary Clinton lobbying for broadcasters against White-Fi Our system demonstrated that we can reuse unused spectrum without hurting broadcasters KOMO (Ch. 38) KIRO (Ch. 39) White-Fi (Ch. 40) THANK YOU! 46
