specnet
DESCRIPTION
SpecNet : Spectrum Sensing Sans Frontieres. SpecNet. Anand Padmanabha Iyer (MSRI) Krishna Kant Chintalapudi (MSRI) Vishnu Navda (MSRI) Ramachandran Ramjee (MSRI) Venkat Padmanabhan (MSRI) Chandra Murthy ( IISc ). A Case for Sub-GHz in Rural India. - PowerPoint PPT PresentationTRANSCRIPT
SpecNet
SpecNet : Spectrum Sensing Sans Frontieres
Anand Padmanabha Iyer (MSRI)Krishna Kant Chintalapudi (MSRI)
Vishnu Navda (MSRI)Ramachandran Ramjee (MSRI)Venkat Padmanabhan (MSRI)
Chandra Murthy (IISc)
A Case for Sub-GHz in Rural IndiaCommercial Broadband Connectivity in Rural India is uneconomical
• 70% of Indian population • 500,000 villages 1-2SqKm in area• 80% villages under 1000 people• Low Income, Low user density
Long Distance Wi-Fi (in 2.4 GHz)
• High gain directional wireless links for back-haul• Needs a tower roughly per village• Does not scale economically
Sub-gigahertz license free spectrum
• Excellent range• about 10Km at 300 MHz 30 dBm• A single tower can provide for several tens of village• Has the potential to enable economically viable connectivity
Whitespaces in the Heart of Bangalore
FM
TV
GSM
CDMA
• Over 90% of the spectrum remains unused in the sub-gigahertz spectrum
• Only 16/566 MHz of TV spectrum is used
• Prior studies in U.S.A, Spain, France, Singapore, China etc.
UnusedT.V. Bands
Options for Spectrum Usage in India
Three Options to Reclaim Unused Spectrum
• Auction away to Commercial Providers- no commercial interest in rural deployments
• Create a License Free Band Similar to ISM- can potentially spur tremendous growth- government loses the opportunity to monetize the band
• Opportunistic Usage of Unused Spectrum (e.g. FCC in U.S)- perhaps best of both worlds
Unused T.V. Bands in India
• In U.S almost all allocated T.V. bands are in use at one or more locations
• A large number of T.V. bands are not used anywhere in India!
Mapping Spectrum Usage
How can we construct and maintain spatio-temporal spectrum usage maps?
• A collaborative measurement platform is the key!
• A network of spectrum sensing devices.
The first step is to understand the nature spectrum usage
• India is a large country
• Information is not as readily available as in developed countries
- e.g. no online T.V. tower location database
2500 Km
2000 Km
SpecNet
SpecNet : A platform that enables development of collaborative spectrum measurement based applications using networked spectrum analyzers
Remote User
Spectrum Analyzer
The Power of SpecNet
Construction and Maintenance of Real-Time White Space Spatio-Temporal Usage Maps • Can help future white space service providers to plan their infrastructure deployments
• Can aid the operation of white space devices
Enable remote measurements• Help cognitive researchers to access real data from across the world to validate their models
Remote UserSpectrum Analyzer
Real-Time Distributed Applications that Utilizes Spectrum Measurements• Researchers can implement and test their ideas using real-time sensing data
SpecNet Operation
XML RPC
SpecNetUser
SpecNetServer
import xmlrpclib;apiServer = xmlrpclib.ServerProxy(“https://research.microsoft.com/specnet/api”);devices = apiServer.getDevice();
User CodeSpectrum Analyzers
• Volunteering spectrum analyzer (SA) owners register and connect to SpecNet• SA owners specify times of public usage • Connect to SpecNet server
Users• Use SpecNet API to write applications • SpecNet API provides an easy to use abstraction layer implemented as XML-RPC for flexibility
SpecNet Server
• Interprets the API commands to task individual spectrum analyzers
• Schedules task intelligently to optimize resource utilization
Fundamental Tradeoffs
Time versus Resolution Bandwidth
• Sort of like the Heisenberg’s uncertainty principle• The finer frequencies you wish to resolve, the longer it takes
Time versus Noise Floor• A lower resolution bandwidth implies lower noise• Can detect weaker signals• Also means it takes longer to detect weaker signals
Resolution Bandwidth
• Ability to distinguish between two nearby parts of the spectrum
spanscan fRBWct )/1(1
RBWcN 2)/1(3 Nctscan or
Log(RBW)N
f
102 106
-80
-110
A Simple First Example
(Lat,Lng)r
Fc
BW
Behind the Scenes
• For each spectrum analyzer SpecNet maps the required noise floor Nf to its resolution bandwidth.
• It then issues commands to each spectrum analyzer to scan.
• Collects the results and sends them back to the user
import xmlrpclib;apiServer = xmlrpclib.ServerProxy(“https://research.microsoft.com/specnet/api”);
for d in devices: val = apiServer.getPowerSpectrum(‘NOW’,d,Fc,BW,Nf);
devices = apiServer.getDevices([lat,lng,r]);
Example II: Occupancy Detection
Occ
upan
cy
1
0
r
[lat,lng]d
import xmlrpclib;apiServer = xmlrpclib.ServerProxy(“https://research.microsoft.com/specnet/api”);
oc = apiServer.getOccupancy(NOW,[lat,lng,r], Fc,BW,P)
• Must detect a transmitter with power P anywhere within the circle
P
d
)log(100 dPPd
• SpecNet server chooses a resolution bandwidth such that noise floor is Pd - 5dB
Scheduling Multiple Spectrum Analyzers
Goal : To minimize scan time e.g. 300-600 Mhz
Strategy I : Partition the frequency space
• S1 scans 300-400 MHz, S2 scans 400-500 MHz, S3 scans 500-600 MHz• Time taken reduces linearly i.e. by a factor of 3
Strategy II : Partition the geographical space
• All spectrum analyzers scan 300-600Mhz• Scan only a part of the geographical area• Scan time = max( k1d1
, k2d2 , k3d3
)• Scan time decreases super-linearly
S1
S2
S3
d1
d2
d3
d
Strategy III : A Hybrid Partitioning
• Find an optimal combination of area and frequency partitioning
kdtscan
Example III : Estimating Transmitter’s Footprint
Locating T.V Transmitter Towers in India
• There is no readily available database that provide this information in India like in the U.S• We tried to obtain this information using RTI
- Incomplete information (100/700+)- Erroneous information
• Provided to the SpecNet users as an API
Typically use a path loss model
Log Distance Path Loss Model
- Longley Rice Model
)log(100 dPPd
Localizing Bangalore T.V TowerHow can one localize the T.V. transmitter?
• Basic Idea : Use a path loss model and find the location that fits the data the best
• [loc,P] = estimateTransmitterParams (pos, power, model)
6 Km error in the RTI Data
Predicting T.V. Signal Strength
125 Locations in Bangalore
• 5 – 8 dB variation due to fading at various locations• 60 Test Locations spread across Bangalore• Performance is within the variation limits
A Real-Time Demo App
Find the strongest FM Station!
1. Scan from 50-150 Mhz at a high resolution
2. Find the strongest point in the spectrum
3. Scan ± 500Khz around the strongest point at a finer resolution bandwidth
Opportunistic Spectrum Usage in U.S
FCC Ruling (2008) : Permits opportunistic usage of T.V whitespaces in the sub-gig Hz in US• Will lead to tremendous innovation and development in wireless
communication
Putting things in Perspective
ISM Band Before 1985• Wasteland for emissions due to Industrial, Scientific and Medical equipment
ISM Band Today • Tremendous innovation • WiFi, Bluetooth, Zigbee, WiBree, Cordless phones, etc.