To Fragment or Not To Fragment: Viability of NC OFDMA in Multihop Networks Muhammad Nazmul Islam WINLAB, Rutgers University
Availability of Non Contiguous Spectrum Demand for wireless services is increasing rapidly Qualcomm predicts a 1000x increase by 00 FCC has opened up 300 MHz in TV bands Plans to open up additional 00 MHz license by rule bands by 00 Any radio can use these bands if it abides by FCC rules If uncoordinated networks use these bands, they will adjust spectrum usage according to traffic demands Available bands will become non contiguous TV white space is itself non contiguous in nature
Case for Noncontiguous OFDMA I Three available channels 1 3 C Node A transmits to node C via node B. B Node B relays node A s data and transmits its own data to node C. X Node X, an external and uncontrollable interferer, transmits in channel. A If we use max min rate objective and allocate channels, node B requires two channels; node A requires one channel Scheduling options for Node A and Node B? 3
Case for Noncontiguous OFDMA II #1: Contiguous OFDM 1 C #: Multiple RF front ends 1 3 C #3: Non Contiguous OFDM (NC OFDMA) Nulled Subcarrier 1 3 C B B B 3 X X X A Transmission in link BC suffers interference in channel A Spectrum fragmentation limited by number of radio front ends A NC OFDM accesses multiple fragmented spectrum chunks with single radio front end 4 4
NC OFDM Operation #3: Non Contiguous OFDM X[1] X[3] X[] = Serial to Parallel 0 X[1] X[3] IFFT x[1] x[] x[3] Parallel to Serial D/A Nulled Subcarrier Modulation B 1 3 AP Node B places zero in channel and avoids interference X Node A, far from the interferer node X, uses channel. Both nodes use better channels. Node B spans three channels, instead of two. Sampling rate increases. A NC OFDM accesses multiple fragmented spectrum chunks with single radio front end
Benefits and Challenges of NC OFDMA Benefits: Avoids interference, incumbent users. Uses better channels Challenges: Increases sampling rate Increases ADC & DAC power Increases amplifier power 6
Power Consumption Model ADC and DAC power depend on the sampling rate Other blocks deal with analog signals power consumption does not depend on sampling rate We ignore programmable amplifier s power consumption here. Tx Power, pt fs pm Rx Power, pr 1 f 1 s m M. 1,, 1, constants f s - Sampling rate pm - Allotted power in channel m. 7
Optimization Formulation Our Approach Formulation min i N P sys, i ( PTxCkt, j N, i PRxCkt, i Pij ) Psys i i N gijpij W* log ( 1 ) fij N W 0 (i,j) (N,N) Interference, flow conservation, Notation P sys i Total System Power Minimization half N Set of nodes., System power of node Individual System Power Constraint Capacity Constraint duplex constraints P TxCkt, i Ckt power of node i's Tx path, Pij Emitted power at link ij i. W Bandwidth, N0 Noise Spectral Density, g ij Link gain of ij, f Flow at ij. ij
Transmit Power Minimization Waterfilling Formulation j N min P ij P i N Tx i P Tx, i, i N Total Trasnmit Power Minimization Individual Transmit Power Constraint gijpij W* log ( 1 ) fij N W 0 (i,j) (N,N) Interference, flow conservation, Notation P Tx i half N Set of nodes., Transmit power of node P ij Emitted power at link ij Capacity Constraint duplex constraints i. W Bandwidth, N0 Noise Spectral Density, g ij Link gain of ij, f Flow at ij. ij
Comparison with Waterfilling I Point to Point Link Channel width 3 MHz. Minimum required rate 18 Mbps. Waterfilling selects good channels across the whole list. Our approach selects two non contiguous good neighbors. 10
Comparison with Waterfilling II Point to Point Link Waterfilling consumes less transmit power. Our approach consumes less system power. 11
Multi hop Scenario (TV Bands in Wichita, KS) D(1) = 3 D() = 6 Channel width 6 MHz. Two sessions. Required rate 10 Mbps. Node 1 transmits to 3. Node transmits to 4. S(1) = 1 S() = 4 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 1
Waterfilling Approach D(1) = 3 D() = 6 3 4 S(1) = 1 S() = 4 Channel Gain (db) 18 130 13 134 Link 1 6 17 3 4 47 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 13
Waterfilling Approach D(1) = 3 D() = 6 17 3 4 S() = 4 S(1) = 1 Channel Gain (db) 13 133 134 13 136 Link 3 6 17 3 4 47 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 14
Waterfilling Approach D(1) = 3 D() = 6 17 3 4 S() = 4 S(1) = 1 17 Channel Gain (db) 19 131 133 13 Link 4 6 17 3 4 47 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 1
Waterfilling Approach D(1) = 3 D() = 6 17 3 4 S() = 4 S(1) = 1 4 17 Channel Gain (db) 18 130 13 134 136 Link 6 6 17 3 4 47 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 16
Our Approach (Low Power ADC & DAC) 6 D(1) = 3 D() = 6 3 4 6 S(1) = 1 S() = 4 Index Location (MHz) 6 17 3 4 47 7 79 8 491 7 33 671 17
Comparison with Waterfilling (Low Power ADC & DAC) Our approach reduces system power consumption by 4 db. 18
Our Approach (USRP ADC & DAC) D(1) = 3 D() = 6 17 4 S(1) = 1 Index Location (MHz) S() = 4 6 17 3 4 47 7 79 8 491 7 33 671 19
Comparison with Waterfilling (USRP ADC & DAC) USRP ADC and DAC s power consumption curves are steeper. Waterfilling spans more spectrum and consumes a lot of system power Our approach reduces system power consumption by 10 db! 0
Conclusion Researchers focus on channel gains and traffic demands to determine power control, scheduling and routing variables Our results reveal that hardware configuration of the radio front ends, e.g., slope of ADC and DAC power consumption curves, can influence these variables. 1
Future Works Power consumption models of programmable amplifier Non linear due to bandwidth gain product Investigate multi frond end radio s power consumption Analog power may increase since there are multiple components Digital power may decrease since each ADC/DAC spans narrower spectrum
Questions? Thank You! 3
Power Minimization Transmit Power Minimization (Waterfilling) System Power Minimization (Our Approach) min s.t. p m m M p m M W m log (1 0 m M pmg N W 0 m ) r min s.t. p m m M p m M W m log (1 0 m M f s pmg N W Spectrum Span Constraints 1 0 m 1 ) r f s
References 1. Qualcomm data challenge, accessed March 013, http://www.qualcomm.com/media/documents/wireless networks risingmeet 1000x mobile data challenge. C. Cordeiro, K. Challapali, D. Birru, and S. Shankar, IEEE 80.: the first worldwide wireless standard based on cognitive radios, in Proc. IEEE DySPAN 00, Nov. 00, p. 38337. 3. Enabling innovative small cell use in 3. GHZ band NPRM & order, accessed March 013, http://www.fcc.gov/document/enablinginnovative small cell use 3 ghz band nprm order 4. S. Cui, A. Goldsmith, and A. Bahai, Energy constrained modulation optimization, IEEE Transactions on Wireless Communications, vol. 4, pp. 349 360, SEP 00. Y. Shi and Y. T. Hou, Optimal power control for multi hop software defined radio networks, in Proc. IEEE INFOCOM 007, May 007, pp. 1694 170.