Interference Alignment and RAMCOORAN Co-Ordinated Multi-Point with 802.11ac-feedback: Testbed Results Per Zetterberg 1
Astonishing Result RAMCOORAN Cadambe/Jafar, ” Interference Alignment and Degrees of Freedom of the K-User Interference Channel”, K-transmitters and K-receivers, K-links: IEEE Trans, Information Theory K/2 simultaneous interference-free links. 2008. Requires coding over multiple channel realizations. Global channel knowledge required. 2
Channel extension (Cadambe/ Jafar) RAMCOORAN ( ) ( ) ( ) [ ] = ij ij ij H diag h f , 1 t , , h f , t 1 m m = + m 2 n 1 Channel extended MIMO channel (my + 3 n 1 => 1 . 5 wording) + 2 n 1 3
The ”alignment” RAMCOORAN ( ) = + + + = H H y w x x ... x w x D 1 N D 4
Implementation IA RAMCOORAN Feedback: Wired ethernet 𝒗 1 MS 1 BS 1 𝒘 1 𝒗 2 BS 2 MS 2 𝒘 2 𝒗 3 MS 3 𝒘 3 BS 3 5
Implementation: CoMP RAMCOORAN Feedback: Wired ethernet 𝒗 1 MS 1 BS 1 𝒘 1 𝒗 2 BS 2 MS 2 𝒘 2 𝒗 3 MS 3 𝒘 3 BS 3 6
Beamformer RAMCOORAN 2 * u H v = k k , k k SNIR ∑ ≠ k 2 ~ + σ * 2 u H v k k , n n n k “Approaching the Capacity of 2 ∑ ~ σ = σ Wireless Networks through 2 2 max , 0 . 001 H N k , n Distributed Interference n Alignment", by Krishna Gomadam, Viveck R. Cadambe and Syed A. Jafar. 7
Beamformer initialization CoMP RAMCOORAN [ ] = = H H H H H U S V , 1 11 12 , 13 1 1 1 [ ] = = H H H , H H U S V 2 21 22 , 23 2 2 2 [ ] = = H H H , H H U S V 3 31 32 , 33 3 3 3 [ ] ( ) ( ) ( ) ~ V = V :, 1 V :, 1 V :, 1 1 2 3 ( ) ~ ~ ~ − 1 = H W V V V 8
3MS The testbed RAMCOORAN 10m 3BS P=10dBm NF=10-11dB 10m 9
IEEE 802.11ac feedback IEEE 802.11ac feedback RAMCOORAN Training symbols One per antenna 10
IEEE 802.11ac feedback, contd. RAMCOORAN H = H U S V [ ] ( ) ( ) − m , 1 i n N N N c r ( ) ~ ∏ ∏ r φ φ = ψ j j T N . 1 , i . V D 1 e , , e , 1 G I . i i r − × i i 1 l i l i N N r c = = + i 1 l i 1 Number of phis and psis : NcNr-Nc^2/-Nc/2 π 0 π Range 0, 2 , / 2 Range Quantization bits =7 or 9 Quantization bits 5 or 7 The whole 6x2 matrix is treated by the three MSs. 11
Frequency granularity RAMCOORAN Our implementation: 14MHz 1 38 -19, -18, -17, -16, -15, -14, -13, -12, -11, -10, -9, -8,-6, -5, -4, -3, -2, -1, 1, 2, 3, 4, 5, 6, 8, 9, 10, 11,12, 13, 14, 15, 16, 17, 18, 19 20 -18, -16, -14, -12, -10, -8, -6, -4, -2, -1, 1, 2, 4, 6,8, 10, 12, 14, 16, 18 2 4 8 -12, -8, -4, -1, 1, 4,8, 12 8 6 -16, -8, -1, 1, 8, 16 4 -16,-1,1,16 16 12
SNR feedback SNR feedback RAMCOORAN 20 -18, -16, -14, -12, -10, -8, -6, -4, -2, -1, 1, 2, 4, 6,8, 10, 12,14, 16, 18"; 14MHz 1 2 10 -16, -12, -8, -4, -1, 1, 4, 8, 12, 16 8 12, -8, -4, -1, 1, 4, 8,12 4 4 -8, -1, 1, 8 8 16 2 -16,16 13
Interpolate between subcarriers RAMCOORAN Reuse V on adjacent subcarriers. Interpolate SNR values Reconstruct H as: H=diag(SNR0,SNR1)*V. 14
Overhead RAMCOORAN • The three short frames ≈ 2*60µs+num_users*60µs • Gaps=2 SIFS + num_users SIFS • Number of feedback bits per V matrix= (NcNr-Nc^2/-Nc) (7+9) • Number of feedback for SNR num_users*8*SNRs*Nr Feedback time in our case≈ 350µs+1065µs/(Ng*R) Ng=4, R=2 => 500µs With update interval : 20ms => Overhead 2.5% Based on: Improved MU-MIMO Performance for Future 802.11 Systems Using Differential Feedback , Ron Porat 15
Our implementation 1. BS1 sends P0,P1 2. BS2 sends P2,P3 RAMCOORAN 3. BS3 sends P4,P5 4. MS1-MS3 sends compressed V matrix to BS1. 5. BS1 un-compresses calculates all beamformes using max SINR on 38subcarriers. 6. Each BS sends beamformed frames. 7. MSs saves all data. 8. Post-processing of received signals. 20ms P P P P P P MCS0-9 MCS0-9 MCS0-9 0 1 2 3 4 5 Training frame Payload frames.. 16
Measurement environment RAMCOORAN C B0 B2 B1 C 17
Results LoS (stationary) RAMCOORAN Measured sum throughput 16 14 Sum througput bits/symbol/subcarrier 12 10 8 6 4 CoMP IA TDMA MIMO 2 full-reuse SIMO full-reuse MIMO 0 1 2 4 8 16 38 Ng 18
Results NLoS (stationary) RAMCOORAN Measured sum throughput 16 14 Sum througput bits/symbol/subcarrier 12 10 8 6 4 CoMP IA 2 TDMA MIMO full-reuse SIMO full-reuse MIMO 0 1 2 4 8 16 38 Ng 19
Why is NLoS better for IA? RAMCOORAN 25 20 15 C/Imin (dB) 10 5 LoS 0 NLoS -5 -10 -10 -5 0 5 10 15 20 C/Imax (dB) 20
Time varying channels RAMCOORAN 21
Time varying channels RAMCOORAN 6 People Minute Minute walking after before 5 Throughput 4 3 2 1 0 0 5 10 15 Frame number 22
Simulated on Time varying channels measured RAMCOORAN channels 6 5 Throughput 4 3 2 1 0 0 5 10 15 Frame number 23
0.5ms feedback Time varying channels delay RAMCOORAN 6 5 4 Throughput 3 2 1 0 0 5 10 15 Frame number 24
Conclusion RAMCOORAN • IA gives 25% sum throughput gain over SU MIMO on stationary channel (LoS and NLoS averaged) • CoMP gives 71% gain over SU MIMO. • All schemes limited by RF impairments. • Geometry factor imortant for IA versus MIMO. • Full reuse SIMO and MIMO worse than SU MIMO. • Next step: analyzing measurements with time varying channels. 25
Results versus EVM model RAMCOORAN Schem e Actual I m pairm ent perform ance m odel IA 11.1 11.7 CoMP 15.2 17.3 SU MIMO 8.9 8.1 FR SIMO 6.5 6.2 FR SIMO 2.3 2.3 26
Recommend
More recommend
