Opportunistic Beamforming with Dumb Antennas for Clustered OFDM Patrick Svedman, Katie Wilson and Len Cimini 1 November 28, 2003
Outline PSfrag replacements OFDM Multiuser Diversity Opp. Beamforming Opp. OFDM Conclusions Future Work Feedback Prel. Results Patrick Svedman, Katie Wilson and Len Cimini 2 November 28, 2003
Orthogonal Frequency Division Multiplexing Instead of equalizing the fading channel, divide the channel into sub-channels, which can be considered as flat. Channel Gain 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 10 5 0 5 10 MHz Patrick Svedman, Katie Wilson and Len Cimini 3 November 28, 2003
Multiuser Diversity Uncorrelated (time-varying) channels of different users the users can be scheduled in a favourable way. Schedule a user when his channel is relatively strong. SNR PSfrag replacements time User 1 User 2 User 3 Patrick Svedman, Katie Wilson and Len Cimini 4 November 28, 2003
Opportunistic Beamforming Traditionally, multiple antennas have often been used to combat fading (e.g. space-time block coding). Opportunistic beamforming using dumb antennas: randomly form the beams for each data block to increase the fading of the users! Also stationary users will experience temporal fading! C.f. Viswanath, Tse and Laroia Opportunistic Beamforming Using Dumb Antennas IEEE Trans. Inform. Theory, vol 48, pp. 1277-1294, June 2002. Patrick Svedman, Katie Wilson and Len Cimini 5 November 28, 2003
Opportunism for an OFDM downlink In an OFDM downlink, different users can be scheduled on different subcarriers. This introduces another dimension on which the users can be scheduled! What if the users experience relatively flat channels? More frequency fading can be induced by having different beamforming weights on different subcarriers. Patrick Svedman, Katie Wilson and Len Cimini 6 November 28, 2003
Time-frequency grid of one user 80 70 60 50 40 30 20 10 0 60 50 40 30 20 10 frequency 0 0 20 40 60 80 100 time Patrick Svedman, Katie Wilson and Len Cimini 7 November 28, 2003
What about feedback? The scheduler requires knowledge of the frequency responses of all users much feedback Improvement 1: Clustered OFDM Divide the subcarriers into clusters (of adjacent subcarriers). If the cluster-size is appropriate, the correlation between the sub-carriers in one cluster is high. Improvement 2: Feed back information only about the strongest clusters. The weakest will not be scheduled anyway! much less feedback. Patrick Svedman, Katie Wilson and Len Cimini 8 November 28, 2003
Preliminary Simulation Results Comparison between 1. Opportunistic OFDM with clustering and reduced feedback 2. Opportunistic OFDM with full feedback 3. A smart antenna solution (beamforming on the largest eigenvalue), with round-robin scheduling Simple scenario HIPERLAN/2 channel model (64 sub-carriers) Equally distributed users Maximum system throughput considered Fairness not considered yet Patrick Svedman, Katie Wilson and Len Cimini 9 November 28, 2003
Throughput vs Number of Users: SNR = 0 db, Channel delay-spread = 100 ns 10 0.5 Throughput 10 0.4 10 0.3 10 0.2 10 0.1 Method 1, waterfilling Method 1, equal power Method 2, waterfilling Method 2, equal power Method 3, waterfilling 0 10 20 30 40 50 users Patrick Svedman, Katie Wilson and Len Cimini 10 November 28, 2003
Throughput vs Channel Delay-spread: SNR = 0 db, 32 users 10 0.5 Throughput 10 0.4 10 0.3 0 20 40 60 80 100 120 140 160 gamma Method 1, waterfilling Method 1, equal power Method 2, waterfilling Method 2, equal power Method 3, waterfilling Patrick Svedman, Katie Wilson and Len Cimini 11 November 28, 2003
Throughput vs amount of feedback information Q: SNR = 0 db, Channel Delay-spread = 100 ns throughput 10 0.4 10 0.3 10 0.2 10 0.1 0 10 20 30 40 50 users Method 1, Q=1 Method 1, Q=2 Method 1, Q=3 Method 1, Q=16 Method 2 Patrick Svedman, Katie Wilson and Len Cimini 12 November 28, 2003
Throughput vs cluster-size N: 10 0.5 SNR = 0 db, 32 users throughput 10 0.4 N=1 N=2 N=3 N=6 N=12 N=16 0 20 40 60 80 100 120 140 160 gamma Patrick Svedman, Katie Wilson and Len Cimini 13 November 28, 2003
One User, on-off waterfilling Relative Doppler 1%, OFDM symbol with 256 tones, 4PSK 10 0 modulation Bits per Assigned Tone 10 1 10 2 1 2 3 4 5 6 7 8 9 10 SNR (db) Patrick Svedman, Katie Wilson and Len Cimini 14 November 28, 2003
One User, on-off waterfilling Relative Doppler 1%, OFDM symbol with 256 tones, 4PSK 10 3 4PSK on off water filling 4PSK all tones modulation Bits per OFDM symbol 10 2 10 1 10 0 1 2 3 4 5 6 7 8 9 10 SNR (db) Patrick Svedman, Katie Wilson and Len Cimini 15 November 28, 2003
One User, on-off waterfilling Relative Doppler 1%, OFDM symbol with 256 tones, 4PSK 1 0.9 modulation Fraction of symbols that are no good to user of interest 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1 2 3 4 5 6 7 8 9 10 SNR (db) Patrick Svedman, Katie Wilson and Len Cimini 16 November 28, 2003
Conclusions: Opportunistic beamforming with clustered OFDM promises throughput on par with a smart antenna solution, with little feedback and possibly a lower complexity basestation. Future work: Design of a fair scheduler Analysis of the impact of channel estimation and feedback error/delay More realistic simulations, especially regarding the channel models and more Patrick Svedman, Katie Wilson and Len Cimini 17 November 28, 2003