Carrier Aggregation and MU-MIMO: outcomes from SAMURAI project Presented by Florian Kaltenberger Swisscom workshop 29.5.2012 Eurecom, Sophia-Antipolis, France
Outline Motivation The SAMURAI project Overview on the investigated techniques Multi-User MIMO Carrier Aggregation Into the Real World Conclusion 28/05/2012 - p 2
Motivation Exponential increase in mobile data traffic LTE-Advanced promises downlink of 1Gbps low mobility and 100Mbps in high mobility conditions Practical implementations are still far from theoretical limit 28/05/2012 - p 3
The SAMURAI Project Spectrum Aggregation and Multi-User MIMO: ReAl-world Impact Two main research lines: Increase in spectral efficiency: MU-MIMO Spectrum exploitation: Carrier Aggregation (CA) Industrial feasibility as main goal: MU-MIMO and CA PHY Proof-of-Concept (PoC) Autonomous Component Carrier Selection (ACCS) PoC 28/05/2012 - p 4
MU-MIMO in LTE Systems Shared resources 2, 4 and 8 antenna supported No practical MU-MIMO deployments yet!. SU-MIMO feedback CQI/PMI/RI residual interference CQI/PMI/RI Interference-aware receiver is essential for robust MU-MIMO performance. 2 or 4 UEs supported Type and accuracy of feedback information are crucial for the reliability of MU-MIMO systems LTE systems. 28/05/2012 - p 5
MU-MIMO in LTE Systems LTE Release 8: Transmission mode 5: fixed codebook precoding, wideband PMI feedback Limited codebook size: 4 for 2 TX antennas, 16 for 4 TX antennas One layer per user Large residual multi-user interference LTE Release 9&10: Transmission modes 8 and 9: user-specific precoding using DM-RS higher granularity feedback Flexible codebooks New DCI format supports transparent switching between SU and MU-MIMO (Rel10) Multi-user interference remains an issue -> need for advanced interference aware receiver design! DM-RS = downlink demodulation reference symbol; 28/05/2012 - p 6 DCI = downlink control information
SAMURAI Interference aware receiver Approximate max-log MAP receiver Based on matched filter outputs, no division operations Inherently exploits the structure of the interference instead of Gaussian assumption Applicable to single-antenna UEs as well Implemented in fixed-point C Jonathan Duplicy, Biljana Badic, Rajarajan Balraj, Rizwan Ghaffar, Péter Horváth, Florian Kaltenberger, Raymond Knopp, István Z. Kovács, Hung T. Nguyen, Deepaknath Tandur and Guillaume Vivier, "MU-MIMO in LTE Systems", EURASIP Journal on Wireless Communications and Networking, vol. 2011, Article ID 496763, 13 pages, 2011. doi:10.1155/2011/496763. Ghaffar, Rizwan; Knopp, Raymond, "Interference-aware receiver structure for Multi-User MIMO and LTE", EURASIP Journal on Wireless Communications and Networking, Volume 2011: 40. 28/05/2012 - p 7
Simulation based investigation 10 0 SCM-B 4 2, 3km/h, ideal channel estimation, 8 TTI FD CQI7 CQI13 Interference unaware receiver unreliable in MU- MIMO transmission. BLER 10-1 CQI10 With SAMURAI IA Receiver MU-MIMO works even for high modulation order!. CQI4 MF IRC SAMURAI IA Receiver 10-2 -4-2 0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 SNR / db 28/05/2012 - p 8
Complexity Performance Trade-Off SAMURAI IA Receiver: optimum MU-MIMO performance and low 28deployment costs! SNR / db @ BLER = 10-2 24 20 16 12 8 4 0 CQI4 CQI7 CQI10 CQI13 poor performance-complexity trade-offs for conventional MF and Max Log MAP. MF IRC SAMURAI IA Receiver Max Log MAP 10 2 10 3 10 4 10 5 Number of real-valued multipcation for LLR per subcarrier 28/05/2012 - p 9
MU-MIMO CQI Feedback Issue: The same SU-MIMO feedback used for MU-MIMO -> low overhead but limits MU-MIMO gains SAMURAI approach: apply interference aware receiver, improve post-processing SINR and report accurate CQI -> multi-user interference considered in CQI calculation 15 ideal MU-CQI feedback MU interference in CQI feedback MU-MIMO with SU-MIMO CQI feedback Throughput[Mbits/s] 10 5 Only applicable in Rel.9&10 due to the user specific signaling (DMRS) To achieve high performance MU-MIMO feedback required!. 0 0 2 4 6 8 10 12 14 16 18 20 22 24 SNR[dB] SU-MIMO feedback limits MU-MIMO gains! 28/05/2012 - p 10
MU-MIMO proof of concept Goal: demonstrate the feasibility of MU-MIMO using advanced receiver techniques Platform: Eurecom OpenAirInterface Software defined radio Open-source implementation of LTE Rel. 8.6 UE: CBMIMO1 card with antennas enb: Express MIMO card with amplifiers and antennas 28/05/2012 - p 11
MU-MIMO proof of concept Use of LTE Rel. 8 transmission mode 5 Use of feedback mode 1-2 instead of 3-1 for exploiting sub-band Precoding Matrix Indication (PMI) information Sub-band CQI and wideband PMI not enough for exploiting the full potential of the designed receivers Use of custom Downlink Control Information (DCI) Signals UE to use the precoder(s) according to subband PMI(s) indicated in the latest feedback report Use of SAMURAI Interference Aware Receiver Use of MU-MIMO scheduler that switches dynamically between MU-MIMO and SU-MIMO depending upon number of orthogonal subbands 28/05/2012 - p 12
MU-MIMO PoC: First Results Link-level performance comparison of TM 1, 5, and modified TM 5 with IA receiver on OpenAirInterface platform (software simulation) 10 0 SCM-C channel, MCS 9 BLER 10-1 10-2 0 5 10 15 20 SNR N TX =2, N RX =1, 25 PRBs TM 1 TM 5 TM 5 with IA receiver 5dB gain at 10% BLER with IA receiver 28/05/2012 - p 13
MU-MIMO PoC: First Results System-level performance comparison of TM 2, 5, and modified TM 5 with IA receiver on OpenAirInterface platform (software simulation) Average System Throughput (Kbps) 800 700 600 500 400 300 200 100 Fig 2. AVERAGE SYSTEM THROUGHPUT FOR DIFFERENT SCENARIOS 1. TM2, 2 USERS, 8-TAP RAYLEIGH CHANNELS 2. TM5, 2 USERS, 8-TAP RAYLEIGH CHANNELS 3. TM5, 3 USERS, 8-TAP RAYLEIGH CHANNELS 4. TM5, 4 USERS, 8-TAP RAYLEIGH CHANNELS 5. TM5, 5 USERS, 8-TAP RAYLEIGH CHANNELS 6. TM5, 2 USERS, ORTHOGONAL RAYLEIGH CHANNELS With high number of users, the scheduler achieves the best case even with real channel conditions 0 N TX =2, N RX =1, 25 PRBs 1 2 3 4 5 6 SCENARIOS 28/05/2012 - p 14
MU-MIMO PoC: Ongoing work Real-time performance evaluation on Express MIMO boards and Agilent PXB channel emulator ongoing 28/05/2012 - p 15
Carrier Aggregation LTE Rel-10 allows aggregation of up to 5 20MHz component carriers (CC) Each CC appears as a Rel-8 serving cell to Rel- 8 UEs Synchronization and reference signals System Information Backwards compatible bandwidths Data aggregation in MAC layer Separate HARQ processes and feedback Individual transmission modes (modulation and coding) 28/05/2012 - p 16
CA Challenges Transceiver design challenge 3GPP specifies >30 operating bands, 6 bandwidths, and up to 5 CCs Contiguous vs. non-contiguous Intra-band vs. inter-band very high number of combinations need to be supported Network management challenge How to exploit multiple carriers adaptively and autonomously in self organized networks Load balancing and scheduling for multiple CCs 28/05/2012 - p 17
CA with OpenAirInterface Express MIMO with LIME RF frontend supports up to 4 CCs with up to 20MHz bandwidth 300MHz-3.9GHz tuning range per CC Inter- or intra-band Contiguous and non-contiguous Already demonstrated 2 x 5 MHz CCs Agilent Signal Generator Lime-based RF Express-MIMO OpenAirInterface 28/05/2012 - p 18
CA on OAI: ongoing and future work Integration of RRC signaling Scheduling and load balancing for multiple CCs CA for Inter-cell interference coordination (ICIC) Autonomous Component Carrier Selection in a network of home enbs (femto cells) Cross-carrier scheduling for ICIC in heterogeneous networks 28/05/2012 - p 19
Conclusions To exploit MU-MIMO gains in LTE and LTE- Advanced, we need Interference aware receivers Sub-band feedback Adaptive scheduling Carrier aggregation is a big challenge On the PHY level On the network level OpenAirInterface experimental platform used to make our research more credible 28/05/2012 - p 20