MIMO On-Frequency Repeater with Self-Interference Cancellation and Mitigation. Peter Larsson, Mikael Prytz Ericsson Research ADHOC 08, 7/5-2008

Transcription

1 MIMO On-Frequency Repeater with Self-Interference Cancellation and Mitigation Peter Larsson, Mikael Prytz Ericsson Research ADOC 8, 7/5-8

2 Outline Introduction MIMO On-Frequency Repeater Basic idea Design MIMO Self-Interference Cancellation MIMO Self-Interference Mitigation Performance potential Low rank channel assumption Impact of channel estimation error Conclusion Peter Larsson, Mikael Prytz 8-4-8

3 The communication problem and some options ow to improve cellular range-rate, system capacity, battery use time performance? Power Antennas Bandwidth Parallelism (e.g. MIMO) Channel gains Shorter com. distance Splitting the path in hops P GCh GAnt GAnt Γ = I + N Rate = f (Γ) G Ch ( D) G Ch α D α α ( D ) = G ( ) ( D ) D Ch Smaller cells DAS Split path in multiple hops 3 Peter Larsson, Mikael Prytz 8-4-8

4 Duplex loss in traditional relaying Alternating between transmission and reception Reduced capacity (duplex loss) TX Link Relay(s) Link RS Resource Resource RX(s) BS MS RS K Freq.. R s R s Time 4 Peter Larsson, Mikael Prytz 8-4-8

5 Repeaters (Amplify and Forward) FTR: Frequency translating repeater OFR: On-frequency repeater TX FTR BS Link Relay(s) Link Resource Resource RS RX(s) MS OFR Resource Resource RS BS MS 5 Peter Larsson, Mikael Prytz 8-4-8

6 OFR basics OFR RX and TX on same freq. Out-in isolation through donor + coverage antennas, directivity, separation, walls Isolation through Self-IC Self-IC 35 db extra isolation. 5 db backoff for lowripple 6 Peter Larsson, Mikael Prytz 8-4-8

7 WCDMA OFR SOTA WCDMA OFR Inter-chip-interference Complex receivers for good performance (nonrake) Direct signal notefficiently exploited (chips can not interfere constructively) Does not enable efficient cooperating RNs and BS (chips can not interfere constructively) BS.6 μs RS Delay = 5 μs Resource Resource Resource h r (n) h d (n) MS 7 Peter Larsson, Mikael Prytz 8-4-8

8 OFDM-OFR Transmitter Link Relay(s) Link Receiver(s) h (n)* CP s(n) CP RS Delay=T RS Resource Resource s(n) h (n)* h V (n)* CP CP s(n) s(n) BS MS RS V Delay=T RS OFDM inherently lend itself to OFR Sufficient long CP! T CP > TRS + TDS + TP Allows the opportunity for direct and relayed signal(s) to constructively interfere with each other, without ISI. 8 Peter Larsson, Mikael Prytz 8-4-8

9 More Antennas... Traditional (repeaters and) OFRs use single RX + TX antennas We consider Multi-antenna OFR Why? OFR OFR 9 Peter Larsson, Mikael Prytz 8-4-8

10 Antenna aspects: SM-MIMO Traditional cellular systems uses MIMO based comm. methods Traditional repeater collapses the channel rank to one Key-hole effect! Graceful rate degradation with distance from BS suggest to use MIMO end-to-end Channel capacity (MIMO) Repeater position Desired (MIMO) Un-desired (SISO) Range from BS TX MIMO -OFR RX Peter Larsson, Mikael Prytz 8-4-8

11 Antenna aspects: Interf. cancelation Problem Interference caused by other BSs, MSs, RSs should not be forwarded Solution Use multiple antennas for interference cancelation Issues Interf. Frequency selective processing may be needed! OFR * Evidently, all antenna schemes, (MIMO,IC ) eats of the same cake. Tradeoff! Peter Larsson, Mikael Prytz 8-4-8

12 Antenna aspects: Interf. mitigation Problem Interference caused to other BSs, MSs, RSs should be avoided Solution Use multiple antennas for interference mitigation to exposed (fixed) nodes Issues Frequency selective processing OFR Interf. Peter Larsson, Mikael Prytz 8-4-8

13 Self-interference cancellation MIMO-OFR s vector signal (SM-MIMO) = Repeater output to input matrix channel W is a matrix selected for self-interference cancellation W=A C s - Σ A Σ n W C 3 Peter Larsson, Mikael Prytz 8-4-8

14 Peter Larsson, Mikael Prytz Self-interference mitigation MIMO-OFR Problem Isolation insufficient? Constraints / Objectives Keep ETE channel rank Adapt A and C based on and Assumption has low rank Solutions Adapt A and C to reduce feedback Null space projection ( ) ( ) ( ) ( ) N n N n svd : :, : :, ],, [ U A V C V U S = = = = σ [ ] [ ]= v v v v v u u u u u σ

15 TX Considered repeater and relay cases -Phase OFR -Phase AF -Phase DF Resource RS RX TX Res. Res. Res. Res. Res. Res. r = Ax + (D) Bw A = + B = P RX () () () [ G I] C = log R x P = tr N G () ( ) + Nσ w det( I + ( AR x A (D) () w w = w G = I )( BR w B (, rx) (, rs) ) P P N RX P E{ xx = } = I = E{ ww } = σ I C = E{ C} N R w ) w r = Ax + Bw (D) A = () G I = RS RX det( ( )( ) I + AR xa BR wb C = log B () P RX R x () G P = tr N C = E{ C} () I ( ) + Nσ w w w = w w G = I (, rx) (, rs) (, rx) P P N RX P E{ xx = } = I R w = E{ ww } = σ I N (D) () w ) TX () r = x + w C R x log () () RX RS Res. () () = det( ( ) I + R x R w C = E{ C} Assuming that: () P E{ xx = } = I = E{ ww } = σ I = Ν, G ) and = Ν, G ) P. erhold, E. Zimmermann, G. Fettweis, On the performance of cooperative amplify-and-forward relay networks, in: 5th Int. ITG Conf. on Source and Channel Coding, Erlangen, Germany, 4. ( N R w () ( ) w 5 Peter Larsson, Mikael Prytz 8-4-8

16 Performance potential N= No direct signal -phase OFR -phase AF with comb. -phase DF Direct signal = G - db phase OFR -phase AF with comb. -phase DF Ergodic channel capacity [b/z/s].5.5 Ergodic channel capacity [b/z/s] N=4 Ergodic channel capacity [b/z/s] 5 5 SNR [db] phase OFR -phase AF with comb. -phase DF 5 5 SNR [db] Ergodic channel capacity [b/z/s] 5 5 SNR [db] phase OFR -phase AF with comb. -phase DF 5 5 SNR [db] 6 Peter Larsson, Mikael Prytz 8-4-8

17 Outage probability N= P(C < B/z/s) No direct signal Direct signal = G - db P(C < B/z/s) Outage probability - Outage probability - N=4 -phase OFR -phase AF with comb. -phase DF SNR [db] P(C < 4 B/z/s) - -phase OFR -phase AF with comb. -phase DF SNR [db] P(C < 4 B/z/s) - Outage probability - Outage probability phase OFR -phase AF with comb. -phase DF SNR [db] -phase OFR -phase AF with comb. -phase DF SNR [db] 7 Peter Larsson, Mikael Prytz 8-4-8

18 Low rank channel assumption? A typical repeater feedback channel Eigenvalues vs. Distance D (d=λ/, λ=5 cm, N=4) TX MIMO Repeater Feedback through a reflex D RX Eigenvalues Identical to LoS-MIMO channel LoS-MIMO channel can be modeled with DFT matrix with jπ dv entries x = e λ D - Distance [m] Channel rank at D > m P. Larsson, "Lattice array receiver and sender for spatially orthonormal MIMO communication," in Proceedings of the IEEE 6st Vehicular Technology Conference (VTC '5), vol., pp. 9-96, Stockholm, Sweden, May 5. 8 Peter Larsson, Mikael Prytz 8-4-8

19 Channel estimation error impact Additional isolation thanks to the null space projection method Impact of channel estimation errors? Assumption Complex Gaussian channel estimation error = LoS-MIMO channel at D = m ~ ~ ~ [ U, S, V ] = svd( + Error ) ~ ~ U V eff = Max eigenvalue determine stability of feedback system I = log max( abs( eig( eff ))) Addtional repeater output-to-input isolation [db] log Error 9 Peter Larsson, Mikael Prytz 8-4-8

20 Summary & Conclusions Summary MIMO OFR concept presented -phase (no duplex loss) MIMO support (keeps the channel rank) Self-interference cancellation idea extended to multiple antenna (MIMO) repeater New method for increased output to input isolation through use of multiple antennas and beamforming Conclusions Performance potential promising at high SNRs Potentially useful at low SNRs due to low complexity Appears robust under influence of channel estimation errors Peter Larsson, Mikael Prytz 8-4-8

21 Peter Larsson, Mikael Prytz 8-4-8

22 MIMO On-Frequency Repeater with Self-Interference Cancellation and Mitigation Peter Larsson, Mikael Prytz Ericsson Research Abstract - In this paper, we consider the idea of a MIMO On- Frequency Repeater where self interference cancellation and beamforming based interference mitigation methods are exploited for increased repeater output to input isolation, and increased ETE performance. I. Introduction Future wireless communication systems require enhanced range-rate performance in a cost efficient manner. Splitting a radio communication path in two or more hops gives significantly increased path gain for each hop due to the power law path loss characteristic. Various two-hop methods have been analyzed, for example, cooperative relaying [], cooperative diversity [], virtual antenna arrays [3], -hop relaying [4], and repeaters. Repeaters could be attractive due to their typically lower processing complexity and W requirements relative to other two-hop schemes, and the short forwarding delay. A classical repeater, like many other -hop methods, is characterized by a so called duplex loss that affects the ETE spectrum efficiency adversely. For the repeater, this classically corresponds to receiving on one frequency and transmitting on another. owever, a recent trend among repeater designers is to use on-frequency repeaters (OFR). Sending and receiving concurrently on the same frequency is possible provided the output-to-input repeater antenna port isolation is larger than the repeater input-to-output gain. Typically the isolation margin needs to be 5 db above the repeater gain to mitigate oscillatory behavior. Various approaches to ensure high isolation include physical separation (by location or walls) of sender and receiver repeater antennas, different receive transmit antenna polarizations, and the more recently used method of selfinterference cancellation (self-ic). Another trend in the communication research and industry communities is the use of multiple antennas for more efficient communication, e.g. spatial multiplexing multiple input multiple output (SM-MIMO) [6], which has emerged as a key-component in today s wireless communication system. II. Proposed methods Since both OFR and SM-MIMO promise enhanced data rates it is interesting to explore a joint design. This is further motivated by considering today s state of the art repeaters, which are of the SISO type, in conjunction with SM-MIMO: when the direct signal is to weak a SISO repeater would act as a keyhole in the ETE channel, i.e., the maximum number of MIMO subchannels, or equivalently the ETE channel rank, becomes one despite multiple antennas at sender and receiver. In this setup the notion of SM is lost entirely. Another issue is that OFRs are practically constrained in amplification gain due to the limited isolation that can be TX Resource MIMO Resource OFR Output-to-input antenna feedback Fig.. MIMO On-Frequency Repeater System S U U U U Matrix dimension min {N,N } min{n,n } N N U Σ Σ A A Σ - W B D B Fig.. MIMO On-Frequency Repeater D C C C Repeater RX Matrix dimension min {N,N } min {N,N } achieved. This in turn has an impact on the maximum achievable communication range. Now, MIMO uses multiple antennas at the repeater input and output which can be exploited for RX and TX beamforming at the repeater to improve isolation and hence also amplification gain. Fig. shows a generic frequency-domain version of a MIMO-OFR with matrix-based self-ic. The vector signal s, where s is the transmitted signal, is received at the repeater input experiencing additive noise W. Internally, the vector signal can be tuned by adjusting matrices A, B, C, and D, and the signal is further affected by matrices A, B, C, D, which represent (minor) internal channel alterations primarily due to delays. As seen in Fig. the output signal U couples to the input via the external feedback channel state matrix. To avoid oscillations, B is adapted to cancel the effective feedback signal A A U. In particular, for optimum self-ic a necessary condition is: B = A C () B A C C Another adaptation aspect is to select A and C such that good input-to-output isolation is achieved for all spatially multiplexed signals. In doing so a necessary condition is that the input-to-output rank in the repeater is greater than or equal to the number of spatially multiplexed streams. When has low rank, which can be assured by an appropriate antenna array design, A and C can easily be selected to exploit this property. One approach is to project as much as possible of the repeater output spatially multiplexed signals into the null space of A C. Similarly the repeater input can suppress feedback signal from the output by selecting receive weight in the null space of A C. Practically, this is implemented by a singular value decomposition approach where the null space is considered to correspond

23 to the N-n smallest singular values where n is selected such that the N-n values are L db smaller than the strongest singular value, for a certain limit value L. ence, the SVD operation and selection of A and C are, V ] = svd ( V( :, n : N ) ( U( :, n : N )) [ U, S AC C = A = While multiple antennas can be used for SM-MIMO and self-interference mitigation, it is further possible to exploit some of the available degrees of freedom to suppress undesired interference and enhance the ETE performance. The proposed MIMO-OFR methods are further complemented with OFDM where a sufficiently large cyclic prefix is selected to allow for relayed signals to arrive nearconcurrently, hence absorbing any OFDM symbol ISI caused by path propagation delay differences for repeated and directed signal and allowing for potential constructive interference and channel rank increase. III. Results The Ergodic ETE channel capacity is ) () where is the aggregate ETE channel, γ is the SNR, and N is the number of TX antennas. The final submission will present results for: i) ETE MIMO-OFR channel capacity compared with combinations of non-ofr and SISO operation. ii) MIMO-OFR output-to-input isolation performance vs. repeater output-to-input channel estimation error. iii) ETE channel capacity impact vs. isolation. IV. References [] A Sendonaris, Advanced Techniques for Next-Generation Wireless Systems, Ph.D. Thesis, August 999 [] J. N. Laneman, Cooperative Diversity in Wireless Networks: Algorithms and Architectures, Ph.D. dissertation, Massachusetts Institute of Technology, Cambridge, MA, Aug.. [3] M. Dohler, Virtual Antenna Arrays, WWRF, electr. Conference CD-ROM, Paris, France, Dec. [4] IST WINNER II, D3.5. v., Relaying concepts and supporting actions in the context of CGs [5] J. Chun, J. Lee, P. Choi, J. C. Yun, S. J. Lee, J.. Lee, Smart antennas for the on-air on-frequency repeater in the 3G mobile communication applications, pp [6] Gerard J. Foschini (autumn 996). "Layered space-time architecture for wireless communications in a fading environment when using multi-element antennas". Bell Labs Technical Journal : γ C = E log I + N, (3)