Full DimensionalDownlink Adaptive Beamforming using Sounding Signals Information

Full DimensionalDownlink Adaptive Beamforming using Sounding Signals Information

Ali A. Esswie , Abdurrahman Fouda Research and Development (R&D) department Intel Corporation, Intel WCL labs

Cairo, Egypt

AbstractWe propose downlink Multi-User MIMO channel state information reference signal (CSI-RS) beamforming scheme,of wireless Frequency Division Duplex (FDD) system with large number of transmit antennas at base station. The developed scheme combines Uplink channel sounding information to efficiently predict the downlink channel in the CSI-RS phase. Scheme is developed to consider the practical case of different user elevations (3D coverage) for maximum received per user throughput. Results show that proposed scheme drastically reduces the user feedback overhead bits than the current LTE-A 3GPP standard dual codebook with achieving enhanced performance. It also enables the freedom of increasing BS antennas with limited DL feedback overheads.

KeywordsLTE-A; Beamforming; CSI; MIMO; User Feedback overhead; SINR.

1. INTRODUCTION

   Future wireless communication systems will providea positive response to the increasing demand ofhigher data rates and larger user and traffic capacities. Major organizations in the wireless industry are working to define what futurerequirements will be. This happens through standard committees such as 3GPP and IEEE. Expansion to current wireless standards (e.g. spectrum) and breaking limitations of existing technologies (e.g. MIMO) are popular areas where efforts are dedicated. For example, Massive MIMO or Full Dimension MIMO (FD-MIMO), where a large number of antennas is employed at the eNB, is strongly developed to become rewarding [1].

   According to 3GPP standards, the future trends of research are contained mainly in user feedback overhead reduction and researching new techniques for maximum achieved throughput per user. In Frequency Division Duplex systems (FDD), users have to estimate DL channel through the transmitted reference signals in the DL path then feedback their serving BSs with standardized Channel Quality Indicator (CQI) and Precoding Matrix Indicator (PMI) feedback bits [2].When the number of antenna elements at the eNB increases, the overhead of DL reference signals and UE feedback will severely limit the system performance and exhaust DL & UL resources [3]. This will add limitations in front of the developments of future 5th Generation (5G) wireless systems. In addition, current and standard CSI-RS beamforming schemes such as the LTE-A dual codebook are only based on space division in the horizontal direction neglecting the elevation (Vertical) direction which will lead to suboptimal per user performance in such a practical case when users have different elevations.

   A common trend for feedback overhead reduction and performance maximization is to divide the precoding process into two stages, training CSI-RS phase and data transmission phase [4]. The first stage is duringthe training phase where DL channel is quantized at the UE side to choose the principal direction of the DL channel per UE. The second stage is during the data transmission phase where Zero Forcing (ZF)is used for interference cancellation

   Following this framework, several approacheshave been introduced and developedfor 3GPP LTE releases. For example, one approach utilizes adaptive codebooks wherean update to codebook base code words happensat the transmitter and receiver according to the channel temporalcorrelation changes [5, 6].Another approach isbased on channel reciprocity which is suboptimal for FDD systems due to the frequency gap between UL & DL channels [7]. In addition, Random beamforming (RBF), one more approach, has proven to be attractive in multi-antenna systems especially when employing large number of transmit antennas with large number of UEs [8, 9].

   In this paper, we propose a novel scheme for CSI-RS transmission phase. Scheme uses the Uplink sounding information to determine and predict the Downlink channel in the 3D space with different-elevation users. Once the principal paths of DL channel per user are identified, eNB distributes group of CSI-RS beams over different elevation angles to achieve maximum received Signal to Noise Ratio (SINR) for each user. The beam architecture through elevation directions is left arbitrary and general for any other developments and to consider other Performance Key Indicators (KPI) such as the scan loss of antennas. Scheme shows enhanced performance than the standard LTE-A codebook while drastically reducing the required user feedback bits.

   This paper is organized as follows: section II presents the test system model for simulations, section III discusses the new proposed 3D beamforming method in details & feedback overhead comparisons with current LTE-A dual codebook, section IV shows the performance results of the proposed scheme. Conclusion is drawn in last section.

2. SYSTEM MODELING

   We used a System Level Simulator (SLS) that considers a multi-antenna downlink MU-MIMO broadcast channel over OFDM system equipped with transmitting antennas at the

   eNB and receivers (UEs) equipped by receiving antennas each as described in Fig. 1 similar to [10].

   directions of the Uplink channel. In DL CSI-RS, eNB shoots 8 DL DFT beams as in Fig. 2 towards the directions of the captured UL directions from UL channel sounding training phase. Users will feedback only 3 bits of PMI feedback each Channel State Information (CSI) period to indicate the strongest serving beam in terms of received SINR maximization.

   AP – 8 Base Beams for 16*1 Antenna Configuration

   150

   120

   90 20

   60

   15

   10 30

   5

   180 0

   210 330

   Fig.1. System block diagram

   The received signal at the i-th UE, is described mathematically as in (1),

   240

   270

   300

   = + , = , , . . (1) where is × channel from the eNB to the i-th UE,

   is the transmitted signal vector and ~ CN (0, N0) is additive white Gaussian noise (AWGN).

   The transmitted signal vector is described as,

   = ()() (2)

   Where the transmitted encoded data are symbols and

   = [1() 2 ()] is the beamforming matrix. is number of independent streams per UE which we limit to unity ( = 1) in order to save UE feedback and decrease control signaling overhead. Total transmitted power is divided equally among co-scheduled UEs. A perfect channel measurements (H) is assumed.

3. PROPOSED CSI-RS BEAMFORMING METHOD

   1. Idea formulation

      In this section we describe our proposed Multi-User downlink transmission scheme and the developments to maximize the Multi-User gain to the maximum.

      According to the fact that Uplink and Downlink waves get bounced along the same set of clusters or paths to their destinations in low mobility environments, the major difference will be the carrier frequency or the frequency gap between Uplink and Downlink channels. Also sub path phases are assumed to be IID between Downlink and Uplink channel as in [11].

      Idea is to estimate the Downlink channel principal directions directly from Uplink sounding signals per user. At eNB, it receives the Uplink sounding signals from users, projects the Uplink energy over a set of K Uplink DFT beams that cover the whole cell directions and estimates strongest and principal

      Fig.2. 2D DL CSI-RS beamforming: 8 Beams for 16x1Tx

      Moreover, we developed the scheme to support beamforming in the vertical direction for best and maximum received INR for different-elevation users. In DL CSI-RS, the horizontal directions are determined from Uplink channel sounding and vertical directions are predefined over whole vertical coverage. Hence, dividing the vertical coverage is arbitrary and can hold up some interesting architectures following pre-set rules to optimally cover the vertical space as depicted in Fig. 3.

      Fig.3. 3D CSI-RS beam distribution over vertical coverage

   2. Comments on feedback overheads

      We followed the standard CSI periodicity of 5 ms in 3GPP (every 5 subframes). Developed CSI-RS scheme showed the need for smaller PMI feedback overheads since beams are shot adaptively in estimated proximity to users hence eNB doesnt have to shoot many beams in CSI-RS phase.

      In 3GPP standards, LTE-A 8-Tx codebook is divided into two separate codebooks (dual codebook) in order to divide the amount of required user feedback at each CSI instant. The UEs feedback two 4-bit indices (PMI1 and PMI2 ) to choose a

      Power Based on DL Channel

      Power Based on UL Channel (0MHz)

      Power Based on UL Channel (100MHz) Power Based on UL Channel (400MHz)

      precoder from each codebook (W and W ). The final precoder

      7000

      6000

      Power Projection along Spatial Directions

      1 2

      will be W1 × W2 . PMI1 Captures the long term channel variations and PMI2 captures the short term channel variations. A usual exercise of feedback overhead reduction is to increase user feedback PMI1 periodicity interval. However, the performance suffers significant degradation in such case. Hence, different frameworks of feedback are defined in 3GPP standards such that (20, 5) ms feedback framework (20 ms per PMI1 feedback and 5ms per PMI2 feedback).

      Proposed scheme requires smaller number of bits for user PMI feedback that are limited to 3 bits per CSI period since eNB shoots only 8 beams each CSIas in TABLE I. Searching a large number of DL CSI-RS beams as in the case of current LTE-A codebooks is replaced by adaptive codebook (set of DL CSI- RS beams) that are estimated and generated from users UL sounding.

      TABLE I. FEEDBACK OVERHEADS, PROPOSED SCHEME VS. CODEBOOK

      5000

      power received along the angle

      4000

      3000

      2000

      1000

      0

      0 20 40 60 80 100 120

      spatial angle from 0 to pi

      Fig. 4. UL & DL total principal paths diversion with Frequency gap

      200

      150

      100

      50

      	Proposed scheme, (5) ms	8-Tx CB, CSI: (5/5) ms	8-Tx CB, CSI: (20/5) ms
      PMI bits/ 5 CSI periods	15	40	28
      CQI bits/ 5 CSI periods	25	25	25
      Total bits	40	65	53

      0

      0 20 40 60 80 100 120

      Fig. 5. UL & DL principal paths diversion for single user

      Where

   3. Detailed algorithm

      1. Users send sounding signals over Uplink channel to serving BSs

      2. BSs extracts the strongest paths of Uplink channel using general energy projection mathematical process over pool of DFT beam directions

      3. Uplink and Downlink principal paths over all channel clusters get bounced similarly and strongest paths in received power diverge in degrees with increasing the frequency gap between uplink and downlink channel as in Fig. 4. In Fig. 5, a random single user is picked to show UL and DL strongest paths diversion for 100MHz frequency gap.

      4. In DL CSI-RS phase, BS shoots 8 beams that are distributed horizontally from Uplink channel directions after energy projection process and vertically as an arbitrary architecture as in Fig. 3. Beams are formulated according to:

         , ,

         = 2 ( sin cos + cos )/ (3)

         • m and n are the indices of antenna element horizontal and vertical placement in the array at BS.

         • is the elevation angle (Vertical) for each UE from the antennas planar axis

         • is the azimuth angle (Horizontal) for each UE from edge of coverage area

      5. Users feedback only 3 bits to choose a beam out of the shot 8 beams compared to 8 needed bits for the LTE-A dual codebook

      6. Zero Forcing beamforming is applied on top of the CSI- RS beamforming for interference cancellation in data transmission phase

4. SIMULATION RESULTS

   The major simulation parameters of our scenarios are listed in TABLE II. Generally, the 3GPP simulation methodology is followed.

   TABLEII. MAJOR SIMULATION PARAMETERS

   Parameter	Value
   Channel model	3GPP 3D model [11]
   Network	21 cells
   DL frequency	2GHz
   UL frequency	1.9GHz
   UE dropping	Uniform 2D, 10 UEs/Cell
   eNB antenna configurations	8
   eNB antenna polarization	ULA, 0.5
   eNB antenna pattern	3GPP
   UE antenna configuration	2
   UE antenna polarization	ULA, 0.5
   Rank per UE	1
   Channel measurement at UEs	Perfect
   CQI mode	Wideband
   PMI mode	Sub band
   CSI feedback delay	5 sub frames
   CSI feedback frequency	Every 5 sub frames
   Maxim number of HARQ Retransmissions	4
   Outer-loop link adaptation target packet error rate	10%
   MU-MIMO rank	8
   Scheduler	Proportional-Fair

   As in Fig. 6 and TABLEIII, proposed CSI RS scheme shows enhanced performance than the exhaustive and impractical scenario of the LTE-A codebook (CSI: [5/5] ms) and significantly improved performance than the practical scenario of the LTE-A codebook (CSI: [20/5] ms). In both cases, required user feedback bits are reduced as in TABLE I.

   Proposed scheme is designed to be full dimensional CSI-RS transmission scheme by considering the vertical coverage in beamforming. As mentioned earlier, eight DL horizontal directions per user are determined from the energy projection over UL channel phase and the vertical directions are defined in a way to optimally cover the vertical space in each CSI. The horizontal directions are distributed over predefined vertical directions to achieve full dimensional beamforming and best received SINR per user. This implementation showed enhanced performance than the conventional 2D techniques of beamforming as in Fig. 7.

   TABLEIII. PROPOSED SCHEME DEMONSTRATES HIGHER THROUGHPUT

   Fig. 6. Higher throughput is achieved by the new 2D method in comparison with LTE-A codebooks

   Fig. 7. Full dimensional proposed scheme demonstrates higher received SINR

5. CONCLUSION

In this paper, we introduced novel full dimensional beamforming method in the CSI-RS training phase for future wireless communication netwoks. Proposed scheme combines and uses the uplink sounding signals information per user to efficiently predict and estimate the Downlink channel. Full dimension beamforming is performed to optimally cover the 3D space and for maximum received throughput for practical different elevations users. Scheme showed lower user feedback overhead bits with maintaining better performance than the standard LTE-A dual codebook since efficient beamforming is performed and targeted directly to user positions.

REFERENCES

1. Y.-H. Nam, B. Loong , K. Sayana, Y. Li, and J. Lee, "Full-Dimension MIMO (FD-MIMO) for next generation cellular technology", IEEE Communications Magazine, Vol. 51, no. 6, pp. 172-179, Jun. 2013

2. O.Y. Bursalioglu, S. A. Ramprashad and H. C. Papadopoulos, "Towards improving LTE SU/MU-MIMO performance: Issues

3. H. Huh, A. M. Tulino and G. Caire, " Network MIMO with Linear Zero- Forcing Beamforming: Large system analysis, impact of channel estimation and reduced-complexity scheduling", IEEE Transactions on Information Theory, Vol. 58 , no. 5, pp. 2911 – 2934 , 2011.

4. J. Nam, J.-Y. Ahn, A. Adhikary, and G. Caire, "Joint spatial division and multiplexing: realizing Massive MIMO gains with limited channel state information", in proceedings of the 46th Annual Conference on Information Sciences and Systems (CISS), Princeton, NJ, USA, 2012.

5. K. Amiri, D. Shamsi, B. Aazhang and J. R. Cavallaro, "Adaptive Codebook for Beamforming in limited feedback MIMO systems", in proceedings of the 42nd Annual Conference on Information Sciences and Systems, CISS 2008, Princeton, NJ, USA, 2008, pp. 994-998.

6. J. Chang I.T. Lu and Y.X. Li, "Adaptive codebook-based channel prediction and interpolation for multiuser multiple-input multiple-output orthogonal frequency division multiplexing systems", IET Communications, Vol. 6, no. 3, pp. 281-288, April 2012.

7. Y. Han, J. Ni and G. du, "The potential approaches to achieve channel reciprocity in FDD system with frequency correction

8. D. Piazza and U. Spagnolini, "Random Beamforming for spatial multiplexing in Downlink multiuser MIMO systems", in proceedings of the 16th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC 2005), Berlin, 2005, Vol. 4, pp. 2161- 2165.

9. M. Kountouris, D. Gesbert, and T. Sälzer, "Enhanced Multiuser Random Beamforming: Dealing with the not so large number of users case", IEEE Journal on Selected Areas in Communications, Vol. 26 , no. 8, pp. 1536- 1545, Oct. 2008.

10. M. Esslaoui and M. Essaaidi, "An efficient user scheduling scheme for downlink Multiuser MIMO-OFDM systems with Block Diagonalization", International Journal of Advanced Computer Science and Applications, (IJACSA) Vol. 4, No. 5, pp. , 2013.

11. 3GPP TR 25.996, "3rd Generation Partnership Project; Technical Specification Group Radio Access Network; Spatial channel model for Multiple Input Multiple Output (MIMO) simulations (Release 11)", V11.0, 2012-09.