Full-Dimension MIMO: Status and Challenges in Design and Implementation

Gary Xu, Yang Li, Young-Han Nam and Charlie Zhang

Samsung Research America (Dallas)

Taeyoung Kim and Ji-Yun Seol DMC R&D Center,

Samsung Electronics Co., Ltd.

May 27, 2014

1

Outline

Current Status of FD-MIMO 1

Challenges of FD-MIMO 2

2

Background of Full-Dimension MIMO • Theory Behind: Massive MIMO*

– Spatial resolution increases as number of eNB antennas

– Narrow beam transmission with little MU interference

*Marzetta, “Non-cooperative cellular wireless with unlimited numbers of base station antennas,” IEEE TWireless Nov. 2010 3

• Active Antenna Array (AAA)

– 2D vs. 1D AAA

Full-Dimension MIMO (FD-MIMO)

4

Elevation

beamforming

Azimuth

beamforming

FD-MIMO simultaneously

supports elevation & azimuth

beamforming and > 10 UEs

MU-MIMO FD-MIMO eNB

Rel-10 FD-MIMO 32-64 Tx

FD-MIMO 100 Tx

Cap

acit

y

3-5x

10x

MU-MIMO with 10s of UEs

2-dimentional AAA & ~100 antennas

3D-spatial channel model

λ=12cm

@ 2.5GHz

Eg. 1: 8x8 array with full digital

beamforming across 64 elements

λ/2

λ/2

0.5m

0.5m

λ/2

λ/2

2λ

0.5m

1m

FD-MIMO 2D AAS Form Factor Examples

Eg.2: 8x8 array with full digital

beamforming across 64 elements

Eg. 1: 8x4 array with each element

4 antennas with analog beamforming

Urban Macro

Urban Micro

0.25m

λ/2

λ/2

0.25m

Eg.3: 8x8 array with cross-pol.

Digital beamforming 64 elements.

FD-MIMO antenna panel form factor is well within practical range

Eg.4: 1x8 array.

Digital elevation

beamforming

λ/2

Small Cell

0.5m

Industry Status and 3GPP roadmap

6

2014 2015 2016 2013

Start 3D channel model study item (SI)

Complete 3D channel model (SI)

Complete channel& baseline calibration

Expected start of Elevation Beamforming (EB)/FD-MIMO SI

EB/FD-MIMO SI Completion

Start EB/FD-MIMO work item (WI)

WID Complete (Dec 2016)

1st FD-MIMO Prototype: 32 antenna LTE base-station

PoC for Small Cell

3GPP development

PoC for Macro Cell

3-Dimension (3D) Channel Model

7

In SCM, channel is a composite response of cluster/subclusters to Tx/Rx antennas:

• Number of cluster/subclusters

• Delay of clusters

• Power of clusters

• Phases (due to e.g. reflection)

• Angle of Departure/Arrival (AoD/AoA)

Note:

AoA/AoD critically determines channel correlations Spatial channel model (SCM)

2D model assumes all clusters zero elevation angles and cannot describe elevation differences. 3D model captures elevation angles and thus clusters can be distinguished in elevation domain

0 5 10 15 20 25 30 35 40 45 500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Channel condition number (dB)cdf

Urban Macro

Urban Micro

Statistics in 3GPP 3D Channel Model

8

70 75 80 85 90 95 100 105 110 115 1200

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

cdf

Elevation angle of depature (degree)

Urban Macro

Urban Micro

• Elevation angle (w.r.t. zenith) has a range of 30-deg for UMa and 50-deg for UMi

• 80% channels have condition number > 5dB

Note: see “3GPP TR 36.873” for more details of 3D channel model and UE distribution.

2 Rx antennas (+ pol.)

System-Level Simulator (SLS) Evaluation

9

Simulation Setup:

• 3D ITU, UMa

• 57 sectors with K=10 UEs per sector

• Center frequency 2GHz, bandwidth 10MHz

• UE speed 3km/h or 30km/h, uniformly distributed

• 40 drops, 4s per drop

• UE: 2 Rx, 1Tx

Overhead: 20%

Ideal SRS estimation

4 ms scheduling delay

Normalized by # of DL subframes

Baseline: SU-MIMO with rank1

0

0.04

0.08

0.12

0.16

0.2

4x2 32x2: 2 UE 32x2: 4 UE

0.045 0.053

0.17

0

0.1

0.2

0.3

4x2 32x2: 2 UE 32x2: 4 UE

0.082

0.163

0.298

SLS Simulation Results (Up to 4 UE MU-MIMO)

0

2

4

6

8

10

4x2 32x2: 2 UE 32x2: 4 UE

2.24

5.09

8.04 127%

58%

99%

83%

0

1

2

3

4

5

6

7

4x2 32x2: 2 UE 32x2: 4 UE

1.95

3.73

6.26 91%

68%

18%

220%

3 km/h

30 km/h

Average cell throughput (bps/Hz) Cell-edge throughput (bps/Hz)

Average cell throughput (bps/Hz) Cell-edge throughput (bps/Hz)

10

4Tx 32Tx 32Tx 4Tx 32Tx 32Tx

4Tx 32Tx 32Tx 4Tx 32Tx 32Tx

Outline

Current Status of FD-MIMO 1

Challenges of FD-MIMO 2

11

FD-MIMO Framework in LTE/LTE-A

12

Antenna Virtualization & CQI Prediction

Cell-wide beamforming

by antenna virtualization

UE-specific beamforming

Issue: (1) How to generate wide beam from a large array

(2) CQI (channel quality indicator) mismatch

• Wide-beam (ant. virtualization) for control signal (coverage)

• Narrow-beam (precoding) for data signal

• UE CQI* is measured based on wide-beam *CQI is a UE feedback value and is essential for eNB to decide transmission scheme, code rate, modulation for each UE. 13

-180

-120

-60

0

60

120

180

-180-120-60060120180

270

300

240

330

210

0

180

30

150

60

120

90

0

2

4

6

8

10

12

Antenna Virtualization & CQI Prediction (2)

Mean of error: 0.0941

Var. of error: 0.0424

𝜌𝑝𝑟𝑒𝑑𝑖𝑐𝑡 =|𝒉𝑘𝒘𝑘|2

|𝒉𝑘𝒘0|2 𝜌𝑚𝑒𝑎𝑠𝑢𝑟𝑒𝑑

14

Synthesized antenna virtual pattern (32 ant.) CQI prediction

FD-MIMO in TDD: Antenna Calibration

15

PA

LNA eNB

transceiver

1

PA

LNA

Ant 1

PA

LNA

Ant M

𝑡1

𝑟1

𝑡2

𝑟2

𝑡𝑀

𝑟𝑀

eNB

transceiver

2

eNB

transceiver

M

…

…

Ant 2 H Reciprocal

Uplink sounding

Downlink transmission

𝑡1 =•••= 𝑡𝑀,𝑟1 =•••= 𝑟M

𝑡1 − 𝑟1 =•••= 𝑡 𝑀

− 𝑟 M

Joint Tx/Rx calibration

Independent Tx/Rx calibration

Challenges:

• Inherent error in calibration circuit

• Complexity grows with antennas

• Prefer independent Tx/Rx calibration

Calibration requirement:

Front-haul Complexity

Possible Solutions:

• Front-haul (CPRI) compression

• New baseband architectures 16

CPRI

Number of sectors 3

System bandwidth (MHz) 20

Sampling rate (Msps) 30.72

Bit width per I/Q-branch 16

Number of TX antennas (paths) 32

CPRI throughput (Gbps) ~96

FD-MIMO in FDD: Exploit Uplink Correlation

CSI acquired by training & feedback in FDD LTE/LTE-A

Pilot & feedback bits proportional to # of Tx antennas

Issue: CSI (channel state information) acquisition

Possible to use uplink channel for downlink precoding?

Uplink

Downlink

Uplink & downlink channels are correlated

FD-MIMO can measure uplink better

*Sana Salous and Hulya Gokalp, “Medium- and Large-Scale Characterization of

UMTS-Allocated Frequency Division Duplex Channels”, IEEE TVT.

17

FD-MIMO in FDD: Exploit Uplink Correlation (2) • Duplex distance: 45 MHz. Channel condition: NLOS.

Downlink: 2300 MHz; uplink: 2250MHz

0 100 200 300 400 500 600 700 800 900 10005

10

15

20

25

30

35

40

45

50

Block Index

Angle

betw

een D

ow

nlink a

nd U

plink B

eam

s

WCS Band

𝐏𝐫 𝐀𝐧𝐠𝐥𝐞 < 𝟑𝟎𝟎 = 𝟗𝟗. 𝟔% 𝐂𝐨𝐫𝐫𝐞𝐥𝐚𝐭𝐢𝐨𝐧 𝐂𝐨𝐞𝐟𝐟𝐢𝐜𝐢𝐞𝐧𝐭 = 0.995

PMI*: Highly correlated CQI: Highly correlated

Angle between eigenvector of downlink & uplink

*PMI (Precoding Matrix Indicator): quantized channel direction. 18

Other Challenges in FD-MIMO

• How to reduce overhead by exploiting channel correlation in azimuth and elevation domain?

• Possible to combine with uplink measurement to provide better accuracy?

Feedback and codebook design in FDD

• How to accurately estimate a large number of channels?

• How to reduce channel estimation complexity?

Uplink sounding in TDD

• How to optimally schedule ~10 MU-MIMO UEs without exponentially increasing complexity?

Scheduling & precoding complexity

19

Summary

20

Elevation

beamforming

Azimuth

beamforming

FD-MIMO simultaneously supports

elevation & azimuth beamforming and

> 10 UEs MU-MIMO

FD-MIMO eNB

• Full-dimension MIMO is a promising technology to

significantly improve cellular capacity (by x3-5)

• Challenges ahead include system design and

implementation

• “The Next Big Thing is Here” in wireless industry

THANK YOU!

21