Post on 23-Aug-2020
Ali Sadri, PhD
Sr. Director
Intel Corporation
iCDG - Intel Communication and Devices Group 2
Past and Future Capacity Improvement
Air Interface
Available Spectrum
Small Cell
Air Interface
Available Spectrum
Small Cell
Interference Mitigation, Full DuplexNew Waveform, MU-MIMOBeamforming, etc
Licensed, Unlicensed, SharedmmWave
DensificationRelayEdge CloudMesh BackhaulFronthaul
3G-4G
4G-5G
iCDG - Intel Communication and Devices Group 3
Search for Alternate Spectrum
10 20 30 40 50 70-80
NoMobile
AllocationIn Region
1 & 2
<1 GHz <4 GHz <4 GHz 7 GHz
Glo
ba
l MS
<3 GHz
Glo
ba
l MS
Glo
ba
l MS
Glo
ba
l MS
24
.25
25
.25
27
31
38
.6
42
.5
47
.2
50
.2
57
64
5+5 GHz
Glo
ba
l MS
60 GHz Band
Unlicensed
40 GHz Band
Licensed
LMDS Band
Licensed
24 GHz Band
Licensed
50 GHz Band
Licensed
70-80 GHz Bands
Minimal Licensed
3
CurrentIMT
bands
1
iCDG - Intel Communication and Devices Group 4
Reuse mmWave Knowledge
Legacy Bands
< 3.8 GHz
New Bands
< 6 GHz
Bands > 6
GHz (+mmWAVE)
Licensed Unlicensed Licensed Unlicensed
28
GHz
39
GHz
45
GHz
70-90
GHz
60
GHzLAA
6-24
GHzWiFi
Licensed Unlicensed
WiFi
* Categorized based on channel models and path loss
** Potentially the same technology elements could be used across a range of frequencies
iCDG - Intel Communication and Devices Group 5
mmWave Path-Loss Comparisons
2.3 GHz
28 GHz38 GHz
73 GHz
Oxygen
Absorbance
iCDG - Intel Communication and Devices Group
eNBAggregation
6
HetNet with mmWave Capable Small Cells (MCSC)
eNB eNB
MCSC
MCSCMCSC
MCSC
MCSC
MCSC
28 ,39 or 60 GHz
iCDG - Intel Communication and Devices Group 7
Network Densification Topology
Fiber Node
Distribution Node
Access Node
iCDG - Intel Communication and Devices Group 8
High Frequency Beam Forming
iCDG - Intel Communication and Devices Group 9
Challenges in mmWave Systems Design
•Higher Path Loss• To compensate with the high path loss higher gain antenna and/or
higher transmit power is required
• EIRP, TX power and RF exposure limit are regulated
•Massive MIMO is required for high gain antennas• Transmission becomes highly directional
• With Narrow beams, tracking of the UE becomes challenging
• Feed line loss• Diminishing return occurs as size of array increases
• Transmission loss increases as function of frequency
iCDG - Intel Communication and Devices Group 10
Challenges in RF & Antenna
• Feed line loss: (8-by-8) elements
Antenna spacing: λ
2=
𝑐/𝑓
2=
5𝑚𝑚
2= 2.5mm
60 GHz
Antenna spacing: λ
2=
𝑐/𝑓
2=
7.69𝑚𝑚
2
@ 28 GHz is 5.36mm and @ 39 GHz is 3.85mm
From 60 GHz to 28 GHz (or 38 GHz),
• The required area getting bigger then feed line getting longer (roughly double).
• Feed loss is also a function of frequency (higher loss at 60 GHz)
28 or 39 GHz
iCDG - Intel Communication and Devices Group
Modular RFEM Configurations
Antenna Side Shield Side
60GHz Operation
16 Elements
25.2 mm x 9.8 mm
64 elements 128 elements32 elements16 elements 128 elements
iCDG - Intel Communication and Devices Group 12
MAA POC Evolution
G3SIndoor
G3Indoor
G3MIndoor
G1Indoor
G2Indoor
• Discrete• MAA 128 (2x4)• Maple-M & R• EIRP ~ 43 dBm• 300 x 220 x 150
• Partial PCB• MAA 128 (1x8)• Maple-M & R• EIRP ~ 43 dBm• 160 x 140 x 110
• Stack up PCB• MAA 128 (1x8)• Maple-M & R• EIRP ~ 43 dBm• 160 x 140 x 110
• Stack up PCB• MAA 128 (2x4)• Maple-M & R• EIRP ~ 43 dBm• Reduce Side lobe
• Stack up PCB• MAA 128 (2x4)• Maple-M • MAA-RFEM• EIRP ~ 48 dBm• 160 x 140 x 110
G4Outdoor
• 190 x 170 x 140
iCDG - Intel Communication and Devices Group
Hardware Overview
• Indoor Design
• Easy access to ports
• Easy signal breakout for chamber tests
• Easy tabletop, tripod, post, ceiling installation
• Antenna Array
• 128 elements - 8x16 array - balanced feeds
• Tiled 8x RFEMs based on Intel WiGig product
• 1x Intel WiGig Baseband Modem Module
GEN3+ Evaluation Kit
13Intel Confidential
iCDG - Intel Communication and Devices Group
Link Budget Calculation
ITU Region N (1 Gbps threshold)
Assumptions• Noise figure + implementation loss: (10.5 + 3) dB• PER < 1%• AWGN channel (phase impairment considered)
Calculate SNR values and find supportable MCS in AWGN channels
LOS Backhaul Access
No rain 650 m 380 m
99.00% availability 600 m 360 m
99.90% availability 470 m 290 m
99.99% availability 350 m 230 m
iCDG - Intel Communication and Devices Group
Antenna Field Regions
D
Anant Gupta, UCSB
Under the direction of:
Professor Madhow of UCSB and Professor Amin of Standard
Oct 31, 2016
iCDG - Intel Communication and Devices Group
Goal: Sparse array of subarrays for directive & steerable beams with:
• Sparse placement of subarrays (4x4 element arrays).• Optimal phase shifts for beamsteering.
Attribute:• Larger aperture Directivity ↑ and BW ↓• Sparse arrays with same/fewer elements
Challenge: • Sub-Nyquist generates aliasing and grating lobes• Problem different from traditional 2D placement (subarray
elements are fixed)
Approach: Non uniform configurations perform better in all metrics• Optimal placement of sub-arrays and phase processing• Algorithmic/application-level resiliency to aliasing (e.g. for
imaging)
Sparse Array of Sub-Arrays
Sparse array
conventional array
Intel MAA-RFEM4x4 Module
iCDG - Intel Communication and Devices Group
Early Insights
Trade-offs in different architectures:
Metrics:
BW, Grating/side lobes, Directivity
Directivity saturates beyond certain aperture size
0 2 4 6 8 10
Subarray seperation (6)
18
19
20
21
22
23
24
25
GD
(d
B)
Directivity v/s Subarray Seperation
plus
Square
Benchmark
Sparse Non-uniform
-60 -40 -20 0 20 40 60
3°
-30
-25
-20
-15
-10
-5
0
No
rmali
ze
d G
ain
Gfinal
Horizontal ?=0°
SEP2
MRA
Uniform
Benchmark
iCDG - Intel Communication and Devices Group
Major Metrics & Approach
0 20 40 60 80 100 120 140
Elements
-9.2
-9
-8.8
-8.6
-8.4
MS
LL
Sequential Steering weight Optimization
Round 1
Round 2
Round 3
Round 4
Cost functions• MSLL: Maximum Side lobe level(relative to main lobe)• Directivity Gain-• 2D Beamwidth: (3 dB beam)Max* (3 dB beam)Min
• ASLL (Average Side Lobe Level)
Sub-Array Placement: Greedy search• Sequentially search for subarray positions on all possible
locations of grid (dx=0.5λ, dy=0.6λ).
Steering weight optimization: Sequential Optimization • Scan for best steering weight across all elements to
reduce MSLL.
iCDG - Intel Communication and Devices Group
Tradeoffs in Performance
Observations and tradeoffs
Tradeoff between beamwidth and sidelobe level as aperture size increases.
Beamwidth ∝ (Aperture area)-1
crc + lin B
Configs
-20
-15
-10
-5
dB
MSLL(rel. to mainlobe)
crc + lin B
Configs
22
24
26
28
dBi
Directivity Gain GD
crc + lin B
Configs
-25
-20
-15
dB
ASLL(rel. to mainlobe)
crc + lin B
Configs
101
102
103
deg2
Beamwidth(deg2)
Naive Seq-phase-Optimized Ideal
-10 0 10
x - 6 units
-10
-5
0
5
10
y - 6
un
its
Plus
-10 0 10
x - 6 units
-10
-5
0
5
10
y - 6
un
its
Circle
-10 0 10
x - 6 units
-10
-5
0
5
10
y - 6
un
its
Pseudo Linear
-10 0 10
x - 6 units
-10
-5
0
5
10
y - 6
un
its
Benchmark
iCDG - Intel Communication and Devices Group
Early Results; trade-offs in beam steering
Observations and tradeoffs
Phase optimization to ↓ MSLL causes ↓ Directivity.
Tradeoff between Directivity gain & sidelobe level with phase optimization
Beamwidth ∝ (Aperture area)-1
0 20 40 60
Steering Angle(El.)
22
24
26
28
GD
(dB
i)
Plus
0 20 40 60
Steering Angle(El.)
22
24
26
28
GD
(dB
i)
Circle
Theory
Ideal
Rounding
Phase-opt
0 20 40 60
Steering Angle(El.)
23
24
25
26
27
28
GD
(dB
i)
Linear
0 20 40 60
Steering Angle(El.)
23
24
25
26
27
GD
(dB
i)
Benchmark
Directivity v/s steering
-10 0 10
x - 6 units
-10
-5
0
5
10
y -
6 u
nit
s
Plus
-10 0 10
x - 6 units
-10
-5
0
5
10
y -
6 u
nit
s
Circle
-10 0 10
x - 6 units
-10
-5
0
5
10
y -
6 u
nit
s
Pseudo Linear
-10 0 10
x - 6 units
-10
-5
0
5
10
y -
6 u
nit
s
Benchmark
iCDG - Intel Communication and Devices Group
Beam width is roughly inverse of physical array aperture width
Beamwidth and Aperture
iCDG - Intel Communication and Devices Group 23
Conclusion
• Substantial effort has been focused in the industry on the 5G access technology to improve capacity, latency, throughput, scalability and quality of service;
• Access technology alone cannot significantly improve network capacity;
• An end-to-end 5G system need be augmented by flexible and high throughput backhaul and fronthaul;
• mmWave technology is a great candidate for both access and backhauling to increase network throughput and capacity, and lower interference;
• Sparse array architecture provides additional feature to optimize array performance