1. No category

5G New Physical Layer
Design and Test
November, 2014
Sang-Kyo Shin
Keysight EEsof EDA
Agenda
– Objective
– Multi-Carrier Waveform Techniques
• OFDM vs. FBMC
• FBMC signal processing
• Reference transmitter and receiver modeling, simulation and test
– MIMO and Digital Beamforming Techniques
• Diversity, Spatial Multiplexing
• Multi-user, Massive MIMO
• Modeling and Simulation Case studies
5G & mmWave
Workshop
Page 2
5G Enabling Devices >> New 5G R&D Challenges
Advanced signal processing
New waveforms
•
•
Legacy OFDM enhancement
FBMC, GFDM, UFDM
•
•
•
Multiple MIMO modes and beamforming
Network interference suppression
Adaptive channel estimation / equalization
Full duplex communications
•
•
•
Amplifier
•
•
•
Envelope tracking
Digital predistortion
Wide, multi bands
Multi-antenna
•
•
•
Access
•
•
Self interference cancellation
Dual polarization antenna
Real time operation
Non-orthogonal
multiple access
Random / scheduled /
hybrid
•
Multi-band
Multiple radio access technologies
•
•
Impedance matching
Mutual coupling
Multi-band, multi-RAT port
sharing
FD / Massive MIMO
GSM/EDGE/WCDMA/HSPA/LTE
WiFi/BT/WiGig/GNSS/5G
•
•
•
Traditional cellular bands <6GH
WiFi, BT, GNSS bands
5G mmWave bands
5G & mmWave
Workshop
Page 3
What we have for early 5G research today?
W1906BEL, 5G baseband exploration library
• 5G candidate physical layer modeling source codes for early research customer
• Committed by Keysight evolving toward world’s first 5G standard compliant library
MATLAB_Script
M1 {MATLAB_Script@Data Flow Models}
Modeling New Physical Layer
– Provide 5G candidate
waveform
• Multi-carrier modem
Tx/Rx processing chain
• FBMC,OFDM, etc…
– Usable with 4G standard
library
Multi-Antenna Techniques
– Advanced / adaptive signal
processing
• MIMO
• Digital beamforming
– Combined 2D/3D MIMO
channel simulation(W1715)
Algorithm Design / Verification
– Real world simulation
environments with
polymorphic language
selection
• Custom C++ model builder
• MATLAB®
• MATLAB® Script™
5G & mmWave
Workshop
Tackling Multi-Domain Issues
– Integrated to advanced
platform
• SystemVue/ADS/EMPro/
GG
• Keysight Instruments
Page 4
Agenda
– Objective
– Multi-Carrier Waveform Techniques
• OFDM vs. FBMC
• FBMC signal processing
• Reference transmitter and receiver modeling, simulation and test
– MIMO and Digital Beamforming Techniques
• Diversity, Spatial Multiplexing
• Multi-user, Massive MIMO
• Modeling and Simulation Case studies
5G & mmWave
Workshop
Page 5
Waveform Requirements
Figure 1.
– OFDM vs. FBMC
Spectrum Using
different filter overlap
factor
• Enable efficient multiple access
• High density of users
• Carrier assignment schemes in asynchronous
context
Figure 2.
– FBMC Fragmented
Spectrum
• Efficient usage of the allocated spectrum
• Robustness to narrow-band jammers and impulse
noise
Figure 3.
• High performance spectrum sensing
– Prototype Filter
Design
• Low computational complexity
– Filter overlap factor
K : number of
multicarrier symbols
which overlap in the
time domain
• Compatibility OFDM vs. NEW
5G & mmWave
Workshop
Page 6
Waveform Design Considerations for 5G
Waveform
Bandwidth /
Frequency
Advanced Multi-Carrier Waveforms1
OFDM
3GHz
FBMC / OFDM / Others
10GHz
Single carrier
30GHz
90GHz
>> Wider BW, Higher Fc, More robustness against phase noise
New RAT
OFDMA/
NOMA?
NOMA/
OFDMA?
Note1:
•
•
•
•
•
Orthogonal Frequency Division Multiplexing(OFDM)
Filter Bank Multicarrier(FBMC)
Universal Filtered Multicarrier(UFMC)
Generalized Frequency Division Multiplexing(GFDM)
Biorthogonal Frequency Division Multiplexing(BFDM)
5G & mmWave
Workshop
Page 7
OFDM
Advantage
Drawback
– Good spectral efficiency
– Some loss of spectral efficiency due to Cyclic
Prefix insertion
– Resistance against multipath interference
– Efficiently implemented using FFTs and IFFTs
– Imperfect synchronization cause loss of
orthogonality
– Subcarrier nulls correspond to peaks of
adjacent subcarriers for zero inter-carrierinterference
– Large peak to average power ratio(PAR) leads to
amplifier inefficiency
– High out-of-band power
– Subcarrier intermodulation must be reduced
frequency
f1
f2
5G & mmWave
Workshop
Page 8
Synthesis Filter bank
Symbol
de-mapping
Symbol
de-mapping
FFT
S / P
P / S
IFFT
Sub-carrier
mapping
Symbol
mapping
Sub-carrier
de-mapping
OFDM baseband signal processing blocks
Sub-carrier
de-mapping
post processing
OQAM
FFT
Poly Phase
Network
S / P
P / S
Poly Phase
Network
IFFT
OQAM
preprocessing
Sub-carrier
mapping
Symbol
mapping
OFDM vs. FBMC
Analysis Filter bank
FBMC baseband signal processing blocks
5G & mmWave
Workshop
Page 9
FBMC Signal Processing Block
OQAM preprocessing
𝐶2𝑅𝑘
𝑑0, 𝑛
Synthesis Filter Bank
𝜃0, 𝑛
𝛽0, 𝑛
x
x
𝛽1, 𝑛
x
x
.
.
.
.
.
.
𝑑1, 𝑛
𝛽 0, 𝑛
𝐴0(𝑧 2 )
↑ 𝑀/2
+
𝑧 −1
𝐴1(𝑧 2 )
𝑰𝑭𝑭𝑻
.
.
.
↑ 𝑀/2
.
.
.
↓ 𝑀/2
x
𝜃 0, 𝑛
𝑆𝑢𝑏𝐶𝐻
Proc
x
x
.
.
.
𝐴𝑀 − 1(𝑧 2 )
↑ 𝑀/2
.
.
.
𝛽 1, 𝑛
↓ 𝑀/2
𝐵 1(𝑧 2 )
.
.
.
.
.
.
Transform
Poly phase
filtering
FBMC transmitter
P/S
Conversion
𝑭𝑭𝑻
↓ 𝑀/2
S/P
Conversion
𝐵 𝑀 − 1(𝑧 2 )
𝑆𝑢𝑏𝐶𝐻
Proc
𝑅𝑒
x
x
.
.
.
𝛽 𝑀 − 1, 𝑛
𝜃𝑀 − 1, 𝑛
𝑆𝑢𝑏𝐶𝐻
Proc
Transform
x
Sub
channel
processing
𝑅2𝐶𝑘
𝑑 1, 𝑛
𝑅𝑒
.
.
.
x
Poly phase
filtering
𝑑 0, 𝑛
𝜃1, 𝑛
x
𝑧 −1
𝑑𝑀 − 1, 𝑛
Staggering
𝐵 0(𝑧 2 )
𝑧 −1
+
𝑧 −1
𝜃𝑀 − 1, 𝑛 𝛽𝑀 − 1, 𝑛
𝐶2𝑅𝑘
OQAM postprocessing
𝑠[𝑚]
𝜃1, 𝑛
𝐶2𝑅𝑘
Analysis Filter Bank
𝑅2𝐶𝑘
𝑑 𝑀 − 1, 𝑛
𝑅𝑒
𝑅2𝐶𝑘
Destaggering
FBMC receiver
5G & mmWave
Workshop
Page 10
OQAM Preprocessing
𝑐𝑜𝑚𝑝𝑙𝑒𝑥 𝑡𝑜 𝑟𝑒𝑎𝑙 𝑐𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛
𝑐𝑘 𝑙
𝑅(. )
𝜃𝑘 𝑛 = 1, 𝑗, 1, 𝑗, 1, . .
↑2
𝑓𝑜𝑟 𝑘 𝑒𝑣𝑒𝑛
+
𝑐𝑘 𝑙
𝑗𝐼(. )
↑2
𝑧 −1
𝑅(. )
↑2
𝑧 −1
↑2
𝑑𝑘 𝑛
x
𝑥𝑘 𝑛
𝜃𝑘 𝑛 = 𝑗, 1, 𝑗, 1, 𝑗. .
+
𝑓𝑜𝑟 𝑘 𝑜𝑑𝑑
𝑗𝐼(. )
𝜃 pattern 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛
𝑑𝑘 𝑛
x
𝑥𝑘 𝑛
• A time offset of half a QAM symbol period(T/2) is applied to either the real part or the
imaginary part of the QAM symbol
• For two successive sub-channels, say m and m+1, the offset are applied to the real part of
the QAM symbol in sub-channel , while it is applied to the imaginary part of the QAM
symbol in sub-channel m+1.
5G & mmWave
Workshop
Page 11
Poly Phase Network Filter Bank
𝑑0, 𝑛
𝜃0, 𝑛
𝛽0, 𝑛
x
x
𝐴0(𝑧 2 )
↑ 𝑀/2
+
𝑀−1
𝑠𝑚 =
∞
.
𝑘=0 𝑛=−∞
𝜃1, 𝑛
𝛽1, 𝑛
x
x
𝑑1, 𝑛
𝑠𝑚
.
.
.
.
.
.
𝑧 −1
𝑰𝑭𝑭𝑻
𝐴1(𝑧 2 )
.
.
.
↑ 𝑀/2
.
.
.
+
.
.
.
* Filter overlap factor K : number of multicarrier symbols which
overlap in the time domain.
𝑤ℎ𝑒𝑟𝑒:
𝑑𝑘 , 𝑛 𝜃𝑘 , 𝑛 𝑔𝑘 𝑚 − 𝑛𝑀/2
M is number of subcarriers
𝑑𝑘 , 𝑛 𝑖𝑠 𝑡ℎ𝑒 𝑟𝑒𝑎𝑙 𝑣𝑎𝑙𝑢𝑒𝑑 𝑠𝑦𝑚𝑏𝑜𝑙
𝜃𝑘 , 𝑛 𝑖𝑠 𝑗
(𝑘 + 𝑛 )
𝑔𝑘 (m) is impulse response of the filters
* OFDM can be implemented by set K as 1
5G & mmWave
Workshop
Page 12
Sub-channel Equalization
Maximal ratio combined diversity reception
t[𝑘]
transmitted
symbol
Channel
Estimation
𝑤i
H[z]
Evaluation of MRC weighted target values
𝑦[𝑘]
distorted subcarrier
sequence
2
𝑍-1
𝑍-1
𝑣𝑘 𝑛 =
X
𝑤0
X
+
𝑤1
X
+
𝑙=0
𝑤2
𝑤𝑘 , 𝑙, 𝑛 𝑦𝑘 𝑛 − 𝑙
𝑙 = number of tap
𝑣𝑘 𝑛
3-tap Complex FIR frequency sampling-design
5G & mmWave
Workshop
Page 13
OQAM post processing
𝑟𝑒𝑎𝑙 𝑡𝑜 𝑐𝑜𝑚𝑝𝑙𝑒𝑥 𝑐𝑜𝑛𝑣𝑒𝑟𝑠𝑖𝑜𝑛
𝜃pattern 𝑚𝑢𝑙𝑡𝑖𝑝𝑙𝑖𝑐𝑎𝑡𝑖𝑜𝑛
𝜃𝑘 𝑛
𝑓𝑜𝑟 𝑘 𝑒𝑣𝑒𝑛
𝑥𝑘 𝑛
x
𝑑𝑘 𝑛
↓2
𝑅(. )
𝑐𝑘 𝑙
𝑧 −1
+
−1
↓2
𝑗
↓2
𝑧 −1
𝑗
𝑧
𝜃𝑘 𝑛
𝑓𝑜𝑟 𝑘 𝑜𝑑𝑑
𝑥𝑘 𝑛
x
𝑑𝑘 𝑛
𝑅(. )
𝑐𝑘 𝑙
+
𝑧
−1
↓2
5G & mmWave
Workshop
Page 14
Modeling / Simulation Example for FBMC Systems
Random
bit
generation
Symbol
Mapping
FBMC
Reference
Source
LO source
Phase/
Power
Modulator
Wireless
Channel
AWGN
FO,IQ Im
Demodulator
FO,IQ Im
ADC
Jitter /
Q noise
FBMC
Reference
Receiver
BER/FER
Measurem
ent
TEST
O1 {Oscillator@Data Flow Models}
REF
Im
11010
•••
•••
•••
•••
B1 {RandomBits@Data Flow Models}
M1 {Mapper@Data Flow Models}
ModType=QPSK [ModType]
Re
FBMC_Source_1
BERFER {BER_FER@Data Flow Models}
QUAD
OUT
Mod OUT
FBMC_Source
MAPPER
Freq
Phase
Q
I
Amp
Taps
Channel
Out
Noise
Density
Freq
Phase
Q
DeMod
I
Amp
DEMAPPER
Im
FBMC_Receiver
Re
C4 {CxToRect@Data Flow Models} M2 {Modulator@Data Flow Models}
C1 {CommsChannel@Data Flow Models}
D3 {Demodulator@Data Flow Models} R3 {RectToCx@Data Flow Models}
ModelType=Pedestrian_A A1 {AddNDensity@Data Flow Models}
NDensityType=Constant noise density
OutputType=I/Q
NDensity=10e-12 W [NDensity]
FCarrier=1e9 Hz
FBMC_Receiver_2
• • • Node
•••
•••
Bits
D2 {Demapper@Data Flow Models}
ModType=QPSK [ModType]
Simulation parameters
5G & mmWave
Workshop
•••
Page 15
Performance Analysis with Real Hardware
SYSTEMVUE
RFIC DUT
M8190A 12 GSa/S Arbitrary
Waveform Generator
Reference Source
I
Digital Modem Source
for Linear Modulation
Frame Structure
Idle
Preamble
Data Payload
Automatic waveform
creation & download
DSSS System
X
Custom modem
design
: Replaceable
in C++, .m or
SV DSP
parts formats
BBIQ - RF
Q
Spreading Code
Generator
BPSK, QPSK, ..., up to 4096-QAM
8-PSK, 16-PSK, 16-APSK, 32-APSK
16-Star QAM, 32-Star-QAM,
and Custom APSK
Preamble_ModType=BPSK [Preamble_ModType]
Payload_ModType=16-QAM [Payload_ModType]
I
5G Reference
Library
Q
Digital Modem Receiver
Feedward
Filter
-
Decision
Device
-
Feedback
Filter
RF - BBIQ
Reference
Receiver
M9703A AXIe 12-bit High-Speed
Digitizer/Wideband Digital Receiver
Interleaving to get 4ch @ 3.2 GSa/s
Decision Feedback Equalizer
Fast Computation Algorithm
CIR--->DFE coefficients
{DigMod_ReceiverL_FastDFE}
FrameSync_Algorithm=DiffCorr
FreqSync_Mode=CIR Corr
TrackingAlgorithm=LMS
TEST
REF
BERFER {BER_FER@Data Flow Models}
•
•
Wider BW (63 GHz BW)
Higher Sampling (160 GSa/s)
BER/FER Measurement
Infiniium 90000 Q-Series Oscilloscope
5G & mmWave
Workshop
Page 16
Agenda
– Objective
– Multi-Carrier Waveform Techniques
• OFDM vs. FBMC
• FBMC signal processing
• Reference transmitter and receiver modeling, simulation and test
– MIMO and Digital Beamforming Techniques
• Diversity, Spatial Multiplexing
• Multi-user, Massive MIMO
• Modeling and Simulation Case studies
5G & mmWave
Workshop
Page 17
Motivation
– Higher requirement for system capacity and spectral efficiency(bits/s/Hz)
– To overcome traditional approaches ( expand bandwidth, higher modulation order,
multiple access)
– The MIMO for better use the spatial resource
• The capacity is increased by a multiplication of the number of antennas
S

C  B  log 2 1  bit / s  M
 N
5G & mmWave
Workshop
Page 18
Classification
y1
X1
Multi-user
Increase system
efficiency
Multi streams/users
Spatial division multiplexing
Receive Diversity
MIMO
Multi-user MIMO
.
.
.
.
.
.
X1, X2
y1, y2
Use spatial channel
information?
-X2, X1*
Matrix
Space-time block coding (STBC)
Transmit Beamforming
• Open-loop MIMO
• Closed-loop MIMO
5G & mmWave
Workshop
M >> K >> 1
Massive MIMO
Page 19
K terminals
y2
X2
Transmit Diversity
Massive multi-users
S streams
Spatial Expansion
Spatial multiplexing
Improve user throughput
M antennas
Spatial diversity
Improve robustness
Transmit Diversity
– Use transmit diversity to diminish the effects of fading by
transmitting the same information from two different
antennas
X1, X2
y1, y2
– The data from the second antenna is encoded differently
to distinguish it from the primary antenna
-X2, X1*
– The transmit diversity feature uses ST(space-time) or
SF(space-frequency) block encoding to differentiate the
signals between Antenna 1 and Antenna 2
– The user equipment (UE) must be able to recognize that
the information is coming from two different locations and
properly decode the data.
SFBC:
Tx0
Tx1
f1
f2
x2 
 x1
 x 2 * x1 *


STBC:
t1
t2
* complex conjugate
5G & mmWave
Workshop
Page 20
Spatial Multiplexing
– Operation Concept
• Transmission of multiple spatial data streams over
different antennas in the same RB
y1
X1
h11
h21
• The dimension of spatial channels is increased and
system capacity increased
h12
h22
– Relevant signal processing
• Perform Layer mapping and Pre-coding to lower the
receiver complexity and reduce the signal interference
between antennas
• Statistic correlation between vector(h11,h12) and
vector(h21,h22 )
X2
y2
x: transmitted signal,
y: received signal,
H: spatial channel matrix,
Hij: channel coefficient from the jth transmit
antenna and the ith receive antenna.
y=Hx
y1=h11x1+h12x2+n1
y2=h21x1+h22x2+n2
5G & mmWave
Workshop
Page 21
Modeling and Simulation for MIMO
– MIMO Tx/Rx simulation under Rayleigh fading and AWGN channel
– Explore different decoding algorithms and performance evaluation
• ML, MMSE-SIC, ZF-SIC, MMSE-Linear, ZF-Linear
AWGN
Fading Channel
I5 {IID_Gaussian@Data Flow Models}
StdDev=0.707 V [1/sqrt(2)]
Im
[ ]
Re
I6 {IID_Gaussian@Data Flow Models}
StdDev=707.1e-6 V [StdDev]
Im
Re
R3 {RectToCx@Data Flow Models}
P1 {Pack_M@Data Flow Models}
NumRows=2 [ChannelNumRows]
NumCols=2 [ChannelNumCols]
Format=ColumnMajor
I7 {IID_Gaussian@Data Flow Models}
StdDev=0.707 V [1/sqrt(2)]
R1 {RectToCx@Data Flow Models}
[ ]
P2 {Pack_M@Data Flow Models}
NumRows=2 [RxNumRows]
NumCols=1 [RxNumCols]
Format=ColumnMajor
I2 {IID_Gaussian@Data Flow Models}
StdDev=707.1e-6 V [StdDev]
DEMAPPER
M o d Type
Ch a n n e l Re s ponse
•••
11010
•••
•••
[ ]
•••
MAPPER
B2 {RandomBits@Data Flow Models}
Re c o v e re d Data
M5 {Mapper@Data Flow Models}
ModType=QPSK [ModType]
MIMO_Decoder
MIMO_Encoder
P3 {Pack_M@Data Flow Models}
NumRows=2 [TxNumRows]
M1 {MIMO_Encoder@5G Advanced Modem Models}
NumCols=1 [TxNumCols]
Mode=Spatial Multiplexing [Mode]
Format=ColumnMajor
NumTx=2 [NumTx]
Transmit with MIMO coding
Re c e i v e d Data
M2 {Mpy@Data Flow Models}
A4 {Add@Data Flow Models}
M3 {MIMO_Decoder@5G Advanced Modem Models}
Mode=Spatial Multiplexing [Mode]
DecoderMethod=ML [DecoderMethod]
ModType=QPSK [ModType]
DebugFlag=0
[ ]
U1 {Unpack_M@Data Flow Models}
NumRows=2 [TxNumRows]
NumCols=1 [TxNumCols]
Format=ColumnMajor
•••
•••
Node
•••
•••
Bits
D1 {Demapper@Data Flow Models}
ModType=QPSK [ModType]
MIMO decoding and demapper
5G & mmWave
Workshop
Page 22
Multi-User MIMO
Capacity Comparison
SU−MIMO: 𝑀𝑙𝑜𝑔(1 + 𝑆𝑁𝑅)
𝑆𝑁𝑅
MU-MIMO: 𝑀𝑙𝑜𝑔 1 + 𝑀 𝑙𝑜𝑔𝑈 , 𝑈 → ∞
M: TX antenna number, U: Total user number
MU-MIMO Scenario
Received signal at UE k:
The challenge for MU-MIMO is to find orthogonal
users and design precoding W to minimize the
second term with the restrictions of user grouping,
power, latency and complexity
Hk: kth user’s channel, W k: weight vector, Sk: data symbol
5G & mmWave
Workshop
Page 23
Multi-User MIMO
Advantages
Disadvantages
– Multiple access capacity gain (proportional to
BS antennas)
– BS needs to know channel state information at
transmitter (CSIT). The challenges include
– more immune to propagation limitations such
as channel rank loss, antenna correlation and
LOS
– Maintain spatial multiplexing gain without large
antenna number at terminals
• TDD vs. FDD for CSIT
• CSI feedback path bandwidth, Code book
design
– Complexity of the scheduling procedure at BS
• User grouping scheduling, power allocation
and latency requirements
5G & mmWave
Workshop
Page 24
Modeling and Simulation for Capacity Estimation
Channel transfer matrix
I1 {IID_Gaussian@Data Flow Models}
StdDev=0.707 V [1/sqrt(2)]
Im
[ ]
Re
Capacity measurement
User scheduling
H
Power_Selected
P
W_Selected
W
User Scheduler
R2 {RectToCx@Data Flow Models}
P4 {Pack_M@Data Flow Models}
NumRows=1 [NumRx]
NumCols=4 [NumTx]
Format=ColumnMajor
D2 {Distributor@Data Flow Models}
BlockSize=1
UserScheduler {MATLAB_Script@Data Flow Models}
TotalUsers=100 [TotalUsers]
NumTx=4 [NumTx]
NumRx=1
TotalPower=10 [SNR]
123
R
Channel Capacity
H_Selected
H
SumRate {MATLAB_Script@Data Flow Models}
NumTx=4 [NumTx]
Noise=1
NumRx=1
A1 {Average@Data Flow Models}
NumInputsToAverage=100
S4 {Sink@Data Flow Models}
StartStopOption=Samples
I3 {IID_Gaussian@Data Flow Models}
StdDev=0.707 V [1/sqrt(2)]
Simulation condition
Sum Capacity
– Transmit antenna number (M) : 4
– Total number of user : from 4 to 100
– SNR=10dB
– Power allocation by waterfilling algorithm
User K: 4->100
5G & mmWave
Workshop
Page 25
Massive MIMO
– Originally envisioned for time division duplex(TDD1), but can
potentially be applied in frequency division duplex(FDD)
.
.
.
.
.
.
K terminals
– Brings huge improvements in throughput and energy efficiency
when combined with simultaneous scheduling of a large number of
UEs
S streams
– System Model : M transmit antenna with maximum S streams, K
users each with a single antenna
Massive multi-users
M antennas
– The use of a very large number of service antennas operated fully
coherent and adaptive
M >> K >> 1
Massive MIMO
Note1 : Prefer TDD as not enough resources for pilots and CSI feedback.
5G & mmWave
Workshop
Page 26
Massive MIMO Operation and Challenges
Operation
Challenges
– Acquire Channel State Information from uplink
Pilots / Data
– Pilot contamination: interference from other cells
– Reciprocity calibration and adjustment
– Pre-coding1 to support multi-stream
transmission
– MMSE receiver with beamforming
• Maximum ratio combining(MRC) : interference
and noise are both white in the space
• Interference rejection combining(IRC): colored
interference
Note1 : Linear pre-coding [maximum ratio transmission(MRT), zero-forcing(ZF)].
Non-linear pre-coding [Dirty paper coding(DPC)], full CSI required
• Blind channel estimation?
• Coordination and planning?
– New pre-coder with low-complexity, low-PAPR
– Hardware performance
• I/Q imbalance, A/D resolution, PA linearity
• Phase noise, clock distribution
– Synchronization at low SNR
– Understand mmWave MIMO channel
5G & mmWave
Workshop
Page 27
Modeling and Simulation for Large Number of Antennas
Transmit
Beamformer
Multi-CH
Modulator
Multi-CH
Envelope Adder
Multi-CH
AWGN
Multi-CH
De-Modulator
Receive
Beamformer
O1 {Oscillator@Data Flow Models}
Frequency=1000000 Hz
Power=.010 W
T1 {Tx_Beamformer@5G Advanced Modem Models}
BeamformingType=Calculate by antenna …
NumOfAntx=4
NumOfAnty=4
Dx=0.5
Dy=0.5
Theta=0 °
Phi=0 °
M2 {MultiCh_Demodulator@5G Advanced Modem Models}
NumChannels=1
FCarrier=1e6 Hz
InitialPhase=0 ° [[0]]
AmpSensitivity=1 [[1]]
MirrorSignal=NO
ShowIQ_Impairments=NO
M4 {MultiCh_AddEnv@5G Advanced Modem Models}
OutputFc=Center
InPhi
LO
weights
Tx
Beamformer
InTheta
quad_output
ref
MultiChannel
Modulator
output
input
input
MultiCh
Noise Density
output
MultiChannel
Demodulator
weights
Rx
Beamformer
input
output
Env
M1 {MultiCh_Modulator@5G Advanced Modem Models}
NumChannels=1
FCarrier=1e6 Hz
InitialPhase=0 ° [[0]]
AmpSensitivity=1 [[1]]
ConjugatedQuadrature=NO
MirrorSignal=NO
ShowIQ_Impairments=NO
M6 {MultiCh_AddNDensity@5G Advanced Modem Models}
NDensityType=Constant noise density
NDensity=0.0 W
Scripting
•
•
•
•
Multiuser scheduling
Capacity analysis
Quick algorithm
implementation and test
Calibration
R1 {Rx_Beamformer@5G Advanced Modem Models}
NumOfTxAnts=16
ABF_Algorithm=Sample Matrix Inversion
BlockSize=1024
Plotting
•
•
•
Antenna pattern review
Interference analysis
between different
streams
Beam pattern vs. precoding analysis(MRT,ZF)
5G & mmWave
Workshop
Page 28
Channel Sounding / Parameter Extraction / Simulation
Channel
impulse
response
Reference transmit
signal(chirp/pn)
channel
H[z]
t[𝑘]
𝑧[𝑘]
∑
•
•
•
PDP (Path delay, path loss)
AOA, AOD
Doppler shift
Channel
parameters
Estimation
algorithms
CIR
correlation
Channel sounding
Parameters estimation
Statistics & modeling
•
•
•
Scenario selection
Network layout
Antenna parameters
•
•
•
•
•
AS AoA/AoD
PAS
Doppler spectrum
Correlation
Rician K factor
Large/Small scale
parameters
generation
Fading coefficient
generation
Input signal
𝑥[𝑘]
¤
ESPRIT
Subspace based algorithm
Maximum estimating
number of path is limited by
number of Rx, will be fail
under NLOS scenario
cannot estimate path loss
and path delay
 small computing amount
SAGE
Maximum likelihood
estimation algorithm
 No limitation for number
of path, suitable for both
LOS and NLOS scenarios
 Can estimate all the
channel parameters
including path loss and path
delay of each path
Iteration needed, large
computing amount
faded signal
𝑦[𝑘]
SystemVue Simulation
5G & mmWave
Workshop
Page 29
Prototyping and Testing in Real Time Hardware
– Move forward from largely theoretical massive MIMO research to real hardware
implementation and test
– Open FPGA and download custom algorithms for MIMO and Beamforming
– Test and measure in real-time
CUSTOM
ALGORITHMS
FPGA
ARRAY
M9703
REAL-TIME PROCESSING
Up to 40 Channels x 1GHz wide
5G & mmWave
Workshop
Page 30
Summary
– Early research task start with huge math and software simulation
– Connected solution between Design/Simulation and Test/Verification
– Real time, Multi-channel, wide-band are key themes for 5G research
– Hardware-Software-People
5G & mmWave
Workshop
Page 31