E-Band Wireless Backhaul: System Design & Test Challenges · PDF fileE-Band Wireless Backhaul: System Design & Test Challenges Presented by: Agilent Technologies, Inc. Daren Sang McClearnon

E-Band Wireless Backhaul: System Design & Test Challenges

Presented by:

Agilent Technologies, Inc.

Daren McClearnon

Sang-kyo Shin

Webcast, 27 March 2014

Outline System-level approach to modeling, simulating and verifying E-Band Systems

E-Band Webcast , March 2014

© Agilent Technologies, Inc.

2

HARDWARE &

MEASUREMENTS 4

ADVANCED MODEM

TECHNOLOGIES 2

E-BAND SYSTEM

OVERVIEW 1

RF INTEGRATED

VERIFICATION & TEST 3

Future mobile network infrastructure Where does the data actually go?

Device to device

communications

Vehicular

communication

Wireless

backhaul

Ultra dense

deployments

Machine

type devices

Multi-hop

Macro

Sub-GHz

100GHz

200GHz

300GHz

E-band

(71-76/81-86GHz)

E-band

(71-76/81-86GHz)



3

E-band applications Small Cell Backhaul – inexpensive, compact, capacity, easy deployment

Macro site



4

E-band applications Mobile Backhaul – flexible locations, multiple services

BBU

RRU

Macro site

CPRI



5

E-band applications Lower-cost infrastructure complement to fiber

o

p

t

i

c

o

p

t

i

c

o

p

t

i

c Optic Ring

Part

ial re

pla

cem

en

t

Fiber like connectivity



6

Light Licensing of ITU “e-band” Spectrum

71-76/81-86 GHz License light regime in USA and Europe

- 2005, USA FCC Light licensing

- 2007, UK office of communications (OFCOM)

- 2007, Australian Communication and Media Authority (ACMA)

• FCC Memorandum Opinion and Order 05-45

• FCC Public Notice DA 05-311

• ECC Recommendation (05)07

• ETSI EN 302 217-2-2,3

Why e-band for wireless backhaul?

• Uncongested frequency band

• High antenna directivity with pencil

beam

• Separate two frequency band without

interference concern

• Light licensing scheme

60GHz 70GHz 80GHz 90GHz

V-Band

Unlicensed

E-Band

Light Licensed

W-Band

Fragmented

5GHz 5GHz

ITU frequency classification in millimeter wave band



7

ITU-R Radio-Frequency Arrangements F.2006



8

E-Band Products, Technical Information

Functionality 1st Generation 2nd Generation Scalable 2nd

Generation Future

Modulation Single Carrier, QPSK/BPSK

Adaptive Modulation – 64 QAM or Higher

Adaptive Modulation – 64 QAM or Higher

Multi-level

modulation or

other

Channel Width

Fixed or Variable

Channels

250 MHz – 1 GHz

Variable Channels

250 MHz – 500 MHz

Variable Channel Size

< 250 MHz

Variable

Channel Size

Capacity Less Than 1 Gbps 1 ~ 2 Gbps

2.5 Gbps, Extension

to 5 Gbps with

polarization

10 Gbps or More

Latency >150 µSeconds 21 – 150 µSeconds 20 µSeconds or Less ?

Channel Coding RS RS, LDPC RS, LDPC LDPC

Power Consumption

35+ Watts 21-34 Watts 20 Watts or Less ?



9

Design and Verification Challenges

• Signal routing & losses

• Limited performance

• Lack of characterization

data

RF / mm-WAVE

• System-level specs

• Accurate cross-domain

Modeling & Simulation

Architectures, Partitioning

Troubleshooting, Verification

SYSTEM ARCHITECT

• Advanced modem techn.

• Concurrent TX Channels

• Adaptive - Coding

- Modulation

- Channel spacing

- Channel compensation

BASEBAND

• Higher sampling rate

• Wider bandwidth

T&M VERIFICATION

• Complexity, density

• Concurrent teams

• Evolving Stds, regulations,

vendors, macro-economy, etc

ECOSYSTEM, REQUIREMENTS



10

SYSTEM-LEVEL

DESIGN COCKPIT

HARDWARE &

MEASUREMENTS 4

E-BAND SYSTEM

OVERVIEW 1

RF INTEGRATED


ADVANCED MODEM

TECHNOLOGIES 2




11

Signal processing approaches to RF/System challenges

• Millimeter-wave frequencies are difficult, expensive to

engineer

• “Where is the least expensive place to address key issues?”

RF Challenge Signal Processing Approach

Channel / Propagation Adaptive Equalization can maximize

available analog dynamic range vs. time

RF performance margins Powerful Coding techniques maximize the

data throughput for a finite SFDR



12

#1 - Adaptive Signal Processing

Changing environment

• Decision-directed adaptive equalizer

Analog impairments

• Automatic correction of TX and RX quadrature impairments in DSP

• Closed-loop adaptive digital pre-distortion (DPD)

Baseband efficiency

• Adaptive coding and modulation, with software-defined profile

• Re-configurable symbol rate and bandwidth

• Reed-Solomon and LDPC FEC, with configurable codeword length and payload amount

• Convolutional interleaver, with configurable depth

All about

ADAPTIVE

1



13

Channel Equalization

Why use EQ?

• Inter-symbol Interference (ISI) is created by multipath with

time dispersive channels (W>BC)

• “Equalization” compensates for ISI

EQ Types

• Linear: Time-invariant channels

• Nonlinear: Time-varying severe channel distortion - Decision Feedback Equalization (DFE)

- Maximum Likelihood Symbol Detection

- Maximum Likelihood Sequence Estimator (MLSE)



14

• Equalizer suppress time-varying communications channel

distortion by set d[k] = t[k-u]

• Relationship between d[k] and y[k] determines the mode of

the equalizer

𝑒[𝑘]

𝑒[𝑘] 𝑦[𝑘]

𝑦[𝑘] equalized symbol

Adaptive Equalization

∑ adaptive

filter

𝑑[𝑘]

𝑥[𝑘] t[𝑘]

+ −

transmitted symbol

communication

channel H[z]

𝑍-u



15

Decision Feedback Equalizer (DFE)

Use feedback of detected symbols

∑

𝑚[𝑘]

𝑒[𝑘]

𝑥[𝑘] 𝑦[𝑘] t[𝑘]

+ −

transmitted symbol

Figure 2. Decision Feedback Equalizer

adaptive

filter

∑ +

+

adaptive

filter detector

channel

H[z]

demodulated samples input

forward Filter

feedback Filter

data output

quantized samples

error signals for adjusting weights

equalized samples



16

EQ convergence and Mean-Squared Error (MSE)

• RLS converges faster than

LMS algorithm towards the

final mean squared error

• But, LMS is simpler

• RLS/LMS algorithm

successfully combines

speed & efficiency

Figure 5. MSE Curve of Equalization Results



17

#2 Coding Algorithms

Typical behavior of the bit error probability versus bit signal-to-noise ratio for uncoded and coded systems

2 Transmitter

Noisy

Channel Receiver

uncoded

coded

coded

uncoded

Pb(e)

Eb / No (dB)

Potential coding gains of coded transmission v.s uncoded

Pb(e)

Eb / No (dB)

10 -1

10 -2

10 -3

10 -4

10 -5

10 -6

10 -7

10 -8

-2 -1 0 1 2 3 4 5

“coding gain”



18

High Performance Coding Algorithms being adopted in E-Band Modems

Ingredients of Shannon’s proof?

• Long, structured, “pseudorandom” codes

• Practical, near-optimal decoding algorithms

For example?

• Turbo codes (1993)

• Low-density parity-check (LDPC) codes

(1960, 1999)

• Bring Shannon limits to within reach, on a wide

range of channels.

What’s the problem?

• complexity

Claude Elwood Shannon

(1916-2001)



19

Bit Error Rate & Channel Coding Performance

Figure. Simulated BER v.s Eb/N0 curves in AWGN channel.

Optimum receiver with feed forward AGC, differential correlation

frame synchronization, correlation frequency synchronization,

phase estimation, no equalizer, DD-PLL frequency tracking.

LDPC

Coding Gains

Figure. Simulated BER v.s Eb/N0 curves with Custom LDPC Code.

Optimum receiver with FEC sub-system (BCH and LDPC Coding)

• 1st Gen : BPSK, QPSK

• 2nd Gen : 16, 32, 64 QAM

• 3rd Gen : Higher Order ?

Major E-band Modem Modulation Schemes



20

Digital Modem – Algorithmic Verification

Building a robust DSP chain, then validating the architecture

CODE MOD DE-

CODE

CHAN

NOISE

INTERF

DE-

MOD

FRAME CODE MOD DE-

CODE FRAME

SYNC

ADAPT

EQ AGC FILTER

CHAN

NOISE

INTERF

FREQ

PHASE DE-

MOD FILTER

LDPC

CODE

DE-

CODE

ALGORITHMIC

REFERENCE

RECEIVER

CHAN

NOISE

INTERF

ALGORITHM IC

REFERENCE

SOURCE



21

Digital Modem – Algorithmic Verification

Building a robust DSP chain, then validating the architecture

CODE MOD DE-

CODE

CHAN

NOISE

INTERF

DE-

MOD FRAME CODE MOD

DE-

CODE FRAME

SYNC

ADAPT

EQ AGC FILTER

CHAN

NOISE

INTERF

FREQ

PHASE DE-

MOD FILTER LDPC

CODE

DE-

CODE

ALGORITHMIC

REFERENCE

RECEIVER

CHAN

NOISE

INTERF

ALGORITHM IC

REFERENCE

SOURCE

InitialDelay=0N=100000 [NumOfBitsPerFrame]D4 {InitDelay@Data Flow Models}

InitialDelay=0N=10000 [PayloadLength*DSSS_factor]

D1 {InitDelay@Data Flow Models}

(X-Y)2

N

DisplayOption=dBFrameLength=10000 [PayloadLength*DSSS_factor]

FramesToAverage=10StartFrame=1

EVM {RMSE@Data Flow Models}

Noise

Density

NDensity=-131.021dBm [NDensity]

NDensityType=Constant noise densityA1 {AddNDensity@Data Flow Models}

1 1 0 1 0

DataPattern=PN15B2 {DataPattern@Data Flow Models}

MathLang

M1 {MathLang@Data Flow Models}

TEST

REF

BitsPerFrame=100000 [NumOfBitsPerFrame]StartStopOption=Samples

BERFER {BER_FER@Data Flow Models}

MathLang


BPSK, Q PSK, . . . , up t o 4096- Q AM8- PSK, 16- PSK, 16- APSK, 32- APSK

16- St ar Q AM , 32- St ar - Q AM ,

and Cust om APSK

Dat a PayloadPr eam bleI dle

Fr am e St r uct ur e

Spr eading Code

G ener at or

X

Digital Modem Sourcefor Linear Modulation

DSSS Syst em

Rolloff=0.35ShapingFilter=Raised Cosine

DSSS=NO [DSSS]IdleInterval=2e-6s [IdleInterval]GuardLength=0 [GuardLength]Payload_ModType=1024-QAM

PayloadLength=10000 [PayloadLength]Preamble_ModType=BPSK [Preamble_ModType]

PreambleSequence=(400x1) [1; -1; 1; 1]PreambleSource=YES

PreambleLength=400 [PreambleLength]SampsPerChip=4 [SampsPerChip]

ChipRate=10e+6Hz [ChipRate]InitialPhase=0°

Power_dBm=-10dBm [Power_dBm]FCarrier=5e+9Hz [FCarrier]

Source {DigMod_SourceL_RF@DigMod Models}

Decision

Device

Adapt ive

Equalizer

+

Equalizer(LE and DFE)

LMS and RLS

Adaptive Algorithm

Digital Modem Receiver

PLL_Bandwidth=4HzFreqTrackMode=DD-PLL

Fraction=1EQ_Type= NO Equalizer

Np=197

PhaseEstimator=YESLf=200Nf=200

FreqSync_Mode= Corr

SearchLength=100FrameSync_Algorithm=DiffCorr

AGC_Algorithm=FeedForward_AGCReceiver_Filter=NO Filter




ChipRate=10e+6Hz [ChipRate]FCarrier=5e+9Hz [FCarrier]

Subnetwork1 {DigMod_ReceiverL_EQ_RF@DigMod Models}PRBS

BER

EVM

CHAN

REF. TX REF. RX



22

InitialDelay=0N=100000 [NumOfBitsPerFrame]D4 {InitDelay@Data Flow Models}

InitialDelay=0N=10000 [PayloadLength*DSSS_factor]

D1 {InitDelay@Data Flow Models}

(X-Y)2

N

DisplayOption=dBFrameLength=10000 [PayloadLength*DSSS_factor]

FramesToAverage=10StartFrame=1

EVM {RMSE@Data Flow Models}

Noise

Density

NDensity=-131.021dBm [NDensity]

NDensityType=Constant noise densityA1 {AddNDensity@Data Flow Models}

1 1 0 1 0

DataPattern=PN15B2 {DataPattern@Data Flow Models}

MathLang


TEST

REF

BitsPerFrame=100000 [NumOfBitsPerFrame]StartStopOption=Samples


MathLang


BPSK, Q PSK, . . . , up t o 4096- Q AM8- PSK, 16- PSK, 16- APSK, 32- APSK

16- St ar Q AM , 32- St ar - Q AM ,

and Cust om APSK

Dat a PayloadPr eam bleI dle

Fr am e St r uct ur e

Spr eading Code

G ener at or

X


DSSS Syst em

Rolloff=0.35ShapingFilter=Raised Cosine



PreambleSequence=(400x1) [1; -1; 1; 1]PreambleSource=YES


ChipRate=10e+6Hz [ChipRate]InitialPhase=0°

Power_dBm=-10dBm [Power_dBm]FCarrier=5e+9Hz [FCarrier]

Source {DigMod_SourceL_RF@DigMod Models}

Decision

Device

Adapt ive

Equalizer

+

Equalizer(LE and DFE)

LMS and RLS

Adaptive Algorithm


PLL_Bandwidth=4HzFreqTrackMode=DD-PLL

Fraction=1EQ_Type= NO Equalizer

Np=197

PhaseEstimator=YESLf=200Nf=200

FreqSync_Mode= Corr

SearchLength=100FrameSync_Algorithm=DiffCorr

AGC_Algorithm=FeedForward_AGCReceiver_Filter=NO Filter




ChipRate=10e+6Hz [ChipRate]FCarrier=5e+9Hz [FCarrier]

Subnetwork1 {DigMod_ReceiverL_EQ_RF@DigMod Models}

Algorithm Verification Methodology

Integrate your BB algorithm IP into the system-level verification

SystemVue

Reference Design

Custom

LDPC

Algorithm Working TX/RX

Individual

algorithms

parameterized

Probe inside

Make test vectors



23

ADVANCED MODEM

TECHNOLOGIES 2

SYSTEM-LEVEL

DESIGN COCKPIT

HARDWARE &

MEASUREMENTS 4

E-BAND SYSTEM

OVERVIEW 1

RF INTEGRATED





24

E-Band RF Transceiver



25

TX

RX

RX

TX

UP Stream / Down Stream Communication

X 4

REF PLL

÷ 8

71-76GHz TX

81-86GHz RX

E-Band Transceiver Block Diagram Example

RFIC

Modeling RF System Architectures

Source1=FTXIF MHz at -12 dBm

TX_IF_IN {MultiSource}

Fhi=21750MHz [FTXIF+1500]

Flo=18750MHz [FTXIF-1500]

N=3

IL=0.01dB

BPF_Butter_3 {BPF_BUTTER}

RI

L

LO=7dBm

ConvGain=0dB [TXRFMIXERG]

Mixer_3 {MIXER_BASIC}

NF=9.49dB [TXPANF]

G=14dB [TXPAG]

RFAmp_4 {RFAMP}

IP1dB=100dBm [TPAKIP1dB]

L=0dB [TXPAKG+1]

Attn_1 {ATTN_NonLinear}

Fhi=86.5GHz

Flo=80.5GHz

N=5

IL=0.01dB


ZO=50Ω

TRANSMITTERROUT {*OUT}

NF=5dB [TXIFAMPNF]

G=7dB [TXIFAMPG]

RFAmp_3 {RFAMP}

IP1dB=16dBm [TXIFATTIP1dB]

L=13dB [TXIFATTG]

Attn_2 {ATTN_NonLinear}

Pwr=7dBm

F=60750MHz [3*FTXIF]

LO4 {PwrOscillator}

TRANSMITTER

RF/IF First Frequency Conversion – SystemVue SpectraSys Models

R I

L

IP1dB=5dBmLO=7dBm

ConvGain=0dBMixer_1 {MIXER_BASIC}

PhaseN=(4) [-60; -80; -100; -96]dB

Pwr=7dBmF=60750MHz [3*FTXIF]LO7 {PwrOscillator}

Fhi=20375MHz [FTXIF+CS/2]

Flo=20125MHz [FTXIF-CS/2]N=3

IL=0.01dB


OP1dB=8dBm [RXIFOP1dB]NF=10dB [RXIFNF]

G=4dB [RXIFG]RFAmp_2 {RFAMP}

OP1dB=4dBm

NF=26.16dBG=0dB

RFAmp_5 {RFAMP}

ZO=50ΩRECEIVEROUT {*OUT}

Fhi=86.5GHz

Flo=80.5GHzN=5

IL=0.01dB


IP1dB=100dBm

L=1dBAttn_4 {ATTN_NonLinear}

Fhi=86GHzFlo=81GHz

N=3IL=0.01dB

BPF_Butter_6 {BPF_BUTTER}Source1=FTXRF MHz at -12 dBm, BW: CS MHzTX_IF_IN {MultiSource}

OP1dB=4dBm [RXLNAOP1dB]

NF=8dB [RXLNANF]G=22dB [RXLNAG]RFAmp_1 {RFAMP}

RECEIVER



26

Rx Front End Modeling – Direct Conversion

LNA81-86GHz

EBAND

2nd IF

PRIMARY LO

IFAI-channel

Q-channelIL=3.02dB

Split2_1 {SPLIT2}

M

D

InDrv=13dBm

HL=0dB [zzz;0;0;0,0]DIV=4

CG=0dB

Disabled: OPENFreqDiv_1 {FREQ_DIV}

IL=3.02dBSplit2_2 {SPLIT2}

OP1dB=60dBm

NF=3dB [IFA_NF]G=20dB [IFA_GAIN]

IFA1 {RFAMP}

IL=3.02dB

Split2_3 {SPLIT2}

A=90°

PhaseShift_1 {PHASE}

Fhi=19.25GHz [(VCOTUNE_FREQ/4)+RFBW/2]

Flo=14.25GHz [(VCOTUNE_FREQ/4)-RFBW/2]

N=3IL=0.01dB

BPF_Xband1 {BPF_BUTTER}

OP1dB=10dBm

NF=5dBG=20.5dB

LNA1 {RFAMP}

Fpass=1GHz [IF2_FILTERBW]

N=3

IL=0.01dBLPF_Baseband {LPF_BUTTER}

Source2=EBAND_FREQ GHz at EBAND_PWR dBm, w Phase Noise

Disabled: OPEN

RF_PhNoise {MultiSource}

Source1=EBAND_FREQ GHz at -60 dBm, BW: 2.4 GHz

Disabled: OPEN

RF_wide {MultiSource}

Source1=EBAND_FREQ GHz at EBAND_PWR dBmRF_in {MultiSource}

ZO=50Ω

IF2_Out_I {*OUT}

Fpass=1GHz [IF2_FILTERBW]

N=3IL=0.01dB

LPF_Baseband1 {LPF_BUTTER}

R I

L

IP1dB=0dBmNF=7dB

LO=7dBm

SUM=DifferenceConvGain=0dB

DOWNMIX2BQ {MIXER_BASIC} ZO=50Ω

IF2_Out_Q {*OUT}

R I

L

IP1dB=20dBmNF=7dB

LO=7dBm

SUM=Difference

ConvGain=0dBDOWNMIX1 {MIXER_BASIC}

Fhi=19.25GHz [(VCOTUNE_FREQ/4)+RFBW/2]

Flo=14.25GHz [(VCOTUNE_FREQ/4)-RFBW/2]

N=3IL=0.01dB

BPF_Xband {BPF_BUTTER}

Fhi=86.25GHz [(5*VCOTUNE_FREQ/4)+RFBW/2]

Flo=81.25GHz [(5*VCOTUNE_FREQ/4)-RFBW/2]

N=3IL=0.01dB

BPF_mmWave {BPF_BUTTER}

EnablePN=NO

Pwr=13dBm [VCOTUNE_PWR+6]

F=67GHz [VCOTUNE_FREQ]VCO_LO3 {PwrOscillator}

EnablePN=NO

Pwr=13dBm [VCOTUNE_PWR+6]

F=16.75GHz [VCOTUNE_FREQ/4]VCO_LO6 {PwrOscillator}

R I

L

IP1dB=0dBm

NF=7dBLO=7dBm

SUM=Difference

ConvGain=0dB

DOWNMIX2BI {MIXER_BASIC}



27

Key system questions

• Do the baseband PHY algorithms work with the RF?

• What are the root causes of each problem?

• Which domain is the best place to address these problems?

• Can you correlate measurements with simulations?

• If not, what assumptions are “baked into the results”?

? ? ? E-Band Webcast , March 2014


28

System Level Modeling – integrated approach

E-band

Source E-band Golden Transmitter

Spe c t rum A nalyzer

SpecRxIn


SpecRxOut

123SigRxIn

123SigRxOut

NoiseDensity

Noise_Power

Attenuation

Attenuation=54 [Atten]

Subnetwork1 {Attenuation}

1 1 0 1 0

DataPattern=PN15

B2 {DataPattern@Data Flow Models}

TEST

REF


Re

ImFc

CxEnv

E2 {EnvToCx@Data Flow Models}


SpecRxWnoi

ModOUT

QUADOUT

FreqPhaseQ

IAmp

InitialPhase=0°

FCarrier=81e+9Hz [FCarrier]

InputType=I/Q

RF_Link

SYS

BPSK, QPSK, ..., up to 4096-QAM8-PSK, 16-PSK, 16-APSK, 32-APSK

16-Star QAM , 32-Star-QAM, and Cus tom APSK

Data Pay loadPreambleId le

Fram e Struc ture

Spreading Code

Generator

X

D ig it a l M ode m Source

f o r L ine a r M odu lation

DSSS Sy s tem

Payload_ModType=16-QAM [Payload_ModType]

Preamble_ModType=BPSK [Preamble_ModType]

Dec ision

Dev ice

Feedward

Fi l ter-

-

Feedback

Fi l ter

Dec is ion Feedbac k Equal izer

Fas t Com putation Algori thm

CIR--->DFE c oeffic ients

D ig it a l M ode m R e ceiver

TrackingAlgorithm=LMS

FreqSync_Mode=CIR Corr

FrameSync_Algorithm=DiffCorr

{DigMod_ReceiverL_FastDFE}

R I

L

IP1dB=5dBmLO=7dBm


PhaseN=(4) [-60; -80; -100; -96]dB




IL=0.01dB




OP1dB=4dBm

NF=26.16dBG=0dB

RFAmp_5 {RFAMP}


Fhi=86.5GHz

Flo=80.5GHzN=5

IL=0.01dB


IP1dB=100dBm


Fhi=86GHzFlo=81GHz

N=3IL=0.01dB





SpecRxIn


SpecRxOut

123SigRxIn

123SigRxOut

NoiseDensity

Noise_Power

Attenuation



1 1 0 1 0

DataPattern=PN15


TEST

REF


Re

ImFc

CxEnv



SpecRxWnoi

ModOUT

QUADOUT

FreqPhaseQ

IAmp

InitialPhase=0°


InputType=I/Q

RF_Link

SYS




Fram e Struc ture

Spreading Code

Generator

X



DSSS Sy s tem



Dec ision

Dev ice

Feedward

Fi l ter-

-

Feedback

Fi l ter









E-band

Receiver

BER &

FER



29

System Level Modeling


SpecRxIn


SpecRxOut

123SigRxIn

123SigRxOut

NoiseDensity

Noise_Power

Attenuation



1 1 0 1 0

DataPattern=PN15


TEST

REF


Re

ImFc

CxEnv



SpecRxWnoi

ModOUT

QUADOUT

FreqPhaseQ

IAmp

InitialPhase=0°


InputType=I/Q

RF_Link

SYS




Fram e Struc ture

Spreading Code

Generator

X



DSSS Sy s tem



Dec ision

Dev ice

Feedward

Fi l ter-

-

Feedback

Fi l ter









E-band Receiver Sensitivity Measurements

E-band

Source

E-band

Receiver

BER &

FER

E-band Golden Transmitter

DATAFLOW

SIMULATION

R I

L

IP1dB=5dBmLO=7dBm


PhaseN=(4) [-60; -80; -100; -96]dB




IL=0.01dB




OP1dB=4dBm

NF=26.16dBG=0dB

RFAmp_5 {RFAMP}


Fhi=86.5GHz

Flo=80.5GHzN=5

IL=0.01dB


IP1dB=100dBm


Fhi=86GHzFlo=81GHz

N=3IL=0.01dB




RF SYSTEM

ANALYSIS

SpectraSys Receiver Model



30

Transmit Chain Analysis - EVM

E-Band Reference Source

• Randomized payload data generation

• Baseband transmission source

• Generalized data frame

• Multiple type of modulation

Up-Converter Impairments

• D-to-A converter level and range

• LO phase noise

• Modulator gain and phase imbalance

Tx Non-linearity and channel

• Fast circuit envelop co-simulation

• ADS co-sim

• White Gaussian noise

Measurement and waveform download

• EVM, CCDF, Spectrum measurement

• Waveform download into wide band AWG

• Vector signal analysis

BPSK, QPSK, ..., up to 4096-QAM

8-PSK, 16-PSK, 16-APSK, 32-APSK16-Star QAM, 32-Star-QAM,

and Custom APSK

Data PayloadPreambleIdle

Frame Structure

Spreading CodeGenerator

X


DSSS System

ModOUT

QUADOUT

FreqPhaseQ

IAmp

NoiseDensity

FastCircuitEnvelope

File=82GHz_pa_fce_1.fce

VSA_89600B_Sink

EVM

CCDF

ADS Cosim

1 1 0 1 0



31

Fast Circuit Envelope Simulation with Jitter Injection

NoiseDensity

Noise_Power

Fc

CxEnv


FastCircuitEnvelope

File=LNA_75G_Vmax_0_05_BW_10G.fce

F1 {FastCircuitEnvelope@Data Flow Models}

SYSTEMVUE

E-band circuit

envelope model

@75GHz

GoldenGate

Modulation Envelope is

Represented in the Time Domain

Carrier is Represented in the

Frequency Domain

Circuit Envelope Simulation

TEST

REF


BER without jitter : 0.0

BER with jitter : 0.37

FCE nonlinear model advantages:

• Direct behavioral-modeling extraction method

• Convenient use model

• Fast & accurate system simulations: - RF Nonlinearities

- Freq Response Variations

- Memory Effects

- Noise spectral density (NEW)

- Frequency conversion between pins

• Supports multiple I/O pins and internal nodes

At lower frequencies, X-parameters

are also a good modeling option



32

RF INTEGRATED


ADVANCED MODEM

TECHNOLOGIES 2

SYSTEM-LEVEL

DESIGN COCKPIT

E-BAND SYSTEM

OVERVIEW 1

HARDWARE &

MEASUREMENTS 4




33

Bit Error Rate Easy math but difficult test bench implementation

𝑷𝒆 =𝟏

𝟐(𝟏 − 𝒆𝒓𝒇) 𝑬𝒃/𝑵𝒐

Probability of error

VAR = (1 - Pe) / (Pe × N) Variance of Pe N: transmitted bits

# of error bits received

# of transmitted bits Bit error rate BER =



34

BER Measurement Considerations

• Transmit enough number of bits

- for a Pe of 10 -6 , with a relative variance of 0.01, a sample size N of

approximately 10 8 bits is required

• Use PRBS to avoid deterministic jitter

- Use a randomizer or scrambler in transmitter

- Make sure all constellation points are approximately equally

populated

• Synchronize test and reference signals

TEST

REF


… 0 1 1 0 0 1 0 0 1 …

… 0 1 0 1 1 0 1 0 0 0 1 …

t0 t1 t2

… 0 1 1 0 0 1 0 0 1 …

wrong sync:

correct sync:



35

E-Band RFIC Loopback BER Measurement

• Wider BW (8 GHz BW)

• Higher Sampling (40 GSa/s)

DSA90804A-803-200 Digital Signal Analyzer

SYSTEMVUE

TEST

REF


BPSK, QPSK, ..., up to 4096-QAM

8-PSK, 16-PSK, 16-APSK, 32-APSK16-Star QAM, 32-Star-QAM,

and Custom APSK

Data PayloadPreambleIdle

Frame Structure

Spreading CodeGenerator

X


DSSS System



Decision Device

FeedwardFilter

-

-

FeedbackFilter

Decision Feedback Equalizer

Fast Computation Algorithm

CIR--->DFE coefficients






RFIC DUT

BBIQRF

M8190A 12 GSa/S Arbitrary

Waveform Generator

I

Q Automatic waveform

creation & download

E-band Golden Source

E-band Reference Receiver

BER/FER Measurement

Custom modem

design

Replaceable

Digital modem

library

: C++, .m or SV DSP parts formats



36

M9703A AXIe 12-bit High-Speed

Digitizer/Wideband Digital Receiver

Interleaving to get 4ch @ 3.2 GSa/s

BBIQRF I

Q

System Level Verification


SpecRxIn


SpecRxOut

123SigRxIn

123SigRxOut

NoiseDensity

Noise_Power

Attenuation



1 1 0 1 0

DataPattern=PN15


TEST

REF


Re

ImFc

CxEnv



SpecRxWnoi

ModOUT

QUADOUT

FreqPhaseQ

IAmp

InitialPhase=0°


InputType=I/Q

RF_Link

SYS




Fram e Struc ture

Spreading Code

Generator

X



DSSS Sy s tem



Dec ision

Dev ice

Feedward

Fi l ter-

-

Feedback

Fi l ter









Leveraging test vectors out of the simulation environment

E-band

Source

E-band

Receiver

BER &

FER

E-band Golden Transmitter

DATAFLOW

SIMULATION

RFIC DUT

BBIQRF RFBBIQ I

Q

I

Q



37

A Top-down Design & Verification Flow Consistent EVM, BER measurement across RF/BB domains, and throughout the R&D lifecycle

FPGA

Synthesis

Handwritten

HDL

Custom IP

Algorithms

C++, .m

MEASUREMENT, ANALYSIS

VSA software FlexDCA software

DIGITAL BITS, or MODULATED CARRIERS

MXG / ESG

Infiniium Scope

Logic Analyzer

MXA / PXA

Wideband AWG Wideband Digitizer

Target-neutral

HDL Generation

System design

RF Architecture

Baseband design

PHY Reference

HDL Simulator(s)

SIMULATED H/W

Dataflow Simulation

.m/C++ ALGORITHM

.bit

Files FPGA Target

REAL HARDWARE



38

http://images.google.com/imgres?imgurl=http://pixhost.eu/avaxhome/avaxhome/2008-03-22/ModelSim6.0.jpg&imgrefurl=http://freebooksource.com/?p=7631&h=349&w=450&sz=35&hl=en&start=6&sig2=7-s0hqg0LU4SkJ5WBWEacQ&usg=__hNoKfEeDy-m1uYKoF8LdEzwyO9s=&tbnid=47cfFwpXxR0y6M:&tbnh=98&tbnw=127&ei=07L8SIDpMp3aNLHW1SM&prev=/images?q=modelsim&gbv=2&hl=en

Design and Verification Challenges

SystemVue Dataflow , Reference IP,

RF-DSP co-verification, Links to Test

System–level simulation

Math algorithms, C++,

VHDL/Verilog, SystemC,

Hardware-in-Loop

Baseband Algorithm

Modeling, Implementation

Design & Test

Integration

3rd Party API

Scenarios, Embedding

Automation

RF System Arch.

IC/Board/Module

Physical Design

Pkg/Connectors

3DEM Antenna

Concurrent approach

Earlier validation

Higher confidence

ECOSYSTEM, REQUIREMENTS



39

Conclusions

Confidentiality Label

Allows concurrent development and joint RF/BB architectures

for earlier, more robust E-Band Systems.

E-Band measurements are challenging, so using simulation to fill

gaps and leverage test-vectors can reduce costs, re-use assets

A system-level approach integrating both RF & BB reduces

guesswork in setting specifications and produces higher-performing

system designs.

The industry is moving into very challenging areas. A step forward

is needed for concurrent DSP, RF, and T&M design lifecycle

methodologies, not just raw technology.

“Golden Reference” blocks and open, system-level modeling

alongside RF effects provides earlier confidence in modem

algorithm development, reducing net overall verification effort

RF INTEGRATED


ADVANCED MODEM

TECHNOLOGIES 2

SYSTEM-LEVEL

DESIGN COCKPIT

E-BAND SYSTEM

OVERVIEW 1

HARDWARE &

MEASUREMENTS 4



40

Questions? - Connect with Us

Visit Agilent’s websites at

• agilent.com/find/systemvue

• agilent.com/find/lte-advanced

• www.agilent.com/find/5g

to see the latest solutions that can be used for 4G and 5G

Specifically featured: W1902 Digital Modem library

• cp.literature.agilent.com/litweb/pdf/5991-3123EN.pdf



41

http://www.agilent.com/find/lte



http://www.agilent.com/find/lte-advanced




http://www.agilent.com/find/5g

http://cp.literature.agilent.com/litweb/pdf/5991-3123EN.pdf





Documents

E-Band Wireless Backhaul: System Design & Test Challenges · PDF fileE-Band Wireless Backhaul: System Design & Test Challenges Presented by: Agilent Technologies, Inc. Daren Sang McClearnon