Color figures from Rappaport et al., Millimeter Wave Wireless

Color figures from Rappaport et al., Millimeter Wave Wireless

Communications (ISBN‐13: 9780132172288)

Note: Only selected figures in the book are available in color. Black and white figures have been omitted from this file.

From Rappaport et al., Millimeter Wave Wireless Communications, ISBN-13: 978-0-13-217228-8. Copyright © 2015 Pearson Education, Inc.

PTG-Rappaport Rappaport Ch01 2014/8/12 19:47 Page 4 #4

4 Chapter 1 Introduction

Cellular, 50 MHz,1983

PCS, 150 MHz,1995

UNII, 300 MHz, 1997

LMDS, 1300 MHz, 1998

60 GHz Unlicensed, 5000 MHz, 1998

Figure 1.1 Areas of the squares illustrate the available licensed and unlicensed spectrum bandwidthsin popular UHF, microwave, 28 GHz LMDS, and 60 GHz mmWave bands in the USA. Other countriesaround the world have similar spectrum allocations [from [Rap02]].



1.1 The Frontier: Millimeter Wave Wireless 5

Figure 1.2 Wireless spectrum used by commercial systems in the USA. Each row represents a decade in frequency. For example, today’s 3G and 4G cellular and WiFi carrier frequencies are mostlyin between 300 MHz and 3000 MHz, located on the fifth row. Other countries around the worldhave similar spectrum allocations. Note how the bandwidth of all modern wireless systems (throughthe first 6 rows) easily fits into the unlicensed 60 GHz band on the bottom row [from [Rap12b] U.S.Dept. of Commerce, NTIA Office of Spectrum Management].

AM radio

FM radio

Wi-Fi

3G / 4G LTEcellular

TV broadcast

28 GHz – LMDS(5G cellular)

Active CMOS IC research

77 GHzvehicular

radar

60 GHz unlicensedWiGig (802.11 ad)

UNITEDSTATESFREQUENCY ALLOCATIONSTHE RADIO SPECTRUM

38 GHz5G cellular




Operating frequency (GHz)

Exp

ecte

d at

mos

pher

ic lo

ss (

dB/k

m)

010−2

10−1

100

101

102

50 100 150 200 250 300 350 400

Figure 1.3 Expected atmospheric path loss as a function of frequency under normal atmosphericconditions (101 kPa total air pressure, 22 Celsius air temperature, 10% relative humidity, and0 g/m3 suspended water droplet concentration) [Lie89]. Note that atmospheric oxygen interactsstrongly with electromagnetic waves at 60 GHz. Other carrier frequencies, in dark shading, exhibitstrong attenuation peaks due to atmospheric interactions, making them suitable for future short-range applications or “whisper radio” applications where transmissions die out quickly with distance.These bands may service applications similar to 60 GHz with even higher bandwidth, illustrating thefuture of short-range wireless technologies. It is worth noting, however, that other frequency bands,such as the 20-50 GHz, 70-90 GHz, and 120-160 GHz bands, have very little attenuation, well below1 dB/km, making them suitable for longer-distance mobile or backhaul communications.




2.5 mm

3.5 mm

Figure 1.4 Block diagram (top) and die photo (bottom) of an integrated circuit with four transmitand receive channels, including the voltage-controlled oscillator, phase-locked loop, and local oscillatordistribution network. Beamforming is performed in analog at baseband. Each receiver channel containsa low noise amplifier, inphase/quadrature mixer, and baseband phase rotator. The transmit channel alsocontains a baseband phase rotator, up-conversion mixers, and power amplifiers. Figure from [TCM+11],courtesy of Prof. Niknejad and Prof. Alon of the Berkeley Wireless Research Center [ c© IEEE].




Figure 1.5 Third-generation 60 GHz WirelessHD chipset by Silicon Image, including the SiI6320 HRTX Network Processor, SiI6321 HRRX Network Processor, and SiI6310 HRTR RF Transceiver. These chipsets are used in real-time, low-latency applications such as gaming and video, and provide 3.8 Gbps data rates using a steerable 32 element phased array antenna system (courtesy of Silicon Image) [EWA+11] [©c IEEE].




1990

III-V HBT

Si CMOSSiGe HBT

III-V HEMT

0

100

200

300

400

f T (

GH

Z)

500

600

700

800

1995 2000Year

2005 2010

Figure 1.6 Achievable transit frequency (fT ) of transistors over time for several semiconductortechnologies, including silicon CMOS transistors, silicon germanium heterojunction bipolar transistor(SiGe HBT), and certain other III-V high electron mobility transistors (HEMT) and III-V HBTs.Over the last decade CMOS (the current technology of choice for cutting edge digital and analogcircuits) has become competitive with III-V technologies for RF and mmWave applications [figurereproduced from data in [RK09] c© IEEE].




Figure 1.7 Wireless personal area networking. WPANs often connect mobile devices such as mobilephones and multimedia players to each other as well as desktop computers. Increasing the data-ratebeyond current WPANs such as Bluetooth and early UWB was the first driving force for 60 GHzsolutions. The IEEE 802.15.3c international standard, the WiGig standard (IEEE 802.11ad), and theearlier WirelessHD standard, released in the 2008–2009 time frame, provide a design for short-rangedata networks (≈ 10 m). All standards, in their first release, guaranteed to provide (under favorablepropagation scenarios) multi-Gbps wireless data transfers to support cable replacement of USB, IEEE1394, and gigabit Ethernet.




Figure 1.8 Multimedia high-definition (HD) streaming. 60 GHz provides enough spectrum resourcesto remove HDMI cables without sophisticated joint channel/source coding strategies (e.g., compres-sion), such as in the wireless home digital interface (WHDI) standard that operates at 5 GHzfrequencies. Currently, 60 GHz is the only spectrum with sufficient bandwidth to provide a wirelessHDMI solution that scales with future HD television technology advancement.




wired network

WLAN access point

Figure 1.9 Wireless local area networking. WLANs, which typically carry Internet traffic, are a popu-lar application of unlicensed spectrum. WLANs that employ 60 GHz and other mmWave technologyprovide data rates that are commensurate with gigabit Ethernet. The IEEE 802.11ad and WiGigstandards also offer hybrid microwave/mmWave WLAN solutions that use microwave frequencies fornormal operation and mmWave frequencies when the 60 GHz path is favorable. Repeaters/relays willbe used to provide range and connectivity to additional devices.




Figure 1.10 Wireless backhaul and relays may be used to connect multiple cell sites and subscriberstogether, replacing or augmenting copper or fiber backhaul solutions.




700 MHz

BandUplink(MHz)

Downlink(MHz)

Carrier bandwidth(MHz)

AWS

GSM 900

GSM 1800

PCS 1900

Cellular 850

Digitaldividend

UMTS core

IMTextension

746–763

1710–1755

880–915

1710–1785

1850–1910

824–849

470–854

1920–1980

2500–2570

776–793

2110–2155

925–960

1805–1880

1930–1990

869–894

2110–2170

2620–2690

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

1.25 5 10 15 20

Figure 1.11 United States spectrum and bandwidth allocations for 2G, 3G, and 4G LTE-A (long-termevolution advanced). The global spectrum bandwidth allocation for all cellular technologies does notexceed 780 MHz. Currently, allotted spectrum for operators is dissected into disjoint frequency bands,each of which possesses different radio networks with different propagation characteristics and buildingpenetration losses. Each major wireless provider in each country has, at most, approximately 200 MHzof spectrum across all of the different cellular bands available to them [from [RSM+13] c© IEEE].




Interferingbasestation

LOScommunication

mmWaverelay

mmWavebackhaul

Servingbasestation

NLOScommunication

Figure 1.12 Illustration of a mmWave cellular network. Base stations communicate to users (andinterfere with other cell users) via LOS, and NLOS communication, either directly or via heteroge-neous infrastructure such as mmWave UWB relays.




SIRCIM 6.0: Impulse response parameters

Nor

mal

ized

line

ar p

ower

00

0.2

0.4

0.6

0.8

1

100 200 300Excess delay (nanoseconds)

400 5000

0.1

0.2

Distan

ce (m

eters)

Path loss reference distance = 1 meterTopography = 20% obstructed LOSRMS delay spread = 65.9 nanosecocndsOperating frequency = 60.0 GHzBuilding plan = open

Figure 1.13 Long delay spreads characterize wideband 60 GHz channels and may result in severeinter-symbol interference, unless directional beamforming is employed. Plot generated with Simu-lation of Indoor Radio Channel Impulse Response Models with Impulse Noise (SIRCIM) 6.0 [from[DMRH10] c© IEEE].



1.3 Emerging Applications of MmWave Communications 21

Signalling rate (Gbps)00.1

1

10

5 10 15

Cost (3.5 m)

Optics densityalready better**

2004 2007±1 2010±1 2013±2

Optics better

Copper betterCost (10 m)

Rat

io o

f opt

ical

to e

lect

rical

per

form

ance

Power

20

Figure 1.14 Comparison between optical and electrical performance in terms of cost and powerfor short cabled interconnects. The results show that optical connections are preferred to electricalcopper connections for higher data rates, assuming wires are used [adapted from [PDK+07] c© IEEE].




Figure 1.15 MmWave wireless will enable drastic changes to the form factors of today’s computingand entertainment products. Multi-Gbps data links will allow memory devices and displays to becompletely tetherless. Future computer hard drives may morph into personal memory cards and maybecome embedded in clothing [Rap12a][Rap09][RMGJ11].




Figure 1.16 Future users of wireless devices will greatly benefit from the pervasive availability ofmassive bandwidths at mmWave frequencies. Multi-Gbps data transfers will enable a lifetime of con-tent to be downloaded on-the-fly as users walk or drive in their daily lives [Rap12a][Rap09][RMGJ11].




cloud

10-50 Gbpslinks

Hundreds ofwireless post-it

notes

Desk

OfficeBookshelf

10-50 Gbps links

Passivewireless-to-fiber node

1-10 Tbps link1-10 Tbps link

Desktopcomputer with

wirelessmemory

connectivity

Desk

Hundreds ofwireless post-it

notes withbooks/movies

OfficeBoo

kshe

lf

Passive wireless-to-fiber node

Figure 1.17 The office of the future will replace wiring and wired ports with optical-to-RF inter-connections, both within a room and between rooms of a building. UWB relays and new distributedwireless memory devices will begin to replace books and computers. Hundreds of devices will beinterconnected with wide-bandwidth connections through mmWave radio connections using adap-tive antennas that can quickly switch their beams [Rap11] [from [RMGJ11] c© IEEE].




Vehicle-to-vehicle communication

Vehicular radar

Vehicular-to-infrastructure communication

Figure 1.18 Different applications of mmWave in vehicular applications, including radar, vehicle-to-vehicle communication, and vehicle-to-infrastructure communication.




seat back entertainment wireless Internet access

Figure 1.19 Different applications of mmWave in aircraft including providing wireless connec-tions for seat-back entertainment systems and for wireless cellular and local area networking. Smartrepeaters and access points will enable backhaul, coverage, and selective traffic control.



92 Chapter 2 Wireless Communication Background

Application Layer

Data LinkLayer

LogicalLink

Control

MAC

Physical Layer

Hardware Layer

Network Layer

Transport Layer

Session Layer

Presentation Layer

Figure 2.32 Reference system architecture for a communication network. The Physical Layer(PHY) is considered the lowest layer, and the Application Layer is the highest layer. We proposea new layer, called the Hardware Layer, that is below PHY, in order to account for complexitiesinvolved with the creation of new hardware and devices for mmWave communications.



100 Chapter 3 Radio Wave Propagation for MmWave

100

10

1

0.1

0.010 20

Sea

leve

l atte

nuat

ion

(dB

/km

)

40 60 80 100 120 140 160 180 200

Frequency (GHz)

220 240 260 280 300 320 340 360 380 400

Figure 3.1 The attenuation (dB/km) in excess of free space propagation due to absorption in air atsea level across the sub-terahertz frequency bands. The far left (unshaded) bubble shows extremelysmall excess attenuation in air for today’s UHF and microwave consumer wireless networks, and otherbubbles show interesting excess attenuation characteristics that are dependent on carrier frequency[from [RMGJ11], c© IEEE].



3.2 Large-Scale Propagation Channel Effects 107

100

50

20

10

5

2

1

0.5

0.2

0.1

0.05

0.02

0.011 2 5 10 20 50

Frequency (GHz)

Spec

ific

atte

nuat

ion

(dB

/km

)

100 200 500 1,000

Heavy rainfall 8 GHz1.4 dB attenuation @ 200 m

150 mm/h

100 mm/h

50 mm/h

25 mm/h

5 mm/h

1.25 mm/h

0.25 mm/h

Figure 3.2 Rain attenuation as a function of frequency and rain rate in the mmWave spectrum[from [AWW+13][RSM+13][ZL06] c© IEEE].




Figure 3.5 Indoor penetration measurements at 72 GHz in a building in Brooklyn, New York. TheTX location is marked by a triangle, the RX locations are shown as numbered dots. The primary raypaths for signal penetration are shown with arrows [reproduced from [NMSR13] c© IEEE].




Path 2

Zone 2Zone 2

Zone 1

d1 d2

Receiver

Receiver

Obstacle

Transmitter

Transmitter

Path 1Path 3

Path 2 is l/2 longer than path 1.Path 3 is l/2 longer than path 2.

Figure 3.6 Example of a diffraction object blocking the line-of-sight (LOS) path between trans-mitter and receiver. At millimeter wave frequencies, objects such as trees and people may inducefading and scattering as they move.



LOS: σ = 6.63dBNLOS: σ = 9.45dBnLOS = 3.76

nNLOS = 4.69


nNLOS = 4.51


nNLOS = 4.30


nNLOS = 4.08

120

(a)

(b)

(c)

(d)

n = 5

n = 4

n = 3

nNLOS = 4.69

nLOS = 3.76

Path loss corresponding to one best power versus distance

TX-RX Separation (m)

Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.51

nLOS= 3.57

Path loss corresponding to non-coherently combined two best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.30

nLOS= 3.28

Path loss corresponding to coherently combined two best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.08

nLOS= 3.03

Path loss corresponding to coherently combined three best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

Figure 3.11 Scatter plots of measured 28 GHz cellular path loss in New York City[SR13][RSM+13][AWW+13][SWA+13]. The plots illustrate the reduction in path loss that can beachieved when a mobile handset using 10 steerable beams combines individual multipath signalsarriving at different angles from the same transmitter. In (a), the single best beam pointing direc-tion is used to make a link at each RX location. In (b), the two best beam pointing directionsare non-coherently combined (where the powers in each unique beam are simultaneously added). In(c) and (d), the two and three best beams, respectively, are coherently added (where the total voltagein each unique beam is simultaneously added and then squared to produce power).



nNLOS = 4.69


nNLOS = 4.51


nNLOS = 4.30


nNLOS = 4.08

120

(a)

(b)

(c)

(d)

n = 5

n = 4

n = 3

nNLOS = 4.69

nLOS = 3.76

Path loss corresponding to one best power versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.51

nLOS= 3.57

Path loss corresponding to non-coherently combined two best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.30

nLOS= 3.28

Path loss corresponding to coherently combined two best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

120

n = 5

n = 4

n = 3

nNLOS= 4.08

nLOS= 3.03

Path loss corresponding to coherently combined three best powers versus distance


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

100

100105

80

60

40

20

0

Figure 3.11 Scatter plots of measured 28 GHz cellular path loss in New York City[SR13][RSM+13][AWW+13][SWA+13]. The plots illustrate the reduction in path loss that can beachieved when a mobile handset using 10 steerable beams combines individual multipath signalsarriving at different angles from the same transmitter. In (a), the single best beam pointing direc-tion is used to make a link at each RX location. In (b), the two best beam pointing directionsare non-coherently combined (where the powers in each unique beam are simultaneously added). In(c) and (d), the two and three best beams, respectively, are coherently added (where the total voltagein each unique beam is simultaneously added and then squared to produce power).




Lobe 2

Lobe 1

120°90°

Track position 1

Track position 10

60°

30°

330°

300°270°

240°

210°

180°

150°

0°

120°90°

60°

30°

330°

300°270°

240°

210°

180°

150°

0°

Track position 5

120°90°

60°

30°

330°

300°270°

240°

210°

180°

150°

0°

Track position 21

120°90°

60°

-72-82-92-75-85-95

30°

330°

300°

Threshold270°

240°

210°

180°

150°

0°-75-85-95 -75-85-95

Figure 3.12 Four polar plots of 28 GHz propagation at track positions 1, 5, 10, and 21 along a21-step linear track with λ/2 step sizes show two lobes of received power across azimuth. Measure-ments are for a partially obstructed NLOS RX environment in downtown Brooklyn using 24.5 dBihorn antennas at both the TX and RX. The TX was placed on the rooftop of NYU’s Rogers Hall135 m away from the RX. Each dot represents the received power level at a particular RX azimuthangle. For NLOS RX locations, a threshold of 20 dB below maximum power level was defined for athreshold (shown as a solid-line circle) to determine lobe statistics, whereas 10 dB was used for theLOS threshold [reproduced from [SWA+13] c© IEEE].



3.7 Outdoor Channel Models 141

−100

−110

−120

−130

−140

−150

−160

−200 −100 100Frequency (MHz)

Cha

nnel

pow

er g

ain

|H(f

)|2 (d

B)

TX location WRW-A RX location 14channel frequency response track 7

Relative threshold = 24 dBRX Atten = 0dBDistance = 70 mTX: 15°/−2.5°

RX: 0°/2.5°

Measurement 125 track 7

2000

−90

Figure 3.13 Frequency-selective fading occurs about the 38 GHz carrier frequency in outdoor urbanNLOS channels. Note the periodic 50 MHz fades in frequency about the carrier correspond to a RMSdelay spread that is approximately 20 ns. Here we see a channel that has deep fades as low as 30 dBfrom the peak channel gain.




1

0.9On the campus of the university of Texas at AustinTX on roof of building WRW37.625 GHz carrier1 MHz sub-bands condiered +/- 222 MHz from carrier46.1dBm EIRP25 dBi 7° Beamwidth TX Antenna25 dBi 7° Beamwidth or 13.3 dBi dBi 38°Beamwidth RX AntennaLink distances 61 - 265 m

CDF of frequency selectivity outdoor 38 GHz cellular channel

0.8

0.7

0.6

Cum

ulat

ive

dist

ribu

tion

func

tion

0.5

0.4

0.3

0.2

−40 −30 −20 −10 01 MHz sub-band deviation from mean channel gain (dB)

10 20

0.1

0

Figure 3.14 When a channel frequency representation such as that shown in Fig. 3.13 is consideredover 1 MHz subbands (i.e., we evaluate the average channel gain at 1 MHz intervals and comparethese small intervals to the overall average channel gain across the band), we see that fading isnot severe for outdoor urban cellular mmWave channels. The time delay spread and the number ofresolvable multipath components directly contribute to the fading characteristics across the occupiedspectrum. Directional antennas change small-scale fading from today’s common Rayleigh fadingcharacteristics (for omnidirectional antennas) into much narrower fade depths over much widerfrequency bands.




0.9LOS 38 GHz

LOS 60 GHz

NLOS 38 GHz

NLOS 60 GHz

0.8

0.7

0.6

0.5

0.4

0.3

Prob

abili

ty R

MS

dela

y sp

read

< a

bsci

ssa

0.2

0.1

00 20 40 60

RMS delay spread (ns)

80 100

E[σLOS38]=1.2ns

E[σNLOS38]=23.6ns

E[σNLOS60]=0.8ns

E[σNLOS60]=7.4ns

Max[σLOS38]=1.3ns

Max[σNLOS38]=122ns

Max[σNLOS60]=0.9ns

Max[σNLOS60]=36.6ns

120 140

1

Figure 3.15 Differences in RMS delay spread and their distribution at 38 and 60 GHz in variousoutdoor environments [from [RBDMQ12] c© IEEE].




1

0.9

0.8

0.7

0.6Pr

obab

ility

RM

S de

lay

spre

ad <

abs

ciss

a0.5

0.4

0.3

0.2

0.1

00 20 40 60 80 100


LOS(25dBi RX Ant.)

LOS(13.3dBi RX Ant.)

NLOS(13.3dBi RX Ant.)

NLOS(25dBi RX Ant.)

120 140 160 180 200 220 240

E[σLOS25]=1.5ns

E[σNLOS25]=14.3ns

E[σLOS13]=1.9ns

E[σNLOS13]=13.7ns

Max[σLOS25]=15.4ns

Max[σNLOS25]=225ns

Max[σLOS13]=15.5ns

Max[σNLOS13]=166ns

Figure 3.16 Greater transmitter antenna heights resulted in decreased 90% RMS delay spreadcompared with situations in which the 38 GHz transmitter is near the ground, and the worst-caseRMS delay spread was found to be 225 ns on a Texas college campus using the tallest transmitterantenna height [from [RSM+13] [RGBD+13] c© IEEE].




180

150

120

90

−30

−60

−90

−120

−150

−180

−180 −150 −120 −90 −60 −30 30 90 150 180

TX

RX

Positiveangles

Positiveangles

Negativeangles

Negativeangles

TX azimuth angleRX

azi

mut

h an

gle

0

60

30

0 60 120

38GHz60GHz

Figure 3.17 MmWave applications in which the transmitter and receiver are close to the ground(such as peer-to-peer or vehicle-to-vehicle) will provide a wide distribution of angles at which linksmay be established [from [RBDMQ12][RGBD+13] c© IEEE].




100

150

120

90

30

0−30 30 60 90 120 150 180

Fewer negativeRX azimuth linksbecause of ENSbuilding to leftof RX antenna

0°

0°

TX azimuth angle

RX

azi

mut

h an

gle

Transmitterantennaazimuth

40

25 20 15 10 5 0

30

2010

0

−60

−60 −30

−90

−120

−150

−180

Link distribution10° TX angle bins

0

Lin

k di

stri

butio

n10° R

X a

ngle

bin

s

Lin

kco

unt

Linkcount

60

Figure 3.18 The antenna pointing angles found with a 37.625 GHz carrier and highly directionalantennas at the receiver and transmitter. The transmitter was elevated to 18 m [from [RBDMQ12][RGBD+13] c© IEEE].




120 38 GHz

38 GHz linear fit60 GHz linear fit

0.09 θcomb + 1.1

0.13 θcomb + 11

60 GHz

100

80

60

RM

S de

lay

spre

ad (

ns)

40

20

0

Angle combination = |TX Azimuth| + |RX Azimuth| (deg)

0 50 100 150 200

Figure 3.19 Steeper azimuth pointing angles are associated with higher RMS delay spreads for out-door peer-to-peer channels. The measurements from this plot were taken with 25 dBi 7 beamwidthhorns at the transmitter and receiver, and with link distances from 19 to 129 m [from [RBDMQ12]c© IEEE].




0 50 100 150 200

200

25dBi RX Ant.

RX Ant.

Linear fit for 25dBi

Linear fit for

0.22θcomb

0.14θcomb

θcomb = |TX azimuth| + |RX azimuth| + |∆Elev.| (deg)

−0.56

+ 3.7

13.3dBi

13.3dBi

RM

S de

lay

spre

ad (

ns)

150

100

50

0

Figure 3.20 Steeper antenna pointing angles are associated with higher RMS delay spreads. Thesemeasurements were taken at 38 GHz with at 25 dBi TX antennas, and a 25 dBi or 13.3 dBi RXantennas. Link distances ranged just beyond 900 m [from [RQT+12] c© IEEE].




30

10

20

30

40

Path

loss

abo

ve 3

m r

efer

ence

(dB

)

50

60

LOS

n = 5

n = 4

n = 3

n = 2snLOS = 2dB

snNLOS = 10.12dB

snNLOS-best = 10.16dB

NLOS

nLOS = 2.25

nNLOS-all = 4.22

nNLOS-best = 3.76

70

Transmitter to receiver separation (m)101 102

Figure 3.21 Due to the very high value for the break-point distance, LOS links at mmWavefrequencies are very close to free space in terms of path loss. This plot was generated with highlydirectional antennas at the receiver and transmitter with 25 dBi gain and 7 beamwidths at 60 GHz[from [RBDMQ12] c© IEEE]. Note that the oxygen absorption causes the path loss exponent to beslightly greater than 2.0.




70 LOS

NLOS

n = 5

n = 4

n = 3

n = 2

nLOS = 2.0

σnLOS = 3.79dBσnLOS = 11.72dBσnNLOS-best = 8.57dB

nNLOS-all = 4.57

nNLOS-best = 3.71

60

50

40

30

20

10

0

Transmitter to receiver separation (m)

Path

loss

abo

ve 3

m r

efer

ence

(dB

)

3 101 102

Figure 3.22 This plot shows measured path loss values for 38 GHz peer-to-peer applications withhighly directional 25 dBi 7 beamwidth horn antennas [from [RBDMQ12] c© IEEE].




σnLOS = 4.55dB

σnLOS-all = 11.55dB

σnLOS = 13.39dB

σnNLOS-best = 11.69dB

LOS

LOS partiallyobstructedNLOS

Best NLOS

nLOS-all = 2.30

nLOS = 1.89

nNLOS-all = 3.86n=5

n=4

n=3

n=2

nNLOS-best = 3.2

70

80

60

50

40

30

20

10

0

Path

loss

abo

ve 3

m r

efer

ence

(dB

)


5 101 102 103

Figure 3.23 When a highly directional antenna is used at the receiver, LOS links will be very closeto free space but NLOS links may be more heavily attenuated. This plot is for 38 GHz and themeasurements used the same highly directional antennas at both the transmitter and receiver [from[RQT+12] c© IEEE].




70

60

50

40

30

20

10

n = 5

n = 4

n = 3

n = 2


Path

loss

abo

ve 5

m r

efer

ence

(dB

)

05 101 102 103

Clear LOS

σnLOS = 3.45dB

σnLOS-all = 9.4dB

σnLOS = 11.03dB

σnLOS-best = 8.37dB

NLOS

LOS partiallyobstructed

Best NLOS

nLOS =1.90

nLOS-all =2.21

nNLOS-best =2.56

nNLOS-all =3.18

Figure 3.24 This plot was generated from measurements using a 25 dBi 7 beamwidth horn TXantenna and a less directional 13.3 dBi 40 beamwidth horn at the receiver. NLOS paths are signif-icantly stronger as the receiver cannot filter out multipath as effectively as when a more directionalantenna is used [RQT+12] c© IEEE].




1 30 100 20040

60

80

100

120

140

160

T−R separation (m)

Path

loss

(dB

)

28 GHz omnidirectional PL model 1 m − Manhattanfor (RX at 1.5 m AGL)

NLOSLOSn

NLOS = 3.4, σ

NLOS = 9.7 dB

nLOS

= 2.1, σLOS

= 3.6 dB

(α, β, σ) = (79.2 dB, 2.6, 9.6 dB)n

FreeSpace

UHF (1900 MHz) nNLOS

= 2.6,

σNLOS

= 7.7 dB

Figure 3.25 28 GHz omnidirectional close-in free space reference distance (d0 = 1 m) and floatingintercept path loss models for a non-line of sight (NLOS) urban environment with a receiver antenna1.5 m above ground. A comparison is made to path loss in a 1.9 GHz urban NLOS environment asreported in [BFR+92] [FBRSX94].




160

70

80

90

100

110

120

130

140

150Omni NLOS Path Loss

28 GHz omni NLOS PL model 1 m - Manhattan

nNLOS = 3.4, σNLOS = 9.7 dB

(α, β, σ) = (79.2

10

T-R separation (m)

Path

loss

(dB

)

100

Figure 3.26 28 GHz omnidirectional path loss model from which the TX and RX antenna gains havebeen removed. The close-in free space reference distance model with respect to a 1 m free spacereference distance, and the floating intercept (α, β) model from [RRE14] is shown for distancesranging from 30 to 200 m.




dB, 2.6, 9.6)

10

10

20

30

40

50

60

70

80Omni LOS Path Loss

28 GHz omni LOS PL model 1 m - Manhattan

nLOS = 2.1, σLOS = 3.6 dB

10

T-R separation (m)

Path

loss

abo

ve 1

m F

S re

fere

nce

(dB

)

100

n=3

n=2

Figure 3.27 28 GHz omnidirectional path loss model from which the TX and RX antenna gainshave been removed. The close-in free space reference distance model with respect to a 1 m freespace reference distance is shown. Note that one point at 100 m had excessive path loss due to thefact that the antennas were not aligned on boresight at this location. By removing this single point,it is evident that the LOS path loss exponent is very close to 2.




1 10 20 30 100 2000

20

40

60

80

100

120


Path

loss

abo

ve 1

m F

S re

fere

nce

(dB

)

28 GHz Manhattan path loss versus distance

n = 2

n = 3

n = 4

n = 5NLOS Path Loss

Path LossLOS−Co Path LossLOS−Cross Path Lossn

NLOS = 4.4, σ

NLOS = 10.0 dB

nNLOS−Best

= 3.7, σNLOS−Best

= 9.2 dB

nLOS−Co

= 1.8, σLOS−Co

= 0.99 dB

nLOS−Cross

= 3.6, σLOS−Cross

= 5.1 dB

Figure 3.28 28 GHz Manhattan single beam path loss measurements as a function of T-R sep-aration distance using 24.5 dBi horn antennas with 10.9 half-power beam width at both the TXand RX and 15 dBi (28.8 degree HPBW) horn antennas at both the TX and RX. NLOS path lossesinclude LOS non-boresight and truly NLOS measurements. Co-polarized and cross-polarized LOSmeasured path losses are also shown. The close-in free space reference distance model with respectto a 1 m free space reference distance is shown. All data points represent path loss values calculatedfrom recorded PDP measurements.




1 30 100 200

80

100

120

140

160


Path

loss

(dB

)

73 GHz omnidirectional PL model 1 m − Manhattanfor hybrid (RX at 2 m and 4.06 m AGL)

NLOSLOSn

NLOS= 3.4, σ

NLOS= 7.9 dB

nLOS

= 2.0, σLOS

= 4.8 dB

(α, β, σ) = (80.6 dB, 2.9, 7.8 dB)n

FreeSpace

Figure 3.29 73 GHz omnidirectional path loss model from which the TX and RX antenna gainshave been removed for a combination of cellular and backhaul (hybrid) RX antenna heights. Theclose-in free-space reference distance model for d0 = 1 m and the floating intercept (α, β) modelfrom [RRE14] over 30-200 m are shown.




1 10 20 30 100 20040

60

80

100

120

140

160


Path

loss

(dB

)

73 GHz omnidirectional PL model 1 m − Manhattanfor access (RX at 2 m AGL)

NLOSLOSnNLOS = 3.3, σNLOS = 7.6 dB

nLOS

= 2.0, σLOS

= 5.2 dB

(α, β, σ) = (81.9 dB, 2.7, 7.5 dB)n

FreeSpace(73 GHz)

UHF (1900 MHz) nNLOS

=

σNLOS

= 7.7 dB

Figure 3.30 73 GHz omnidirectional path loss model from which the TX and RX antenna gainshave been removed for mobile RX antenna heights of 2 m. The close-in free-space reference distancemodel for d0 = 1 m and the floating intercept model (α, β) model from [RRE14] over 30-200 mare shown. A comparison is made to path loss in a 1.9 GHz urban NLOS environment as reported in[BFR+92].




1 30 100 200

80

100

120

140

160


Path

loss

(dB

)

73 GHz omnidirectional PL model 1 m − Manhattanfor backhaul (RX at 4.06 m AGL)

NLOSLOSn

NLOS= 3.5, σ

NLOS= 7.9 dB

nLOS

= 2.0, σLOS

= 4.2 dB

(α, β, σ) = (84.0 dB, 2.8, 7.8 dB)n

FreeSpace

Figure 3.31 73 GHz omnidirectional path loss model from which the TX and RX antenna gainshave been removed for backhaul RX antenna heights of 4.06 m. The close-in free-space referencedistance model for d0 = 1 m and the floating intercept (α, β) model from [RRE14] over 30-200 mare shown.




1 10 20 30 100 20070

90

110

130

150

170

190

n = 2

n = 3

n = 4

n = 5


Path

loss

(dB

)73 GHz unique pointing angle path loss versus distance with

RX height: 2 m using 27 dBi, 7° 3dB BW TX antennas and

27 dBi, 7° 3dB BW RX antennas in Manhattan

NLOSNLOS−bestLOSnNLOS = 4.7, σ

NLOS 12.6 dB

nNLOS−best = 3.6, σNLOS−best = 10.6 dB

nLOS = 2.2, σLOS 5.2 dB

NLOS Omni: α = 81.9 dB, β = 2.7,σ = 7.5 dB

Figure 3.32 New York City cellular RX height (2 m) path loss measurements at 73 GHz as a func-tion of T-R separation distance using vertically polarized 27 dBi, 7 half-power beam width TX andRX antennas. All data points represent path loss values calculated from recorded PDP measurements.Crosses indicate all NLOS pointing angle data points, diamonds indicate best NLOS pointing angledata points for each RX location and each T-R combination, and circles indicate LOS data points.The measured path loss values are relative to a 1 m free-space close-in reference distance. NLOS PLEsare calculated for the entire data set and also for the best recorded link. LOS PLEs are calculated forstrictly boresight-to-boresight scenarios. n values are PLEs and σ values are shadow factors. The solidline spanning 30 to 200 m is the omnidirectional (α, β) model from [RRE14] [ALS+14] depicted inFig. 3.30.




1 10 20 30 100 20070

90

110

130

150

170

190

n = 2

n = 3

n = 4

n = 5


Path

loss

(dB

)73 GHz unique pointing angle path loss vs. distance with

RX height: 4.06 m using 27 dBi, 7° 3dB BW TX antennas and

27 dBi, 7° 3dB BW RX antennas in Manhattan

NLOSNLOS−bestLOSn

NLOS = 4.7, σ

NLOS12.7 dB

nNLOS−best

= 3.7, σNLOS−best

= 11.2 dB

nLOS

= 2.4, σLOS

6.3 dB

NLOS Omni: α = 84.0 dB, β = 2.8,σ = 7.8 dB

Figure 3.33 New York City backhaul measurements with RX heights of 4.06 m path losses at73 GHz as a function of T-R separation distance using vertically polarized 27 dBi, 7 half-power beamwidth TX and RX antennas. All data points represent path loss values calculated from recorded PDPmeasurements. Crosses indicate all NLOS pointing angle data points, diamonds indicate best NLOSpointing angle data points for each RX location and each T-R combination, and circles indicateLOS data points. The measured path loss values are relative to a 1 m free-space close-in referencedistance. NLOS PLEs are calculated for the entire data set and also for the best recorded link. LOSPLEs are calculated for strictly boresight-to-boresight scenarios. n values are PLEs and σ valuesare shadow factors. The solid line spanning 30 to 200 m is the omnidirectional (α, β) model from[RRE14] depicted in Fig. 3.31.




0−75

−70

−65

−60

−55

Void 1:Duration: 2.7 ns




20 40 60

Time (ns)

Rec

eive

d po

wer

(dB

m/n

s)

80 100

Cluster 5:Duration: 1.9 ns1 Sub-path

Cluster 3:Duration: 2.00 ns2 Sub-paths




Figure 3.34 Illustration of some of the key temporal modeling parameters used for modeling thetemporal clusters in an omnidirectional SSCM wideband mmWave channel. This example shows fivetime clusters, with time durations ranging from 2 to 31 ns, and voids between clusters ranging from2.7 to 23.9 ns [SR14a].




Lobe segment

Lobe Azimuth Spread

RMS LAS

60°

-84 -74 -64 -54

30°

90°dBm

120°

150°

180°

210°

240°

270°

300°

330°

0°

AOA

Lobe power

Lobe

Figure 3.35 Illustration of some of the key spatial modeling parameters used to model thespatial lobes in an omnidirectional SSCM wideband mmWave channel. The polar plot (in theazimuthal/horizontal plane only) shows five distinct lobes with various lobe azimuth spreads andAOAs. Each dot is a lobe angular segment simulated for a particular discrete pointing angle andrepresents the total integrated received power over a particular beam width (and corresponds to thearea under a PDP for the particular RX pointing angle). The lobe power is the sum of powers fromeach segment within the lobe (e.g., the sum of powers from each lobe segment in a lobe).




60

Path

loss

abo

ve 3

m r

efer

ence

(dB

)50

40

30

20

10

0101


npeer-NLOS = 4.19

nveh-LOS = 2.66

npeer-LOS = 2.23

Peer-to-peer NLOS

Peer-to-peer LOS

Vechicle LOS

n=5

n=4

n=3

n=2

Vechicle NLOS

nveh-NLOS = 7.17

102

Figure 3.36 Path loss for 60 GHz for peer-to-peer applications and communication from a ground-based transmitter to a receiver in a vehicle. These measurements used highly directional 25 dBi 7

beamwidth antennas as the transmitter and receiver [from [BDRQL11] c© IEEE].




All NLOS meas.

E[σNLOS] = 6.48 ns

Max[σNLOS] = 36.6 ns

E[σLOS] = 0.76 ns

E[σpeer] = 6.02 ns

E[σvehicle] = 2.73 ns

Max[σLOS] = 0.88 ns

Vehicleall measurements


Prob

abili

ty R

MS

dela

y sp

read

< A

bsci

ssa

00

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

5 10 15 20 25 30 35 40

Peer-to-peerall measurements

All LOS meas.

Figure 3.37 These measurements used highly directional 25 dBi 7 beamwidth antennas at thetransmitter and receiver. When the transmitter communicates to a receiver inside a vehicle, muchlower RMS delay spreads result than when the transmitter and receiver are in the open [from[BDRQL11] c© IEEE].




Excess time

Tcurser−t

AOA’s of rays areclustered

Clustered in space

Pre-curserpower-growth

Numberof post-curser

rays Nb

Cluster arrivalrate Λ

Post-curserpower-decay

Post-curser MPCarrival rate λbPre-Curser MPC

arrival rate λ f

NLOS clusterCluster decayrate Γ Γe

Channel impulseresponse

LOS componentgain of β

φ

Numberof pre-curser

rays Nf

γf

t−Tcurser

γf

Receiveing antenna

Clustered in time−t

e

e

Figure 3.38 Representation of key parameters used to specify multipath channels. Statistics ofthe key channel parameters are generated from measured data, as collected by wideband channelsounders, to determine the temporal and spatial channel models that can be used by researchers andstandard bodies for modem design and signaling protocols.



4.3 The On-Chip Antenna Environment 203

Metal guard ringSlotted metal

Figure 4.6 There are several considerations for on-chip antennas related to CMOS productionrules: 1) All metal layers must meet a minimum fill requirement. This is reflected in the figure by thefact that there are no large portions of the chip left empty (the lighter-shaded portions of the figure).2) A metal guard ring must often surround the chip to prevent damage during dicing. 3) Largeareas of metal must be slotted to meet design rules. 4) Metal structures must meet a minimum sizerequirement, which in practice is usually satisfied by most designs.



4.3 The On-Chip Antenna Environment 207

100 −10

−20

−30

−40

−50

80

60

40

20

00 50 100 150

Si substrate thickness [µm]

Rad

iatio

n ef

fici

ency

(e r

ad)

(%)

Tra

nsm

issi

on c

oeff

icie

nt (

S 21)

[dB

]

200 250

L= 6 mmp= 10Ω cmd= 5 mm

300

Figure 4.10 This figure indicates that the efficiency of on-chip antennas is reduced greatly by athick substrate [from [KSK+09] c© IEEE].



4.4 In-Package Antennas 209

tanδd=0.001100

10

1

0.1

0.011 10 100

Resistivity (Ω−cm)

Atte

nuat

ion

(dB

/mm

)

1k 10k 100k

tanδd=0.003

tanδd=0.01

tanδd=0.03

Figure 4.12 For low resistivity substrates, the loss due to currents carried by substrate dopants(i.e., conductive losses) is the major loss mechanism hurting performance [from [LKCY10] c© IEEE].



216 Chapter 4 Antennas and Arrays for MmWave Applications

120

50

40

30

20

40 50 60 70 80Width, W (µm)

Thickness (µm)

Ret

urn

loss

(+d

B)

Ret

urn

loss

(+d

B)

90 10040 50 60 70 80 90 100

110

10

0

50

45

35

25

55

40

30

10

8

6

4

2

0

100

L = 4 mm

L = 3 mm

L = 4 mm

L = 5 mm

L = 6 mm

L = 3 mmW = 100 µm

W = 90µm

W = 80 µm

W = 70 µm

L = 6 mm

Cr/Au: 10 nm/60 nm

Cr/Au: 10 nm/60 nm

Cr/Au: 10 nm/60 nm

L = 5 mm

L = 3 mm

L = 2 mm

80

60

40

20

00 5 10 15

Frequency, f (GHz)

Wid

th, W

(µm

)

Len

gth,

L (

mm

)

20 25 30 5 10 15 20Frequency, f (GHz)

25 30 35 4035

Figure 4.18 [MHP+09] presented these plots for the design of a planar dipole antenna on a625 µm GaAs substrate with relative permittivity of 12.9 and 625 µm thick [reproduced from[MHP+09] c© IEEE].




0X

Y

30−30

Silicon substrate

Patch (Pad layer)

Ground (M1−M5)Slot (2 µm wide)

−7.00

−14.00

−21.00

−28.00

60−60

dB(Gain total)

Phi

XZ plane (Phi = 0°)

dB(Gain total)YZ plane (Phi = 90°)

90

120

150−180

−150

−120

−90

YX

Z

Figure 4.26 This figure shows that a patch antenna typically radiates above the top metal layerof the antenna. The top figure illustrates the metal slots that are usually required for on-chip patchantennas due to their large size. In the lower figure, [CGLS09] used two parallel metal strips on theedge of the patch to increase bandwidth. [This figure is a combination of figures from the literature([SCS+08] above, [CGLS09] below) c© IEEE.]



4.5 Antenna Topologies for MmWave Communications 223

Feeding via

Patch ground planePackage substrate

Patch

Chip dielectric

Chip substrate

H1

H2

εr2

εr1Feeding substrate

Packaging

Patch

W

L

A

Air

Ball connector

CPWFA

Hcav

Scav

Figure 4.27 There are various methods for feeding an in-package patch from a packaged chip.The ball connector (left) may, for example, be used in a flip chip connection. [KLN+11] found thatthis type of ball connector improves with a smaller radius of the ball and a smaller metal pad for theball (represented here as a small rectangular piece of metal below the ball) [right portion from[HRL10] c© IEEE].




Lc

Wc

Figure 4.34 This element was cascaded periodically below an on-chip microstrip antenna to forman AMC to achieve a gain of −1.5 dBi [from [CGLS09] c© IEEE].



4.8 Characterization of On-Chip Antenna Performance 255

Manyscatterers

Measurementarea

Proberadiation

Metalstage

Figure 4.58 A probe station is often used to characterize mmWave antennas. These measurementsmay be inaccurate due to radiation from probes and scattered fields from the many surrounding metalobjects [from [MBDGR11] c© IEEE].




180°

Direction of movement

Probe Probe

90°

0°

C1

C2

Figure 4.59 The two antennas were swept in angle across each other. The chips on which theantennas were fabricated are represented by squares, and the antennas are represented by smallerblack boxes [from [MBDGR11] c© IEEE].




With de-embedding

No de-embedding

Yagi gain at chip horizon (65 GHz)

Angle (degrees)

Ant

enna

gai

n (d

Bi)

Figure 4.60 The de-embedding method indicates that the on-chip Yagi pattern was distorted bythe presence of other nearby metal structures [from [MBDGR11] c© IEEE].



4.9 Chapter Summary 257

–15–100 –75 –50 –25 0

Curve infoYagi onlyYagi w/ metal barYagi w/ metal in Q2Yagi w/ metal in all QMeasured

Phi [deg]

Q2

xx

zy

y

yx

60 GHz Yagi antenna pattern normalized

Ant

enna

gai

n-ph

i dir

ectio

n (d

Bi)

25 50 75 100

–13

–10

–8

–5

–3

0

Figure 4.61 Simulations confirmed measurements that indicated the distortion of the antennapattern was caused by surrounding metal structures. This indicates that isolation between integratedantennas and other nearby structures on the chip or in the package is key to successful design [from[MBDGR11] c© IEEE].



5.4 Simulation, Layout, and CMOS Production of MmWave Circuits 271

Probe

PadT-line

Top view

T-line ABCD matrixPad Pad

Z0, β

60 µm 60 µm340 µ

m550 µm

Side viewCpad

Figure 5.7 The S-parameters of a transmission line can be used to determine the effective relativepermittivity and loss tangent of a CMOS process. The effects of the probe pads must be de-embeddedfor this measurement to be accurate [from [GJRM10] c© IEEE].



272 Chapter 5 MmWave RF and Analog Devices and Circuits

Extracted IC dielectric permittivity from measured probepad and transmission line

T line epsilon (real part)

Probe pad epsilon (enhanced)

Probe pad epsilon (simple)

5.5R

elat

ive

perm

ittiv

ity ε

r 5

4.5

4

3.5

10 20 30 40 50Frequency (GHz)

60 70 803

Figure 5.8 The effective relative permittivity may be measured in a number of ways and is a vitalparameter for the design of passive structures [from [GJRM10] c© IEEE].



5.5 Transistors and Transistor Models 273

Loss tangent of IC dielectric from extracted transmissionline complex permittivity

Los

s ta

ngen

t (ta

nδ)

0.2

0.15

0.1

0.05

00 10 20 30 40

Frequency (GHz)50 60 70

Figure 5.9 The effective loss tangent is a vital parameter to predict loss of passive structures [from[GJRM10] c© IEEE].




Metal removed tomeet CMP metalfill requirements

CPW signal traces

Figure 5.26 A common ground plane is evident in this layout. Portions of the metal have beenremoved in order to meet metal fill requirements [based on a figure from [MTH+08] c© IEEE].



5.9 Sensitivity and Link Budget Analysis for MmWave Radios 315

1E+11

Frequency (Hz)

00

–2

–4

–6

–8

–10

–12

5E+10

S11 dB

Figure 5.42 The amplifier accepts energy in only a certain range of frequencies.



5.9 Sensitivity and Link Budget Analysis for MmWave Radios 315

Frequency (Hz)

2.00E + 01

1.50E + 01

1.00E + 01

5.00E + 00

0.00E + 00

– 5.00E + 00

– 1.00E + 01

5E+10 1E+11

S21 dB

0

Figure 5.43 The value of S21 gain is only high in a certain band of frequencies.




Baseband

Mixer

Antennas

Mixer

Baseband

VCOVCO

Poweramp. (PA)

Low noiseamp (LNA)

Figure 5.44 A direct conversion architecture for a transmitter and receiver. This is a populararchitecture for today’s cellphones. In many designs, the VCO is part of a phase-locked loop (PLL).




Ideal

Real(compression)

Input power, mW

Output power, mW20

18

16

14

12

10

8

6

4

2

00 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

Real vs. ideal output power

Psat

OP1dB

P1dB

1dB

Figure 5.46 The non-linearity of most devices results in the compression of the output power ofthe fundamental harmonic.



5.11 Analog MmWave Components 331

3

2

1

0

0

0.5

1

1.5

0

0 1 2 3 4 5 6 7

1 2 3

Amplifier classes

Normalized time

Normalized time

Drain conduction current

IAIBIABIC

Threshold voltageIn

put v

olta

ge

Nor

mal

ize

drai

nco

nduc

tion

curr

ent

4 5 6 7

RL

Vdd

Vin = Vin,DC + Vin,RF

VinAVinBVinABVinC

Figure 5.50 The bias point of the amplifier determines the amplifier’s class. Class A amplifiersconduct current over the entire period. Class B amplifiers conduct over half the period, Class Cconduct over less than half the period, and Class AB conduct over more than half the period, butless than the entire period.




LO v

olta

ge

Normalized time

Normalized time

Bet

a fa

ctor

Beta

Beta

1 2

6

5

4

3

2

1

0

–1

–2

–0.2

0

0.2

0.4

0.6

0.8

1

1.2

3 4 5 60

1 2 3 4 5 60

VddRD

VoutRoCAS

Ro

RS

gmCAS

Vin

p

T

LO voltage

Cascode threshold voltage

Figure 5.60 Based on the nature of the LO signal magnitude and bias point, we may treat thetransconductance of the cascode as a square wave that switches on and off. This figure shows howthe gain of the switching mixer is gated by the LO voltage to create the switching effect of the mixer.This approach is used in double balanced mixers, such as Gilbert cells.




Open-loop gain

00

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

20 40 60 80 100 120 140 160 180

GHz

Phase

–60

–40

–20

00

20

40

60

80

100

20 40 60 100 120 140 160 180

GHz

80

Figure 5.63 The open-loop gain (top) and phase shift (bottom) of the oscillator in Fig. 5.62.




Open-loop gain magnitude

00

0.2

0.4

0.6

0.8

1

1.2

20 40 60 80 100 120 140 160 180

GHz

Open-loop phase

−1.50E+02

−1.00E+02

−5.00E+01

0

5.00E+01

1.00E+02

1.50E+02

0 20 80 100 120 140 160 180

GHz

6040

Figure 5.65 The gain and phase of a simple LC tank oscillator.




Open-loop gain magnitude

00

0.2

0.4

0.6

0.8

1

1.2

20 40 60 80 100 120 140 160 180

GHzOpen-loop phase

−1.50E+02

−1.00E+02

−5.00E+01

0.00E+00

5.00E+01

1.00E+02

1.50E+02

0 20 80 100 120 140 160 180

GHz

6040

Closed-loop gain

00

5

10

15

20

25

30

20 40 60 80 100 120 140 160 180GHz

Figure 5.71 For real-world oscillator circuits, the output spectrum will be polluted by power atfrequencies other than the intended operating frequency.




Noisy waveform

Time

Noiseless waveform

Noise impulse

Figure 5.72 Noise events, such as the noise impulse represented here, will affect both the amplitudeand phase of a circuit. In general, the amplitude impulse response of the circuit will act to removeamplitude noise over time. But phase noise persists, as is evident when we compare the phase of thenoisy waveform to the noiseless waveform.




Output spectrum Q comparison

00

5

10

15

20

25

30

20 40 60 80 100 120 140 160 180

GHz

L = 1 nH

L = 0.1 nH

Figure 5.73 The output spectrum of an LC oscillator becomes more spectrally pure as the qualityfactor of the tank increases.



1.2

0.8

0.6

0.4

0.2

00.00E+00 5.00E+10 1.00E+11

Hertz

Gai

nD

egre

esG

ain

Hertz

Hertz

Open-loop gain

1.50E+11 2.00E+11

1

150

100

50

−50

−100

00.00E+00 5.00E+10 1.00E+11 1.50E+11 2.00E+11

Open-loop phase

250

200

150

100

50

00.00E+00 5.00E+10 1.00E+11 1.50E+11 2.00E+11

Closed-loop gain

Figure 5.76 The transfer characteristics of a subharmonic oscillator.


1.2

0.8

0.6

0.4

0.2

00.00E+00 5.00E+10 1.00E+11

Hertz

Gai

nD

egre

esG

ain

Hertz

Hertz

Open-loop gain

1.50E+11 2.00E+11

1

150

100

50

−50

−100

00.00E+00 5.00E+10 1.00E+11 1.50E+11 2.00E+11

Open-loop phase

250

200

150

100

50

00.00E+00 5.00E+10 1.00E+11 1.50E+11 2.00E+11

Closed-loop gain

Figure 5.76 The transfer characteristics of a subharmonic oscillator.



5.12 Consumption Factor Theory 379

Energy expenditure per bit vs. transmission distance

Distance (m) [Log Scale]

PNP =1 W, max distance=67 m




5 15 50 160 500

RX Gain 30 dbCarrier frequency 20 GHzPath loss exponent: 4

Capacity 1 GbpsHTX = 1HRX = 1

1580 5000

Ene

rgy

per

bit (

dBJ)

−30

−40

−50

−60

−70

−80

−90

−100

Figure 5.86 For a system with high signal path efficiency and high non-path power consumption,we see that the energy expenditure per bit is dominated by non-path power, indicating little advantageto shortening transmission distances [from [MR14b] c© IEEE].




−1005 15 50 160 500 50001580

Ene

rgy

expe

nditu

re p

er b

it (d

BJ)

Distance (m) [log scale]

Energy per bit vs. transmission distance

−90

−80

−70

−60

−50

−40

−30





RX gain 30 dBCarrier frequency 20 GHzPath loss exponent: 4

Capacity 1 GbpsHTX = 0.01HRX = 0.01

Figure 5.87 When signal path components are less efficient, as illustrated here, then shorter trans-mission distances start to become advantageous, as signal path power starts to represent a largerportion of the power expenditure per bit [from [MR14b] c© IEEE].




−1005 15 50 160 500 50001580

Ene

rgy

expe

nditu

re p

er b

it (d

BJ)



−90

−80

−70

−60

−50

−40

−30

PNP =100 W, max distance=40 mPNP =1000 W, max distance=71 m


RX Gain 30 dBCarrier frequency 180Path loss exponent: 4

HTX =1Capacity 10 GbpsHRX =1

Figure 5.88 A higher frequency system that can provide a much higher bit rate capacity withoutsubstantially increasing non-path power consumption may result in a net reduction in the energyprice per bit [from [MR14b] c© IEEE].



5.12 Consumption Factor Theory 381

−1005 15 50 160 500 50001580

Ene

rgy

expe

nditu

re p

er b

it (d

BJ)



−90

−80

−70

−60

−50

−40

−30



HTX = 0.01HRX = 0.01

Capacity 10 GbpsRX gain 30 dBCarrier frequency 180 GHzPath loss exponent: 4

Figure 5.89 Lower efficiencies of signal path components motivate the use of shorter transmissiondistances [from [MR14b] c© IEEE].



6.2 Review of Sampling and Conversion for ADCs and DACs 385

Arbitrary signal1

0.8

0.6

0.4

0.2

-0.2

-0.4

-0.6

-0.8

-10 0.1 0.2 0.3 0.4 0.5

Time (seconds)0.6 0.7 0.8 0.9 1

´10-5

0

Vol

tage

(V

)

Figure 6.1 An arbitrary baseband analog signal having an approximate bandwidth of 100 MHz.



386 Chapter 6 Multi-Gbps Digital Baseband Circuits

Spectrum of arbitrary signal

-0.2-0.3-0.4

-90

-100

-80

-70

-60

-50

-40

-30

-20

-10

0

-0.1 0Frequency (GHz)

Am

plitu

de (

dB)

0.1 0.2 0.3 0.4

Figure 6.2 The spectrum of the arbitrary analog waveform shown in Fig. 6.1. The spectrum hasbeen normalized such that the strongest spectral component has an amplitude of 0 dB.




100 MHz BW signal sampled at 400 MHz

-0.2

-0.3

-0.1

0

4 4.02 4.04 4.06 4.08 4.1 4.12 4.14 4.16 4.18 4.2

Sampled signalOriginal signal

0.1

0.2

0.3

Time (seconds)

Vol

tage

(V

)

´ 10-6

Figure 6.3 A zoomed-in version of the arbitrary signal in Fig.6.1 showing how it has been sampled.The bandwidth (BW) of the original signal is 100 MHz, and the sampling rate is 400 MHz.




Spectrum of 100 MHz bandwidth signal sampled at 400 MHz

-0.2-0.3-0.4-0.5

-90

-80

-70

-60

-50

-40

-30

-20

-10

-0.1 0Frequency (GHz)

Am

plitu

de (

dB)

0.1 0.2 0.3 0.4 0.5

Figure 6.4 The result of sampling the signal in Fig. 6.1 in the time domain is to make the spectrumperiodic in the frequency domain.




Aliased signal0

−10

−20

−30

−40

−50

−60−3 −2 −1 0

Frequency (GHz)

Am

plitu

de (

dB)

1 2 3

Figure 6.5 If the signal of Fig. 6.1 with a baseband bandwidth of 100 MHz is sampled at 100 MHz, the result is an aliased signal. In the frequency domain, overlapping copies of the original signal’s spectrum completely distort the resulting sampled signal.




100 MHz BW signal, sampled at 400 MHz and discretized to 4 bits

-0.6

-20

-15

-10

-5

0

-0.4 -0.2 0Frequency (GHz)

Am

plitu

de (

dB)

0.2 0.4 0.6

Figure 6.6 Quantizing the signal of Fig. 6.1 with 4 bits reduces the dynamic range to 24 dB.




Sampling with jitter

4.164.144.124.1

-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

4.18 4.2 4.22 4.24 4.26

Time (seconds) ´ 10-6

Vol

tage

(V

)

Figure 6.7 An example of sampling with jitter, where uncertainty in the time interval betweensamples results in decreased dynamic range.



438 Chapter 7 MmWave Physical Layer Design and Algorithms

Positive signal limit

Voltage signalQuantization levelSample points

TIME

Negative signal limit

Vol

tage

Figure 7.2 Amplitude thresholding and quantization for a 3-bit ADC with uniform quantizationlevels in the receiver. In a wireless communications receiver, ADCs quantize both the in-phase andquadrature channels independently. Automatic gain control (AGC) is used to normalize the energyof the received complex baseband signal so that the ADC thresholds and quantization levels can befixed (i.e., do not depend on fading). As shown in Chapter 3, fading will be less pronounced withdirectional antennas.




0.5

0.4

0.3

0.2

0.1

0

14

12

10

8

6

4

2

0

Out

put (

volts

)

Out

put p

hase

shi

ft (

degr

ees)

Input (volts)0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4

MeasurementFitted model

Figure 7.3 AM-PM and AM-PM measurements and modified Rapp model for a CMOS 65 nmPA. For more detail on mmWave PA statistics, please consult the tables provided in Chapter 5 [plotcreated with data from [EMT+09] c© IEEE].




–40 dBc/Hz

–50 dBc/Hz

–60 dBc/Hz

–70 dBc/Hz

–80 dBc/Hz

–90 dBc/Hz

–100 dBc/Hz

–110 dBc/Hz

Ideal component

IBM VCO — smoothed

Pole/zero model

2 MHz per division

67.278 GHz 67.296 GHz 67.314 GHz

Figure 7.4 Measured power spectral density of single sideband (SSB) VCO output for desired signalat 67.3 GHz. PSD measurements normalized to desired carrier output power (represented by dBc/Hz).A similar figure is shown in Chapter 5. Here, however, we have included a comparison to a pole/zeromodel that facilitates physical layer performance simulations [data taken from [FRP+05] and smoothedto create the plot c© IEEE].



7.2 Practical Transceivers 443

Pow

er s

pect

ral d

ensi

ty (

dB/H

z)

−30 dB/decade = flicker FM

−40 dB/decade = random walk

−20 dB/decade = white FM

−10 dB/decade = flicker phase

White noise

Frequency (Hz)

Figure 7.5 Different phase noise effects and their contribution to the power spectral density in theLeeson model [Lee66].



7.3 High-Throughput PHYs 449

SC-FDE, RS(255,239) — CM1.3

SC-FDE/OFDM — AWGNSC-FDE/OFDM — AWGN + mask

SNR (dB)

QPS

K B

ER

SC-FDE, LDPC(672,336) — CM1.3

SC-FDE, LDPC(672,432) — CM1.3OFDM, RS(255,239) —

OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336)

100

10-1

10-2

10-3

10-4

10-5

-2 0 2 4 6 8 10 12

Figure 7.6 Bit error rate as a function of SNR for OFDM and SC-FDE in AWGN and LOSchannels with QPSK constellations. The coding options are RS(255,239) and LDPC(672,336). The802.15.3c spectral mask is added to the AWGN channel to demonstrate bandwidth conservation ofeach modulation strategy. Hardware impairments are not considered.




SNR (dB)

SC-FDE, RS(255,239) — CM1.3

SC-FDE/OFDM — AWGNSC-FDE/OFDM — AWGN + mask

SC-FDE, LDPC(672,336) — CM1.3

SC-FDE, LDPC(672,432) — CM1.3OFDM, RS(255,239) — CM1.3

OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336) — CM1.3

4 6 8 10 12 14 16 18 20

16–Q

AM

BE

R

100

10-1

10-2

10-3

10-4

10-5

Figure 7.7 Bit error rate as a function of SNR for OFDM and SC-FDE in LOS CM1.3 channelswith 16-QAM constellations. Hardware impairments are not considered.




SC-FDE, RS(255,239) — CM2.3

SC-FDE, LDPC(672,336) — CM2.3

SC-FDE, LDPC(672,432) — CM2.3OFDM,

OFDM, LDPC(672,432) — CM2.3

OFDM,

SNR (dB)

QPS

K B

ER

100

10-1

10-2

10-3

10-4

10-5

-2 0 2 4 6 8 10 12 14 16 18 20

Figure 7.8 Bit error rate as a function of SNR for OFDM and SC-FDE in NLOS CM2.3 channelswith QPSK constellations. No hardware impairments are considered.




SNR (dB)

16–Q

AM

BE

RSC-FDE, RS(255,239) — CM2.3

SC-FDE, LDPC(672,336) — CM2.3


OFDM, LDPC(672,432) — CM2.3

OFDM, LDPC(672,336) — CM2.3

5 10 15 20 25

100

10-1

10-2

10-3

10-4

10-5

Figure 7.9 Bit error rate as a function of SNR for OFDM and SC-FDE in NLOS CM2.3 channelswith 16-QAM constellations. No hardware impairments are considered.




SC-FDE, RS(255,239) —

w/ CMOS PA, 1 dB Backoffw/ CMOS PA, 2 dB Backoffw/ CMOS PA, 4 dB Backoff

— CM1.3

SC-FDE, LDPC(672,432)OFDM, RS(255,239) — CM1.3

OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336) — CM1.3

SNR (dB)

Increasing

backoff

-2 0 2 4 6 8 10 12

QPS

K B

ER

100

10-1

10-2

10-3

10-4

10-5

Figure 7.10 Bit error rate as a function of SNR for OFDM and SC-FDE in LOS CM1.3 channelsand CMOS PA non-linearity with QPSK constellations.




SNR (dB)4 6 8 10 12 14 16 18 20 22

SC-FDE, RS(255,239) — CM1.3


SC-FDE, LDPC(672,336) — CM1.3


OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336) —

16–Q

AM

BE

R

100

10-1

10-2

10-3

10-4

10-5

Figure 7.11 Bit error rate as a function of SNR for OFDM and SC-FDE in LOS CM1.3 channelsand CMOS PA nonlinearity with 16-QAM constellations.




SC-FDE,


LDPC(672,336) — CM2.3

SC-FDE,OFDM, RS(255,239) — CM2.3

OFDM, LDPC(672,432) — CM2.3

OFDM, LDPC(672,336) — CM2.3

QPS

K B

ER

0 5 10 15 20SNR (dB)

100

10-1

10-2

10-3

10-4

10-5

Figure 7.12 Bit error rate as a function of SNR for OFDM and SC-FDE in NLOS CM2.3 channelsand CMOS PA non-linearity with QPSK constellations.




SC-FDE, RS(255,239) —

w/ CMOSw/ CMOSw/ CMOS

— CM2.3

SC-FDE, LDPC(672,432)OFDM, RS(255,239) — CM2.3

OFDM, LDPC(672,432) — CM2.3

OFDM, LDPC(672,336) — CM2.3

16–Q

AM

BE

R

SNR (dB)

5 10 15 20 25 30 35

100

10-1

10-2

10-3

10-4

10-5

Figure 7.13 Bit error rate as a function of SNR for OFDM and SC-FDE in NLOS CM2.3 channelsand CMOS PA non-linearity with 16-QAM constellations.




SC-FDE, RS(255,239)

w/ 5-bit ADC, 1.0σ peak mag.w/ 5-bit ADC, 1.5σ peak mag.w/ 5-bit ADC, 2.0σ peak mag.

SC-FDE, LDPC(672,336) — CM1.3


OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336) — CM1.3

SNR (dB)

Increasing

peak mag.

QPS

K B

ER

0 2 4 6 8 10 12

100

10-1

10-2

10-3

10-4

10-5

Figure 7.14 Bit error rate as a function of SNR for OFDM and SC-FDE in LOS CM1.3 channelsand 5-bit ADC samples with QPSK constellations.




SC-FDE, RS(255,239)

w/ 5-bit ADC, 1.0σ peak mag.w/ 5-bit ADC, 1.5σ peak mag.w/ 5-bit ADC, 2.0σ peak mag.

SC-FDE, LDPC(672,336) — CM1.3


OFDM, LDPC(672,432) — CM1.3

OFDM, LDPC(672,336) — CM1.3

SNR (dB)

16–Q

AM

BE

R

6 8 10 12 14 16

100

10-1

10-2

10-3

10-4

10-5

Figure 7.15 Bit error rate as a function of SNR for OFDM and SC-FDE in LOS CM1.3 channelsand 5-bit ADC samples with 16-QAM constellations.




bN bN bN... ...

1st sequencecorrelationstarts here.

2nd sequencecorrelationstarts here.

Ns repetitions Ns repetitions

Pre-/post-fixes preventmultipath destruction ofzero sidelobe properties.

... ...

ann = N−NCP

N−1bn

n = N-NCP

N−1an

n = 0

NCP−1bn

n = 0

NCP−1aN aNaN

Figure 7.16 Illustration of how length-N complementary Golay sequence pair, aN and bN , maybe used to construct a training sequence that enables channel impulse response estimation throughcomplementary correlation. Pre- and post-fixes are added before and after Ns repetitions of eachcomplementary sequence to prevent excess multipath from disrupting zero-sidelobe properties. At thereceiver we correlate with the channel distorted versions of each N -length complementary sequenceand add them together to yield an estimate of a single tap. Delayed correlations are computed foreach tap, and each of the Ns repetitions is used to improve estimate robustness in the presence ofnoise.



7.5 Future PHY Considerations 465

Spec

tral

eff

icie

ncy

(bit/

sec/

Hz)

SNR (dB)0

0

0.5

1

1.5

2

2.5

3

3.5

5 10 15 20

Unquantized Capacity1-bit ADC Capacity2-bit ADC Capacity3-bit ADC Capacity

Figure 7.18 Capacity comparison with 1-, 2-, 3-, and, ∞-bit ADC precision for a discrete memory-less channel with perfect synchronization [SDM09].




(a) Analog equalization

analog signal

analog signal

AnalogEQ

ADC

ADC

DAC

Digitalreceiver

Digitalreceiver

Feedbackfilter

Symbol

decisions

Symbol

decisions

Energy-normalized

Energy-normalized

Tap delay line coefficients

(b) Mixed-signal equalization

Figure 7.19 Analog and mixed signal equalization architectures can reduce the ADC bit resolu-tion of the overall receiver (assuming that synchronization and other receiver functionality can bemaintained). Mixed signal equalization can be considered a DFE with an analog feedback filter anda digital feedforward filter.



7.5 Future PHY Considerations 467

Range reference (r)

Transmitter ULA Receiver ULA

8×8 Channel

Antenna element

θ φ

Figure 7.20 LOS MIMO channel with arbitrary uniform linear array (ULA) alignment and Nr =Nt = 8 elements on each ULA. The range reference of the link is denoted by r, and the total antennaarray lengths at the receiver and transmitter are Lr and Lt, respectively.



8.3 Beam Adaptation Protocols 485

−100 −80 −60 −40 −20 0 20 40 60 80 1000

2

4

6

8

10

12

14

16

18

Angle (degree)

Ant

enna

Pow

er G

ain

(dB

)

beams on codebook W3beams on codebook W2beams on codebook W1

Figure 8.7 A multilevel codebook proposed in [HKL+11] for wireless backhaul. Higher levels ofthe codebook have narrower beams, thus enhanced resolution [from [HKL+11, Figure 2] c© IEEE].



488 Chapter 8 Higher Layer Design Considerations for MmWave

1st time slot2nd time slot

Full duplex Half duplex Multi-hop

Figure 8.9 Different relay configurations with source, relay, and destination. In theory, a commu-nication link with a relay may exploit both the direct link from the source to the destination andthe indirect link through the relay. With a full duplex relay, the relay listens and retransmits at thesame time. A practical example of a full duplex relay is a repeater. With a half duplex relay, the relayeither transmits or receives and communication may be broken into two phases: transmission fromthe source, and transmission from the relay. In a multi-hop channel, the relay is half duplex and thesource to destination link is not exploited (from [Hea10]).




0 50 100 150 200 2500

0.360.6

11.2

2

2.5

3

4

5

Coverage range (m)

PHY

dat

a ra

te (

Gbp

s)

LOS-60GHz

LOS-5GHz

NLOS-60GHz

NLOS-5GHz

Figure 8.12 Coverage range and data rate at the physical layer for a 5 GHz link and 60 GHzlink under two different channel conditions: LOS and NLOS. It can be seen that the microwave linkprovides higher coverage at the expense of smaller data rates. A multi-band protocol could obtainthe rate benefits of 60 GHz and the coverage benefits of lower frequencies [from [YP08, Figure 1]c© IEEE].




Data (pixels)

UEPMSB

MSB only No retransmission

scheduled

MSB in error LSB in error

Transmission

Acknow-

ledgment

Transmission

Acknow-ment

Retransmission

MSB only

LSB

MSB LSB

MSB LSB

MSB LSB

MSB LSBUEP

216 bytes 8 bytes RS (224, 216, t = 4)

ParityBad codeword

RS code 1 RS code 100RS code 2 RS code j

RS code 1 RS code 2

RS code 1 RS code 2

RS code 1

RS code 100

RS code 100

RS code 100RS code 2

RS code j

RS code j

RS code j

…

…

…

…

Video packet 1 (partition 1)




(e)

(b)(a)

(d)(c)

Stream 1

Partitionindex

MSB(video pixel)

LSB(video pixel)

Payload-N(video packet)

Payload-1(video packet)

Videoheader

MACheader

PHYheader

Partition 1packet

Pixeltype 1

Horizontal

1234...

1 2 3 4 5 . . .

Pixeltype 2

Pixeltype 3

Pixeltype 4

Partition 2packet

Partition 3packet

Partition 4packet

HCS

MSBCRC

LSBCRC

Length MCS

Payload

MCS0 (EEP) MCS1 or MCS 2

Length MCS

Payload

H and Vpositions

Videoframe #

Interlace field

Video control field of payload NIn

crea

sing

impo

rtan

ce

I channel

Q channel

Firsthalf

g1

g2

Secondhalf

Rate r1

Rate r2

MUX

Stream 2

StreamN–1

RaterN–1

Rate rN Stream N

…

Tim

eV

ertic

al

Type 1Type 2Type 3Type 4

Figure 8.13 Different strategies for supporting video in mmWave systems. (a) Pixel partition-ing. (b) Frame format. (c) Uncompressed video with automatic repeat request. (d) Unequal errorprotection. (e) Error concealment using Reed-Solomon codes. [From [SOK+08, Figure 13] c© IEEE]



Data (pixels)

UEPMSB

MSB only No retransmission

scheduled

MSB in error LSB in error

Transmission

Acknow-

ledgment

Transmission

Acknow-ment

Retransmission

MSB only

LSB

MSB LSB

MSB LSB

MSB LSB

MSB LSBUEP

216 bytes 8 bytes RS (224, 216, t = 4)

ParityBad codeword

RS code 1 RS code 100RS code 2 RS code j

RS code 1 RS code 2

RS code 1 RS code 2

RS code 1

RS code 100

RS code 100

RS code 100RS code 2

RS code j

RS code j

RS code j

…

…

…

…





(e)

(b)(a)

(d)(c)

Stream 1

Partitionindex

MSB(video pixel)

LSB(video pixel)

Payload-N(video packet)

Payload-1(video packet)

Videoheader

MACheader

PHYheader

Partition 1packet

Pixeltype 1

Horizontal

1234...

1 2 3 4 5 . . .

Pixeltype 2

Pixeltype 3

Pixeltype 4

Partition 2packet

Partition 3packet

Partition 4packet

HCS

MSBCRC

LSBCRC

Length MCS

Payload

MCS0 (EEP) MCS1 or MCS 2

Length MCS

Payload

H and Vpositions

Videoframe #

Interlace field

Video control field of payload N

Incr

easi

ng im

port

ance

I channel

Q channel

Firsthalf

g1

g2

Secondhalf

Rate r1

Rate r2

MUX

Stream 2

StreamN–1

RaterN–1

Rate rN Stream N

…

Tim

eV

ertic

al

Type 1Type 2Type 3Type 4

Figure 8.13 Different strategies for supporting video in mmWave systems. (a) Pixel partition-ing. (b) Frame format. (c) Uncompressed video with automatic repeat request. (d) Unequal errorprotection. (e) Error concealment using Reed-Solomon codes. [From [SOK+08, Figure 13] c© IEEE]



8.6 Multiband Considerations 499

Backhaul connections between all base stations

Microwavelink

mmWavelink

Interference

Figure 8.14 A model for coexistence between mmWave and microwave cellular where themicrowave cellular network forms an umbrella network to facilitate the management of manymmWave communication links and to simplify functions like handoff. On the left side, a mobiledevice may connect either to a microwave or mmWave base station or to both simultaneously usingthe phantom cell concept. Interference on the microwave frequencies comes from other microwavebase stations and on the millimeter wave frequencies from other millimeter wave small cells. Asshown in Chapter 3 and elsewhere, the directionality of the beam patterns reduces the impact ofmmWave interferers [BAH14][SBM92][RRE14][RRC14][ALS+14].




-5 0 5 10 15 200.4

0.5

0.6

0.7

0.8

0.9

1

SINR threshold in dB

SIN

R c

over

age

prob

abili

ty

Rc = 50 m

Rc = 100 m

Rc = 200 m

Rc = 300 m

(a) (b)

50 100 150 200 250 3000.4

0.5

0.6

0.7

0.8

0.9

1

Avg. cell radius Rc in meters

Cov

erag

e pr

obab

ility

wit

h SI

NR

> 10

dB

Figure 8.15 SINR coverage probability with different base station densities, where Rc =√

1/πλand λ is the density of base stations.




Cel

l thr

ough

put

in M

Hz

3177.2

432

929

1.2 1.4124

848

0100200300400500600700800900

1000

Average5% tail

mmWave MUMassive MIMOSU-MIMSIS

Figure 8.16 Comparison of cell throughput of mmWave networks and microwave networks.



512 Chapter 9 MmWave Standardization

Freq

uenc

y al

loca

tions

(G

Hz)

65

60

55North America Europe Australia Korea Japan

Figure 9.1 International frequency allocation for 60 GHz wireless communication systems.



9.3 IEEE 802.15.3c 517

ParentPNC

ChildPNC

Parentpiconet

Neighborpiconet

Figure 9.7 Neighbor piconet. The parent piconet (controlled by the set-top box) manages coexis-tence with the neighbor piconet (controlled by the personal computer).




Channel 1(center at 58.32 GHz)




edge at57.2406 GHz

edge at59.4006 GHz

edge at61.5600 GHz

edge at63.7200 GHz

edge at65.8800 GHz

Figure 9.8 Channelization in IEEE 802.15.3c provides four different channels for mmWave PHY.




... ...

......

~1,830 MHz

2,640 MHz

352 Data and pilot sub-carriers 141 Zeroed guard sub-carriers

3 Zeroed DC sub-carriers

... ...

16 Zeroed reserve sub-carriersSpectral mask

Figure 9.15 OFDM symbol formatting in the HSI PHY in IEEE 802.15.3c. The sub-carrier fre-quency spacing is 5.15625 MHz for all 512 sub-carriers. Three null DC tones prevent carrier feedthrough as well as ADC/DAC offset problems. The guard tones are usually nulled to meet spectralmask requirements, although customized guard tone values may optimize front end effects.




... ...

~1,760 MHz

2,538 MHz

... ...

3 Zeroed DC sub-carriersSpectral mask

141 Zeroed guard sub-carriers352 Data and pilot sub-carriers

Figure 9.16 OFDM symbol formatting for the HRP. The sub-carrier frequency spacing is ≈ 4.96MHz for all 512 sub-carriers. Three DC tones and all guard tones are nulled.



9.3 IEEE 802.15.3c 533

... ...

~ 92 MHz

317.25 MHz

34 Data and Pilot Subcarriers 91 Zeroed Guard Subcarriers

3 Zeroed DC Subcarriers

... ...

Spectral mask

98 MHz

Figure 9.18 OFDM symbol formatting for the LRP. The sub-carrier frequency spacing is 2.48 MHzfor all 128 sub-carriers. There are 37 data and null subcarriers, each with a subcarrier width of 2.48MHz, resulting in an occupied bandwidth of 91.76 MHz (∼ 92 MHz). The specified spectral maskpassband bandwidth (at 10 dB down) is 98 MHz, allowing for roll-off in the LRP mode. Three DCtones and all guard tones are nulled.




skewing

QI

1.25 dd

d

d

d

Figure 9.24 UEP Type 3 through skewing of 16-QAM. Here, the minimum distance betweenin-phase constellation points (where MSB bits are mapped) is increased by a factor 1.25.



9.3 IEEE 802.15.3c 545

Quasi-omnipatterns

Sectorpatterns

Hi-respatterns

Beampatterns

Figure 9.29 Four levels of patterns in antenna beamforming codebook for eight-element uniformlinear array (patterns visualized on the azimuthal plane (top view) for a vertical array orientation)[802.15.3-09].



9.3 IEEE 802.15.3c 547

Hi-res beam

Elevation

Azimuthal

Figure 9.30 High-resolution beam cluster in 3-dimensional space.



9.3 IEEE 802.15.3c 549

Q

I

StandardOOK

DAMIOOK

Q

I

Figure 9.32 OOK and DAMI constellations for optional use within the SC-PHY [802.15.3-09].




Host

A/Vcontroller

Adaptationsub-layer

MACsub-layer

PHYsub-layer

Privacy

Stat

ion

man

agem

ent e

ntity

Beamsteering

andtracking

A/Vformatting

Contentprotection

MAC

LRP HRP

Figure 9.33 Layering of WirelessHD device.



9.4 WirelessHD 553

Video/Audio

Source ProcessorSubsystem

Content ProtectionSubsystem

Audio/Video/Data Multiplexer

Data

HRP Data

Switch

Figure 9.34 WirelessHD packetizer diagram [Wir10].




Type CPLCP

Type BPLCP

Type APLCP

Type CPHY

Type BPHY

Type APHY

Arrayconfig.

Type AMAC

Type BMAC

Type CMAC

... Control

Ant

enna

arra

yPH

YM

AC MAC layer

managemententity (MLME)

PHY layermanagement

entity (PLME)

Devicemanagemententity (DME)

Protocol adaptation layer (PAL)

Figure 9.35 Protocol structure of ECMA-387 [ECMA08].




... Control

60 GHz PHYOOB PHY

60 GHz MACOOB MAC

Convergence MAC

Higher layers

Figure 9.36 Out-out-band (OOB) control channel layered architecture in ECMA-387 [ECMA08].




Link change interval

DATA(to STA)

DATA(to STA)

DATA(to STA)

ACK(to STA)

ACK(to Relay)

ACK(to STA)

Never received bydestination STA

No ACK Relay RX focused on source,TX focused on destination

STA(dest)

STA(source)

Relay

DATA(to Relay)

Link change interval Link change interval

Figure 9.46 Transmissions by source STA, amplify-and-forward relay, and destination STA in alink switching example in normal mode.




...STP

... ...

0 1

167

168

169

335

Pair 167Pair 1Pair 0

Figure 9.53 Static tone pairing (STP) in MCS 13-17. Even and odd sub-carriers are paired (outof 336 sub-carriers total) and mapped to maximize the minimum sub-carrier distance between evenand odd sub-carriers (168 sub-carrier distance).


Group 0DTP

0 1

164

1672 3

165

166

G41

+ 1

68G

41 +

169

G41

+ 1

70G

41 +

171

G0 +

168

G0 +

169

G0 +

170

G0 +

171

...

............

Group 41

Figure 9.54 Dynamic tone pairing (DTP) in MCS 13-17. For DTP, the group pair index (GPI) isgiven to the PHY and is defined by the transmitter. The mapping GPI is hence a permutation whereGPI : 0, 1, . . . , 41 → 0, 1, . . . , 41 and GPI : k → Gk. Note that although the even elements ofeach group have a fixed mapping, the odd elements may be mapped more generally. In other words,Gk may vary for a fixed k, depending on the link configuration. The ends of the DTP transformedsub-carriers are not shown to maintain generality, although one of the 42 groups must be mappedto the last DTP group in practice.

Documents

Color figures from Rappaport et al., Millimeter Wave Wireless