5G application trends Orange, Ericsson, “Trafﬁc Model for legacy GPRS MTC” (February 15-19, 2016), document GP 160060, 3GPP GERAN meeting #69 Trafﬁc characteristics of deployed

5G Applications trends and technology needs

Sven MattissonEricsson Research, Lund

5G Applications trends and technology needs | NORCAS 2017 | Page 2

Envisioned 5G plans

Source: Ericsson Mobility Report


Estimated 5G traffic



5G when?



5G research challenges

A set of integrated radio-access technologies jointly enabling the long-term networked society

Ultra-reliable communication

Inter-vehicular / vehicular-to-roadcommunication

Massive machine-typecommunication

Ultra-dense deploymentsDevice-to-device

communication and cooperative devices

Multi-hopcommunication


What is MTC?

Three different 3GPP machine-type communication initiatives:› EC-GSM-IoT (Extended Coverage - GSM - Internet of Things)› LTE-M (LTE M2M)› NB-IoT (Narrow Band Internet of Things, LTE add-on)


5G IoT traffic

Massive IoT traffic scenario

A dense urban environment with 10,000 households per km2 – similar to the central area of London, Beijing or New York – was used as the base for a massive IoT services scenario. A selection of connected device types were assumed to be deployed in the area, including water, gas and electricity meters, vending machines, rental bike position monitors and accelerometers in cars1 monitoring driver behavior. Traffic characteristics for each device are summarized in the diagram above.2 The number of connected devices used in this scenario represents a mature, large scale massive IoT scenario. During an initial rollout phase device densities will be lower and the corresponding traffic load will not be as high.

The services represent a realistic range of massive IoT use cases that are expected to be deployed in an urban environment.3

Deployment environment and traffic models differ for these services: a remote-controlled meter may face an indoor coverage challenge, while a device mounted on a bike is usually found outside. The traffic intensity from meters may be once per day, whereas other devices may need to transmit every 10 minutes.

The data traffic for massive IoT devices is small; the typical data packet for a service is about 100-150 bytes, accounting for a payload of the device ID, time stamp and report data values.

Additionally, each packet has IP overhead and higher layer headers of around 65 bytes; the Media Access Control (MAC) layer overhead is 15 bytes, and standard control signaling within the mobile network is 59 bytes per event for uplink. In total, each event generates around 250-300 bytes to be transmitted by the IoT device.

The figure on the following page shows the resulting traffic demand. It clearly demonstrates that, despite the very high device density, the small traffic per device limits the traffic per area unit to a few kilobits per second (kbps) per km2. As a comparison, mobile broadband traffic is approaching gigabit per second (Gbps) per km2 in dense urban areas.

1 Calculation based on an average of one car per household and every fourth in traffic 2 In this scenario, the traffic is uplink dominated, as downlink traffic for application acknowledgement (ACK)

and control plane signaling (RRC) between the device and the radio access network is comparably small 3 Orange, Ericsson, “Traffic Model for legacy GPRS MTC” (February 15-19, 2016), document GP 160060, 3GPP GERAN meeting #69

Traffic characteristics of deployed massive IoT connected devices in a city scenario

WATER METERS

100 bytes

12 hours

10,000/km2

Typical message size

Message interval

Device density

ELECTRICITY METERS

100 bytes

24 hours

10,000/km2

GAS METERS

100 bytes

30 minutes

10,000/km2

VENDING MACHINES

150 bytes

24 hours

150/km2

BIKE FLEET MANAGEMENT

150 bytes

30 minutes

200/km2

PAY-AS- YOU-DRIVE

150 bytes

10 minutes

2,250/km2

NOVEMBER 2016 ERICSSON MOBILITY REPORT IoT IN FOCUS 31

Estimated traffic from massive IoT devices in a city scenario with 10,000 households per km2.

Per device type:> 100–150 B/msg> 1–150 msgs/day> 0.05–1 #/household

Source: Ericsson Mobility Report – November 2016


NB-IoT Deployment scenarios1. Stand-alone

NB-Io

T

GSM

NB-Io

T

NB-Io

T

GSM

GSM

NB-Io

Tf

[MHz]

GSMisreplacedbyaNB-IoTcarrier.

Future:SeveralNB-IoTresourceblocksaregroupedtogether.

› Two GSM carriers can be replaced by one NB-IoT carrier.

› The NB-IoT system can use more than 1 resource block.

– In the stand-alone case, they are grouped together.

f[MHz]

1PRB=180kHz


…NB-IoT Deployment scenarios

LTE

NB-Io

TLTE

NB-Io

T

f[MHz]

LTEcarrierbandwidth

f[MHz]

LTEcarrierbandwidth

OneormoreNB-IoTcarriers(dependingonLTEbandwidth)canbedeployedintheguardbandoftheLTEcarrier.

OneormoreNB-IoTcarriersisreplacingresourceblocksintheLTEcarrier.

Guardband

2. LTE guard band

3. In-band LTE

NB-Io

T

NB-Io

T

f[MHz]

LTEcarrierbandwidthLTE

Power

boost


5G mm-wave signal properties

› CP-OFDM with windowing/filtering› PAPR on the order of 10dB

CP-OFDM

BWsig: > 0.9 BWch

guard band 10% →1%BWch: 50 – 200 (400) MHz

sub-carrier spacing{60, 120, (240) } kHz

• • •• • •


AAS — array/advance/adaptive antenna systems/steering

beam generationLet one single data stream (i.e. beam) connect to M antenna elements, then we can bombineeither in the analog or digital domain (both RX and TX). Similarly we can tap of several copiesof the antenna signal to N di↵erent cominers to get multiple beams.

w1

wM

w1

wM

w1,1

wM,1

w1,N

wM,Nanalog BF digital BF

M antennas, one beam M antennas, N beams

Valid both for RX and TX.c�Ericsson AB 2015 Ericsson Confidential 12𝑃"# = 𝑃%#𝐺%#𝐺"#

𝜆4𝜋𝑑

+

Friis’ transmission equation→ use many antennas at mm-wave

10x10 cm, 15GHz


Above 6 GHz emissionmask

-13 dBm/MHz

-5 dBm/MHz

BW

BW10

12 ·NRB · fSCS

c�Ericsson AB 2015 Ericsson Confidential 5

Regulatory requirements

› 4G has 90/10% carrier/guard allocation› 5G will use a variable allocation

– up to 98.3/1.7% below 6GHz – up to 95/5% above 6GHz– filter complexity and power consumption– group delay incurred latency– interference in mixed NR and LTE cells

Spectrum confinement

95%

5%

90%

10%

Above 6GHz TX mask


1.96e+09 1.98e+09 2e+09 2.02e+09 2.04e+09

-100

-50

0

Frequency (Hz)

Spec

tral d

ensi

ty (d

Bm/M

Hz)

IP3 30dBmgain+NF 40dB

20MHz AWGN signal with IM3 (in-band IM3 dotted)

P=20 dBmUE E-UTRA mask

kTB

…regulatory requirements

Sample 20MHz PA spectrum, with noise and IM3, @ 2GHz

kTB

Duplex distanceRX band

› Emission in own FDD RX band has to be below the RX noise floor› TDD isolates TX from RX in time, e.g. for NB-IoT and 5G mm-wave bands


Selectivity testingAdjacent channel selectivityConducted

Selectivity test

+ RX demodPsig BER/FER

Psig = Pref sens + 6dBPint


OTA Selectivity test

RXPRX

PnoisePint


Current 3GPP approach

Selectivity = 𝑃,-. − 𝑃0,1

At ref. sens BER/FER

Time consuming when BER/FER threshold is low as well as when there are many RX paths (AAS)

Tentative approach for array antenna systems (AAS)

Simpler, no need for conducted sensitivity, accurate and agnostic but need to be completed with limited over-the-air tests based on minimum radiated sensitivity with a WANTED signal.

Selectivity ~𝑃"# − 𝑃-9,0:

with 𝑃"#, 𝑃-9,0: antenna referred


Of particular interest in 3GPP for mm-wave applications

› Receiver noise figure– Insertion losses in RF filter, switches and substrate– Low-noise amplifier (LNA) input-referred noise– Dynamic range

› Adjacent-channel selectivity (ACS)› Blocking

› Power amplifiers– Linearity

› Error-vector magnitude (EVM)› Adjacent-channel leakage ratio (ACLR)

– Efficiency at back-off› Peak-to-average power ratio (PAPR)


Cut-off frequency and scalingPA technology comparisonCut-off frequency and scalingFundamental trade-off

between voltage and frequency

scaling

Scaling sets speed and integration potential

> Si is superior for integration (so far)

> GI (fT ) = 1, GP(fmax) = 1

gd

s

rg

cgs

cgd

gm

go

fT =gm

2⇡(cgs + cgd), fmax =

fTp4 rg (2⇡fT cgd + go)

c�Ericsson AB 2016 36 (48)Source: del Alamo


Receiver noise figureA simplified receiver model can be derived by lumping the front-end (FE), analog/RF receiver (RX) and ADC into three cascaded blocks.

Receiver noise figureA simplified receiver model can be derived by lumping the front-end (FE),analog/RF receiver (RX) and ADC into three cascaded blocks.

| {z }IL

filter, SW, routing

| {z }DR,CPVFS

| {z }F ,BW ,G

LNA, mixer, analog base band

ADCRXFE

Psig ,N0 SNR

F =Nout

Nin ·G=

Nout · SinNin · Sout

=SNRin

SNRout

NF = 10 · log10(F )

c�Ericsson AB 2016 2 (25)


| {z }IL

filter, SW, routing

| {z }DR,CPVFS

| {z }F ,BW ,G


ADCRXFE

Psig ,N0 SNR

F =Nout

Nin ·G=


=SNRin

SNRout

NF = 10 · log10(F )



Cascade noise factorFriis' formula can be used to find the noise factor at the antenna connector as (linear units unless noted)

Cascade noise factorFriis’ formula [21] can be used to find the noise factor at the antennaconnector as (linear units unless noted)

F = IL ·✓FLNA +

FADC � 1

G

◆.

For example, assuming

NF = 5dB

IL = 0.7 dB

FLNA = 0.8 dB

9>=

>;) FADC � 1

G= 1.5 (1.76 dB)

This factor 1.5 can be used to represent the net allowable SNRdegradation due to the ADC noise floor, base-band processing losses,variability margins etc.


Cascade noise factorFriis’ formula [21] can be used to find the noise factor at the antennaconnector as (linear units unless noted)

F = IL ·✓FLNA +

FADC � 1

G

◆.


NF = 5dB

IL = 0.7 dB

FLNA = 0.8 dB

9>=

>;) FADC � 1

G= 1.5 (1.76 dB)

This factor 1.5 can be used to represent the net allowable SNRdegradation due to the ADC noise floor, base-band processing losses,variability margins etc.



This factor 1.5 can be used to represent the net allowable SNR degradation due to the ADC noise floor, base-band processing losses, variability margins etc.


Wideband LNA noise-figure trends

LNA Noise figuretrends

0 20 40 60 80 1000

1

2

3

4

5

6

7

Carrier frequency (GHz)

NF (d

B)

28nm45nm SOI65nm90nm130nm SOI130nm SiGe250nm GaN150nm GaNtrend

LNA noise figure vs. frequency

By inserting(F0 = 1.12 (0.5 dB)

ft = 168 (GHz)

into the Fmin(fc) equation weget the depicted trend curvethat matches the bestreported noise figures.


By inserting

into the Fmin(fc) equation we get the depicted trend curve that matches the best reported noise figures.

LNA Noise figuretrends

0 20 40 60 80 1000

1

2

3

4

5

6

7

Carrier frequency (GHz)

NF

(dB)

28nm45nm SOI65nm90nm130nm SOI130nm SiGe250nm GaN150nm GaNtrend

LNA noise figure vs. frequency

By inserting(F0 = 1.12 (0.5 dB)

ft = 168 (GHz)

into the Fmin(fc) equation weget the depicted trend curvethat matches the bestreported noise figures.


[Tuned LNAs can do better at the cost of bandwidth]


Total noise figure

› The receiver noise figure increases with the carrier frequency› Total noise figures of 10, 12, and 14dB assumed for 30, 45, and

70GHz, respectively (5dB at 2GHz)


| {z }IL

filter, SW, routing

| {z }DR,CPVFS

| {z }F ,BW ,G


ADCRXFE

Psig ,N0 SNR

F =Nout

Nin ·G=


=SNRin

SNRout

NF = 10 · log10(F )


1.3dB 0.9dB 1.7dB 2.7dB 0.3dB 30HGz

add 3dB variability and implementation margin


PA material propertiesSi GaAs SiC InP GaN Diamon

dμn 1400 10000 800 5400 2000 1900 cm2/Vs

Eg 1.1 1.4 3.26 1.34 3.4 5.45 eV

Ebr 0.3 0.4 3.5 0.5 3.3 5.6 MV/cm

vsat 1.0 1.5 2.0 0.67 2.5 2.7 107 cm/s

JFOM 0.48 0.95 11 0.53 13 24 1012 V/s

Rel. JFOM

1.0 2.0 23 1.1 28 50

Johnson’s figure of merit (JFOM) is defined as <=>?@AB+C

[data1 from Mishra and Rais-Zadeh]

Silicon technology has by far the best integration, cost and scaling properties but suffers from low break-down voltage (Ebr ). If it can be done in (CMOS) Si it will. . . 1 Data is somewhat contradictory


PA capabilities - Psat

P sat

[dBm

]

0

5

10

15

20

25

30

35

40

45

10 100

CMOSBulk

CMOSSOI

GaNHEMT

Frequency [GHz]

-20dB/dec

Source: Johnson, 3GPP R4-1610279


ACLR

For AWGN-like amplitude-modulated signals (like LTE) we have

𝐴𝐶𝐿𝑅 = HAIJKLMANO

= 3 KHMHAIJ

+,

assuming a cubic nonlinearity 𝑓 𝑥 = 𝑎T𝑥 +𝑎V𝑥V, with 𝐼𝑃V =XVYZYM

� .Source: Pedro

1.98e+09 2e+09 2.02e+09 2.04e+09-200

-180

-160

-140

-120

-100

-80

-60

Frequency (Hz)S

pect

ral d

ensi

ty (d

Bm

/Hz)

AWGN signal with IM3 due to cubic nonlinearity (in-band IM3 dash-dotted)

IP3 30dBm

P=20 dBm

10 dBm

0 dBm

𝑃Y?1

𝐼𝑀VY]^


Throughput loss vs ACIR

› Adjacent channel interference ratio

𝐴𝐶𝐼𝑅 =1

1𝐴𝐶𝐿𝑅 +

1𝐴𝐶𝑆

› Adjacent services outside own band may aggravate ACLR, though

ACIR [dB]

Thro

ughp

ut lo

ss [%

]

Source: 3GPP R4-1610477

𝐴𝐶𝑆𝐴𝐶𝐿𝑅

𝑓a 𝑓Y]^


ACLR vs output power

Source: 3GPP R4-168391, R4-1610279

ACLR 30 dB 40 dB

CMOS 16dBm 9dBm

GaN 28dBm 22dBm


𝑃𝐴𝐸 =𝑃9c. − 𝑃,-

𝑃de

Typical bias point(before DPD)

PAE vs ACLR

ACLR: 30 dB 40 dBCMOS 8% 2%GaN 11% 4%

Source: 3GPP R4-168391, R4-1610279


Field tests of 5G technology

› Green flag waves on 5G in Indianapolis, May 2017.– https://www.youtube.com/watch?v=Dw2GT95Vyxc&index=1&list=PLsn61Zheh8ije3EjK_NcyGUAF2PJe34_a

› New world record speed with 5G, February 2017.– https://www.youtube.com/watch?v=UOCM_91n90U&index=2&list=PLsn61Zheh8ije3EjK_NcyGUAF2PJe34_a


Discussion› 5G will augment existing wireless systems with a range of new applications, including, e.g., AAS

for high capacity mm-wave cells and low-power MTC.› Standardization is ongoing but legacy requirements will probably be extended up to 6GHz. Band

allocation above 6GHz will be addressed by WRC-19.› AAS will require OTA testing and new methods will need to be developed to account for spatial

filtering of the transmitted signals as well as received signals.› The need for high integrated mm-wave systems with many transceivers and antennas will require

careful and often complex consideration regarding the power efficiency and heat dissipation in small area/volume affecting the achievable performance as well as over-the-air testing.

› Transceivers for mm-wave frequencies will have increased power consumption due to higher BW, and, considering the thermal challenges given the significantly reduced area/volume for mm-wave products,

– the complex interrelation between receiver linearity, NF, bandwidth and dynamic range in the light of power dissipation should be considered,

– as well as transmitter achievable power versus efficiency as well as linearity.


Acknowledgement

Numerous colleagues have contributed to the activities reported here. In particular I would like to thank:

› Farshid Ghasemzadeh,› Stefan Parkvall, › Kenneth Sandberg, and,› Lars Sundström

for letting me use their material.


References

1. Ericsson Mobility Report. URL: http://www.ericsson.com/mobility-report/.

2. E. Johnson, "Physical limitations on frequency and power parameters of transistors," 1958 IRE International Convention Record, New York, NY, USA, 1965, pp. 27-34.

3. 3GPP R4-164226, On mm-wave technologies for NR.4. 3GPP R4-166526, Discussion on BS and UE noise figure for mm-waves.5. 3GPP R4-168391, On mm-wave ACLR.6. 3GPP R4-1610279, On mm-wave ACLR for 45 and 70 GHz.7. 3GPP R4-1610477, Simulations results for coexistence studies in 30GHz.8. U. K. Mishra et al.“GaN-Based RF Power Devices and Amplifiers 04414367.pdf”. In: 96.2 (2008), pp. 287–305. ISSN: 0018-

9219. DOI: 10.1109/JPROC.2007.911060. 9. M. Rais-Zadeh et al. “Gallium nitride as an electromechanical material”. In: Journal of Microelectromechanical Systems 23.6

(2014), pp. 1252–1271. ISSN: 10577157. DOI: 10.1109/JMEMS.2014.2352617. 10. J. A. del Alamo. Si CMOS for RF Power Applications. Last visited 2016-03-17. 2005.

URL: http://www-mtl.mit.edu/~alamo/pdf/2005/RC-108.pdf. 11. J. C. Pedro and N. Borges de Carvalho. “On the Use of Multitone techniques for Assessing RF Components’ Intermodulation

Distortion”, IEEE Transactions on Microwave Theory and Techniques (Dec. 1999), pp. 2393–2402.


Documents

5G application trends Orange, Ericsson, “Trafﬁc Model for legacy GPRS MTC” (February 15-19, 2016), document GP 160060, 3GPP GERAN meeting #69 Trafﬁc characteristics of deployed