1

© Eridan, 2019

IoT at Millimeter Wave Frequencies: Finding Hope

Earl McCune

CTO, Eridan Communications

2

© Eridan, 2019

Outline

• What is the problem?

• Why is this the way it is?

• What can be done to improve the situation?

3

© Eridan, 2019

Wireless Link Operating Objectives

• communication distance

• power efficiency and link efficacy

• bandwidth

• directivity

• penetration of foliage

• penetration of building structures

• environmental conditions

• human safety

• Costs: capital and operational

4

© Eridan, 2019

Circuit Level

• Linear circuit (PA) efficiency

• Signal bandwidth vs. power draw

• Circuit linearity metric

• Circuit bandwidth (multi-band)

• Ohm’s Law

• Transistor speed (fT)

Key Physics

System Level

• Link Efficacy

• Antenna gain / directivity

• Modulation vs. TX power

• Bandwidth Efficiency (Efficacy)

• Human safety

• Occluded environments

5

© Eridan, 2019

Inherent conflicts

• Modulation vs. Circuit Efficiency and Link Efficacy

• Communication distance vs. coverage

• Signal bandwidth vs. Power draw

6

© Eridan, 2019

Circuit Efficiency

• Efficiency applies to any situation where a conservation principle applies- e.g.

• Efficiency improvement is a solved problem below 5 GHz- Though it is not yet widely implemented

• This technology does NOT scale to mmW frequencies

• For the near term, we appear to be doomed……

PDCPIN

POUT

PD

11% Efficiency

DC IN OUT DP P P P+ = +

7

© Eridan, 2019

0 0.5 1 1.5 2 2.5 3 3.50

0.2

0.4

0.6

0.8

1

VDS

(V)

Dra

in C

urr

ent

(A)

Incre

asin

g V

GS

Increasing PD

Vk(VGS)

Transmitter Physics• Linear transmitters

cannot be efficient– A direct consequence

of Ohm’s Law

• Efficient transmitters cannot have circuit linearity– Class-E approximates

this profile

• If the amplifier transistor is too slow to follow the paths of efficient operation– Then all of the

techniques developed for efficient operation disappear

– Linear operation still remains (see #1)

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.5 1 1.5 2 2.5 3 3.5

I C(A

)

VCE (V)GaAs HBT

Envelope PDF

Signal envelope

Linear Transmitter Efficient Transmitter

8

© Eridan, 2019

Why is Efficiency a Special Problem at millimeter wave frequencies?

• mmWave frequencies are much closer to

transistor speed boundaries

Transistor behavior changes at these frequencies

• Output power drops

• Circuit gain is lower

• Compression begins sooner

• Linear operating range shrinks

• 5G-New Radio uses identical modulation schemes

at all frequencies

100

101

102

103

0

20

40

60

80

100

fT / f

o

Achie

vable

Eff

icie

ncy (

%)

RL/RON=10

RL/RON=30

RL/RON=100

-10

0

10

20

30

40

50

0.001 0.01 0.1 1 10 100 1000

Ava

ilab

le G

ain

(d

B)

Frequency (GHz)mmWmW

9

© Eridan, 2019

Millimeter Wave Power Drop

As frequency increases, the transition times eventually do not settle into ON

and OFF states: slew-rate limiting

• At low frequencies, the transition time is negligibly short and the transistor does switch

• Largest possible sinewave is 2(2/p) times the squarewave magnitude (=1.27x) [+2.1 dB]

• When slew-rate limiting is reached, the largest possible sinewave output is reduced

• Largest possible sinewave is 2(2/p)2

times the squarewave magnitude (=0.81x) [-1.8 dB]

-1.5

-1

-0.5

0

0.5

1

1.5

0 0.5 1 1.5 2

-1.5

-1

-0.5

0

0.5

1

1.5

0 0.5 1 1.5 2

10

© Eridan, 2019

PA Thermal Management

• Available signal swing range is lower at mmW frequencies• Signal PAPR further reduces output power• More circuit stages are needed to make up total needed gain

0%

1%

10%

100%

0 2 4 6 8 10 12 14 16 18

TX E

ffic

ien

cy

Signal PAPR (dB)

< 3 GHz

3-6 GHz

mm Wave

2G 2.5G 3G LTE 5G-NR

0

10

20

30

40

50

60

70

80

90

100

0% 10% 20% 30% 40%

Rat

io t

o O

utp

ut

Po

we

r

Circuit Energy Efficiency

Power Supply Size

Heatsink Size

5G-NR Sub-6

5G-NR mmWave

0

0.005

0.01

0.015

0.02

0.025

0.03

0 0.5 1 1.5 2 2.5 3 3.5

I C(A

)

VCE (V)GaAs HBT

High Frequency

Low Frequency Cu & Fe

Al

( ) 20010dBPAPR

dBPAPR

− =

11

© Eridan, 2019

Defining Amplifier Linear Range

P1dB is a poor definition of amplifier linearity range

Linearity = constant gain, independent of input value

Gain is constant only when S = dPOUT/dPIN = 1.00 dB/dB

Gain expansion when S > 1

Gain compression when S < 1

Nonlinear gain is distorting to the waveform

Linearity: gain must be independent of the input voltage

Proposal:

Amplifier linearity is defined when S = 1.00 within an error

band equal to the allowed distortion

0

5

10

15

20

25

30

35

0 0.01 0.02 0.03 0.04

Gai

n (

linea

r)

Input Voltage (Vrms)

1

10

100

0.0001 0.001 0.01 0.1

Gai

n (

linea

r)

Input Voltage (Vrms)

S=1.2

S=1.1

S=1.0S=0.9S=0.8

S=1.2S=1.1S=1.0S=0.9S=0.8

± x%S=1.0

( )1

,

0

0

S

out rms in

in

V Vgain S g

V V

−

= =

12

© Eridan, 2019

Practical Linear PA: Range and Efficiency

5G-NR EVM spec is -30 dB

Corresponds to 3.3% distortion

Here, this distortion limit is well below P1dB

Sets the upper limit of signal output power (PEP)

Output backoff (OBO) = PSAT - PEP

Also sets the peak linear efficiency

Here 0 = 5% at the top of the linear range

Design goal: maximize the range of S = 1

only needs the POUT(PIN) curve

12

0

5

10

15

20

25

-20 -15 -10 -5 0 5 10

Ou

tput

Pow

er (

dB

m);

Rati

omet

ric

Pow

er G

ain

(dB

); P

AE

(%)

Input Power (dBm)

Output Power (dBm)

Peak PAE

Ratiometric Power Gain

0

0.2

0.4

0.6

0.8

1

1.2

-20 -15 -10 -5 0 5 10

Slo

pe

Po

wer

(d

B/d

B)

Input Power (dBm)

± 3%

P1dB

Linearrange

PSAT

PEP PSATOBO

13

© Eridan, 2019

Add Modulation

0%

1%

2%

3%

4%

5%

6%

0 5 10 15

PA E

ffic

ien

cy

Signal PAPR (dB)

0

5

10

15

20

25

-20 -15 -10 -5 0 5 10

Ou

tput

Pow

er (

dB

m);

Rati

omet

ric

Pow

er G

ain

(dB

); P

AE

(%)

Input Power (dBm)

Output Power (dBm)

Peak PAE

Ratiometric Power Gain

PAPR

PTX

PEP

• The linearity upper limit applies to the modulated

signal upper limit: Peak Envelope Power (PEP)

• Signal average power is lower than PEP by the

modulation PAPR

• Efficiency drops further

• Signal envelope statistics have a major influence

on the final result

• The efficiency upper bound is known

14

© Eridan, 2019

Why are we bothering with mmWaves?

Following the mantra

‘data rates must go higher’

Shannon showed that bandwidth efficiency is

constrained by link SNR

With a maximum link SNR

bps/Hz is upper bounded

More bps then requires more Hz bandwidth: mmW

2log 1 ; 0 1S

N

PR C UU

B B P

= +

15

© Eridan, 2019

0.1

1

10

100

1000

0.1 1 10 100

An

ten

na

Be

amw

idth

(de

gre

es)

Antenna Aperture (wavelengths)

Path Loss Issues

Actual LOS path loss is independent

of frequency

TX forms a EM field density (mW/m2)

What changes with frequency is the

RX antenna behavior

Scales with wavelength

mmW wavelength l is much shorter

Larger area antennas must be

directive:

( )21

, ,4 4

R T T T RpP d P P G G

d

ll

p p

=

RX Antenna

TX

Antenna

LOS

path-loss

rms signal power

into the channel

TX

RX

lPTGT PR

GR

d

p

T

PEPP

PAPR=

$$$

directivity

≥ 2

16

© Eridan, 2019

Efficacy

• Has units (e.g. lumens/watt for lighting, bits per joule for digital

communication)

• Determines how effective a process is when using one resource

to provide another: result/resource

Power (watts) in

Energy (joules) in

Light (lumens) out

Data transfer

(bits) out

Digital CommunicationElectric lighting

bits/joulelumens/watt

17

© Eridan, 2019

Link Efficacy

• Unit arithmetic relates the desired efficacy to its comprising link parameters

▪ These parameters are NOTindependent from each other

• Complete efficacy model includes all link parameters: signal, hardware, environment, and physics

secsec

sec

EE

symbbb sym

JJ Watts

= = ( ),

220

02

10 10

10log

4( ) 10 10

dB

d dBdB

PAPR

p

EE sym T RSNRNFN M f G G d

kTB

l

p

−

−

=

( )2

1 cos360

RG p

=

−

18

© Eridan, 2019

Link Efficacy

• Unit arithmetic relates the desired efficacy to its comprising link parameters

▪ These parameters are NOTindependent from each other

• Complete efficacy model includes all link parameters: signal, hardware, environment, and physics

secsec

sec

EE

symbbb sym

JJ Watts

= =

2log ;b N M

sym=

2

bsymM =

OUTDC

PP

=

fsym

( ),

220

02

10 10

10log

4( ) 10 10

dB

d dBdB

PAPR

p

EE sym T RSNRNFN M f G G d

kTB

l

p

−

−

=

1E-16

1E-14

1E-12

1E-10

1E-08

1E-06

0.0001

0.01

1

0 10 20 30 40 50 60 70 80 90

QA

M B

it-e

rro

r R

ate

(B

ER

)SNR (dB)

4 QAM 16 QAM 64 QAM 256 QAM

10b/sym 12b/sym 14b/sym 16b/sym

18b/sym 20b/sym 22b/sym 24b/sym

0 0

S b b bd BW

N

P E f Esnr

P N B N= = =

0.01

0.1

1

10

100

1000

10000

100000

1000000

2 2.5 3 3.5 4

Ene

rgy

Effi

cacy

(b/J

) ; T

X P

ow

er

(W)

Propagation Constant

b/J, 750 MHz

b/J, 1900 MHz

b/J, 3600 MHz

TX Power, 3600 MHz

TX Power, 1900 MHz

TX Power, 750 MHz

0.001

0.01

0.1

1

10

100

1000

10000

100000

1000000

10000000

100000000

1E+09

10 100 1000 10000

Ene

rgy

Effi

cacy

(b/J

) ; T

X P

ow

er

(W)

Distance (m)

b/J, 750 MHz

b/J, 1900 MHz

b/J, 3600 MHz

TX Power, 3600 MHz

TX Power, 1900 MHz

TX Power, 750 MHz

p = 3.1

d = 1000m

LOS reflection

penetration

occlusion

19

© Eridan, 2019

Antenna Gain and Directivity

Signal power density: mW/cm2

More mW into the RX requires more antenna area (cm2)Independent of frequency!

Directivity (beamwidth) narrows as antenna size increasesMeasured in wavelengthsCorresponds to antenna “gain”Antenna must be aimed

Related by the Fourier Transform“Time domain” – Antenna aperture in wavelengths“Frequency domain” – Antenna directivity in radians

Directivity (gain) and coverage are in opposition

( ) ( )2

,

1 cos360

T RG G p

=

−

0

10

20

30

40

50

60

70

80

0.1 1 10 100

An

ten

na

Gai

n G

A(d

B)

Antenna Beamwidth (degrees)

20

© Eridan, 2019

Align Modulation to mmW Circuit Capabilities

• OFDM (3GPP 5G-NR)- requires circuit linearity

(output backoff)

- and has a large PAPR (~7-10 dB)

• pPSK- PAPR = 0dB

- No linearity required

• Output power increases

• PA efficiency increases 10-fold (or more)

• OpEx decreases by 10x

• Large decrease in power supply and heatsink sizes

0

5

10

15

20

25

-20 -15 -10 -5 0 5 10

Ou

tput

Pow

er (

dB

m);

Rati

omet

ric

Pow

er G

ain

(dB

); P

AE

(%)

Input Power (dBm)

Output Power (dBm)

Peak PAE

Ratiometric Power Gain

PAPR

PTX PEP PSAT

rms 1.000 [+2.5 dB]spacing 0.390 [-1.6 dB]

16-PSK

Output backoff

0

10

20

30

40

50

60

70

80

90

100

0% 10% 20% 30% 40%

Rat

io t

o O

utp

ut

Po

we

r

Circuit Energy Efficiency

Power Supply Size

Heatsink Size

pPSK

OFDM

21

© Eridan, 2019

mmW Safety – Human RF Exposure

IEEE-C95.1 calls for power density of 1 mW/cm2 as an upper limit when there is unrestricted

antenna access (e.g. a mobile device)

Published unit is often EIRP

Beamwidth contributions, more involved calculation

… are any present mmW handset products safe??

# o

f el

emen

ts p

er a

rray

sid

e

# o

f el

emen

ts p

er a

rray

sid

e

Total TX EIRP (far field)Total TX Power

4 antennas (2x2), 4 dBm each

22

© Eridan, 2019

Conclusions

• Dominant contributors to extremely low mmW efficiency

• Requirement for circuit linearity

• Signal PAPR greatly exceeding 3 dB

• mmW transistor operation is never as good as at sub-6 frequencies

• Transistor improvements will likely take a decade

• High efficiency modes familiar at sub-6 are not available (yet)

• Fast improvements require changing the mmW signal modulation

• One option is to adopt circular constellations

• Increases the useful output power from any existing PA

• Circuit linearity is not needed (recover the backoff)

• PAPR is 0 dB (recover PAPR power loss)

• Efficiency improves by at least 15x

• Network coverage improves dramatically

• No compromise on baseline bandwidth efficiency

23

© Eridan, 2019

Summary

There is hope!

But…

• Transistors today are not fast enough to access traditional

high-efficiency behaviors at mmWave frequencies

• Adopting a constant envelope signal with good bandwidth

efficiency (e.g. 16-pPSK) is doable now

• Will recover most of the efficiency loss

• Will greatly expand network coverage

• Requires action from 3GPP

• Operators can demand the performance they really need from

their suppliers – and from 3GPP