the importance of metrology in wireless communication systems · 10 network analyzer power and...

1

The importance of Metrology inWireless Communication Systemsy

From AM/FM to SDR Systems

Nuno Borges [email protected]

www.av.it.pt/nbcarvalhoUniversidade de Aveiro

© 2005, it - instituto de telecomunicações. Todos os direitos reservados.

Presentation Outline

1. Introduction

2 Typical Instrumentation2. Typical Instrumentation

3. The Role of Signal Excitation

4. Software Defined Radio Schemes

5. Some Real World Applications

2


2

1. Introduction

Wireless Systems, started with Tesla, Maxwell, Henry, Marconi, Armstrong, …With the born of these new technology, metrology for this type of systemsimmediately start to emerge, since the knowledge of physical properties of the

3

radio was fundamental for the correct implementation of those systems.

Metrology is defined by the International Bureau of Weights and Measures (BIPM) "the science of measurement, embracing both experimental and theoretical

determinations at any level of uncertainty in any field of science and technology."

1. Introduction

Wireless Communications has started a longtime ago, with the broadcast radio been one ofthe more massive implantation.the more massive implantation.At that time metrology for wireless (radio)applications, mainly addressed the transmittedpower.

– Transmitted power;– Signal Quality measured mainly using

4

human ear;

3

1. Introduction

Other specific measurements were needed forexample for radar, the pulse width, thereflection coefficient of the load, was also areflection coefficient of the load, was also avery important problem due mainly to the hugetransmitted power.

5

1. IntroductionWith the increase in radio transmissions, the spectrum start to be full, increasingthe need for extremely good metrology systems, that can support the income ofnational agencies.

6

With the increase number of systems, measurements as:

• RF Bandwidth• SNR, SIR, etc….

4

1. IntroductionWith the advent of mobilecommunications, the metrologystarts to be more demanding:

• Doppler Effects• SNR• Intermodulation

7

1. IntroductionNext step on wireless communications and thus consequently on metrology wasthe advent of digital communications.

• Error Vector Magnitude• Bit Error Rate• Intermodulation

8

5

1. IntroductionThis measurement procedure can also move to higher levels of abstraction usingQoS measurement systems.

Measure the complete QoSMeasure network parameters

9

1. Introduction

Nowadays metrology has a strongimpact on everyday communications,since it is fundamental and key to:

• Researchers• Communication Operators• Policy standard bodies• Non human devices• …

10

6


1. Introduction





11


2. Typical Instrumentation

Typical Instrumentation, spans from simple measurement of power to morecomplex schemes of complete demodulation of the signal and bit pattern evaluation.For instance some instruments are:

• Power Meters• Spectrum Analyzers• Network Analyzers• Time domain scopes, oscilloscopes• Noise Figure Meters• Vector Signal Analyzers• Protocol Analyzers

12

• Protocol Analyzers• …

7


13

VSA – Vector Signal Analyzer, SA – Spectrum Analyzer, VNA – Vector Network Analyzer, TG/SA –Tracking Generator, SA, SNA – Scalar NA, NF Mtr. – Noise Figure Meter, Imp. An. – Impedance Analyzer,Power Mtr. - Power Meter, Det. Scope – Diode Detector, Oscilloscope

Power can be achieved from:

∫=2/

2)(1)(T

dttvtp


∫− 2/

)()(T

dttvT

tp

Power can be considered in an window interval or in a period if the signal isperiodic .

Power Sensor

14

PowerMeter

ThermistorsThermocouplersDiode Detectors Substituted DC or low

frequency equivalent

Display

Net RF powerabsorbed by sensor

8

A diode detector can be represented by :2.5

x 104


( ) ( ) L++≈⎟⎟⎠

⎞⎜⎜⎝

⎛−= 2

211)( tvktvkeIty TVV

s

The output current will then be:

0 0.2 0.4 0.6 0.8 10

0.5

1

1.5

2

15The Importance of Radio Metrology for Health Impact Evaluation

⎠⎝

( ) ( )[ ] ( ) ( )[ ]12sin2

sinsinsin)(2

212

21 ++=+≈ tAktAktAktAkty ωωωω


The output DC current will thus be:

2A

16

2)( 2

AktyDC ≈

9

From a practical point of view the DC curve is a nonlinear function of thedevice, thus it should be calibrated accordingly.

x 10-3


0.4

0.6

0.8

1

1.2

1.4

squarehigher order

17

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10-3

0

0.2

Calibrationis the process of establishing the relationship between a measuring

device and the units of measure.


18

10

Network Analyzer

Power and FrequencyControl2. Typical Instrumentation

- S Parameters

- VSWR

- Gain (Amplitude and Phase)

- Bandwidth

- 1dB Compression Point

2.400 GHzdBm

DC

19

- AM/AM and AM/PMDC

DUTTest Signal – Sine Wave

Spectrum Analyzer

2.400 GHzdBm

DC + 15 0005 00


DC

DUT

+ 15.00- 05.00

- Gain (Amplitude)-IMD- Bandwidth

IP3

20

- IP3- AM/AM- ACPR-1dB Compression-NPR

Test Signal:• Sine Wave• Two-Tones• Multisines• Arbitrary Waveforms

11

Spectrum Analyzer2. Typical Instrumentation

21

from tektronix

Oscilloscope – for analog domain

2 400Oscilloscope


2.400- 00.53 + 15.00

22

Allows simultaneous measurements of the input and output signals

12

Logic AnalyzersLogic Analyzer


0.400+ 1GHz

ADC

23

Analysis of the bit-sequence in a digital system


1. Introduction





24


13


A nonlinear system do not follow the superposition theorem, thus theresponse to a certain input, can not be found from the superposition ofany other input’s.

|Y(ω)|Non Linear

Systemx(t) y(t)

ωω1

|X(ω)|

25

2ω1 ω3ω1ω1ωω1

A nonlinear system do not follow the superposition theorem, thus theresponse to a certain input, can not be found from the superposition ofany other input’s.


Non LinearSystem

x(t) y(t)ωω1

|X(ω)|

|X(ω)|

2ω1 ω3ω1ω1

|Y(ω)|

|Y(ω)|

ωω1

|X(ω)|

ω2

26

ωω2

| ( )|

ω2 2ω1 ω3ω1

|Y(ω)|ωω1 ω2ω2-ω1

2ω1-ω2 2ω2-ω1

2ω1 2ω2

ω2+ω1

3ω1 3ω2

2ω1+ω2 2ω2+ω1

1 2

14

The best test signal to use in a nonlinear system is thus the realtelecommunications signals from where the system will be excited.


27

-10

0

Similar PSDSi il I t t d P

1. Introduction

-70

-60

-50

-40

-30

-20

Out

put P

ower

[dB

m]

Similar Integrated Power

Different Spectral Regrowth???

AWGN Out

W-CDMA Out

28Test Signals for Nonlinear Distortion Evaluation

-90

-80

-10000 -7500 -5000 -2500 0 2500 5000 7500 10000

Frequency-900MHz [KHz]

AWGN In

AWGN OutW-CDMA In

15

What should be the excitation used for wireless system characterization ?


• One single sinusoid (One tone ?)

• Two signal sinusoids (Two tones ?)

• Gaussian Noise ?

• Multisine Signals ?

Other t pe of signals ?

29

• Other type of signals ?

One single sinusoid normally called one tone

( )tA ωcos


|Y(ω)|

( )tA ωcos

30

2ω1 ω3ω1ω1

16

In Linear systems a one tone is enough for its characterization)()()( ωφωω jjejHjH =


0.6

0.7

0.8

0.9

1

tude

|H(jω)| ( )

0

50

100

150180

se

H(jω) (º)

Linear System

H(jω)x(t) y(t)

31

4 5 6 7 8 9 10 11 120

0.1

0.2

0.3

0.4

0.5

Frequencyf (GHz)

Am

plit

-180-150

-100

-50

0

4 5 6 7 8 9 10 11 12Frequency

f (GHz)

Phas

In nonlinear systems the same approach can be followed for a constant inputpower, changing the input power implies that the transfer function will also changewith input amplitude, transforming itself into the Sinusoidal Input DescribingFunction

)( jAH


),( ωjAHWhich is now dependent both on the frequency and on the input power.

3º

4º

5º

φ (º)

e

1 dB

GP (dB)

r Gai

n

4

6

8

10

AM/PMAM/AM

32

-1º-15 -10 -5 0 5 10 15 20 25

0

1º

2º

Input PowerPin (dBm)

Phas

e

Pin (dBm)

Pin1dB

Pow

er

-15-2

0

2

4

-10 -5 0 5 10 15 20 25Input Power

17

But the previous measurement only takes into account the change of the carrierwith the input signal power and frequency.Nevertheless the nonlinear system will generate harmonic distortion, that can beaccounted for by:


2ω 3ωω

|Y(ω)|

yTotal Harmonic Distortion - THD

33

2ω1 ω3ω1ω1

( ) ( )[ ]

( ) ( )( )[ ]∫

∫ ∑

+

⎥⎥⎦

⎤

⎢⎢⎣

⎡+

=

∞

=T

ioio

T

riroiro

dtAtAAT

dtAtrAAT

THD

0

211

0

2

2

,cos,1

,cos,1

ωφωω

ωφωω

This type of measurement only takes into account the change of the carrier with theinput signal power and frequency.


|Y(ω)|

34

2ω1 ω3ω1ω1No base band excitation No in-band spectrum regrowth

18

In summary a one tone test can:

Measure the gain compression and the phase deviation (Sinusoidal


Measure the gain compression and the phase deviation (SinusoidalDescribing Function)

Measure the degree of harmonic distortion

Can not:

P i b d l h

35

Present any in-band spectral regrowthExcite base band components

)cos()cos()( 2211 tAtAtx ii ωω +=

And the output will be:

In this case the input signal consists in:3. The Role of Signal Excitation

And the output will be:

( ) Ζ∈+=+= ∑∞

= , and wherecos)( 21

1nmnmtAty r

rrorroNL ωωωφω

This will allow to account for the same goals as the one tone as well as for other important in-band studies.

|Y(ω)|

36

ωω1 ω2ω2-ω1

2ω1-ω2 2ω2-ω1

2ω1 2ω2

ω2+ω1

3ω1 3ω2

2ω1+ω2 2ω2+ω1

19

Pout(dBm)

Power sweeps are very important in a two tone test, since they allow us togather important IMD information, specially for small signal, as the wellknown IP3 or IPn.


-10 20

-100

-80

-60

-40

-20

0

20

40

PFund(ω2)1dB/dB

3dB/dB

out( )

Pin(dBm)0

PIMD(2ω2-ω1)

IP3i10 30

IP3

Out

put P

ower

-10 20

-60

-40

-20

0

20

40

PFund(ω2)1dB/dB

2dB/dB

Pout(dBm)

0

P(ω2-ω1)

IP2i

10 30

IP2

Pin(dBm

Out

put P

ower

37

Nevertheless only a small step was done in the process of real signalapproximation.

( )( )

( )( )12

221

122 ωω

ωωω

ω−

=−

=≡P

PP

PPP

IMRIMD

fund

Only a single frequency at the base band is excited for each separation oftones. If the system is dynamic, then it can be studied by doing a sweep in toneseparation, but again we hope that the superposition theorem works….


|Y(ω)|

38

ωω1 ω2ω2-ω1

2ω1-ω2 2ω2-ω1

2ω1 2ω2

ω2+ω1

3ω1 3ω2

2ω1+ω2 2ω2+ω1Only a single frequency, or harmonics

of that single frequency, are excited

20

In summary a two tone test can:

Beyond the one tone informationMeasure the two tone input describing function


Measure the two tone input describing functionObtain in-band distortion informationExcite the long term memory effects at a predetermined frequency

Can not:

Fully excite the long term memory effectsP diff k l l

39

Present different peak envelope values

( )∑=

+=Q

qqqq tAtx

1cos)( θω

A very popular test for nonlinear dynamic systems is the multisine approximation.

ωωω Δqq )1(0 −+=e.g. for


=q 1

In this case not only the amplitude and the separation of tones is important as also thephase of each tone can generate a completely different signal.

mpl

itude

mpl

itude

10

8

6

4

2

0

2

10

8

6

4

2

0

40

Am

time

Am

time

-2

-4

-6

-8

-10

-2

-4

-6

-8

-100 500 1000 1500 2000 2500 3000 3500 4000

0 500 1000 1500 2000 2500 3000 3500 4000

10 Tones of Randomized Phase 10 Tones of Equal Phase

21

In this case the output will be a combination of all the input tones:

( )∑=

+=R

rrrr tAty

1

cos)( φω


where MaxQqQQqqr Ordermmmmmm ≤++++++++= −−− 101100 , LLLL ωωωω

plitu

de [U

]

A typical output signal is:

41

BaseBan

d ω RF

Δω

ω2ω RF

3ω RF

Am

p

The in-band signal will allow us to evaluate the spectral regrowth content and thus study the Adjacent Channel Power Ratio, ACPR, or the Noise Power Ratio, NPR.


ut P

ower

[dB

m]

input spectrum

output spectrum

42

ω

Out

pu

L U

ω0 ωU2ωU1ωL2ωL1

22

In the base band a carefully selection of tone spacing should be done in order to excite the most important characteristics of the system.


Band ω RF ω

2ω RF3ω RF

Am

plitu

de [U

]

43

BaseBa ω

Δω

2 3

In summary a multisine test can:

Beyond the two tone information


Excite the long term memory effects

We should be careful in:

Defining the number of tones and tone separationDefining the relative phase between tones

44

23

Despite those frequency discrete excitation, continuous signals were also used,for instance the white Gaussian noise. First attempts of formal nonlineardynamic system identification led to some important conclusions:

i h hi i i i h h ll b h i l i f i


2 – The non-periodicity of noise stimuli pose several problems on the use ofmodern digital signal processing techniques. Thus, periodic signals arepreferred.

1 – A test with white Gaussian noise is enough to gather all behavioral informationnecessary to extract a full nonlinear dynamic model of fading memory[Schetzen, 1980].

3 – Fortunately it was already shown that the response to noise signals can bei i d b l d i d li i f i di l i i

45

imitated by several randomized realizations of a periodic multisine[Schouckens & Pintelon, 2001].

4 – Several authors are now questioning the validity of the Gaussian excitation fordevice testing [Apparin, 2001].

Probability Density, pdfx(x), as a Weighting Function

Since typical laboratory data like Output Power, Power Spectrum, etc., isaveraged in nature:


ve ged u e:

{ } ∫∞

∞−== dxxpdfxsxEP xin )()( 22

{ } ∫∫∞

∞−

∞

∞−=== dxxpdfxfdyypdfysyEP xNLyout )()()()( 222

46

it is intuitive to expect that, more important than the trajectory of amplitudevalues assumed by the excitation, x(t), should be the Probability with whicheach value is reached, i.e., the excitation pdfx(x) or ccdfx(x).

24

That conclusion led several engineers to abandon the Peak-to-Average Ratioas a faithful excitation metric for predicting output system distortion;

Probability Density, pdfx(x), as a Weighting Function


as a faithful excitation metric for predicting output system distortion;

and led manufacturers of waveform generators to give information of thesignal complementary cumulative distribution function, ccdfx(x).

47

CCDF PLOT

To prove this role of the pdfx(x) we tested a nonlinear system with threesignals of equal integrated power but distinct amplitude distribution:

Multisines for Memoryless Nonlinear Systems


0.03

0.04

0.05

0.06

0.07

Gaussian

Uniform

roba

bilit

y D

ensi

ty

48

-30 -20 -10 0 10 20 300

0.01

0.02Two-Tone

Amplitude

Pr

25


Different modulated signals impose different statistical behavior, and thus should be dealt accordingly.

10-4

10-3

10-2

10-1

100

Pro

babi

lity

[%]

Single Mode Wi-FiSingle Mode WiMAXWi-Fi + WiMAXSingle Mode W-CDMAW-CDMA + 4xGSM1800

4

6

8

10

12

Pro

babi

lity

[%]

Single Mode Wi-FiSingle Mode WiMAXWi-Fi + WiMAXSingle Mode W-CDMAW-CDMA + 4xGSM1800

49

0 2 4 6 8 10 1210-6

10-5

PAPR [dB]

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 10

2

Amplitude [U]


1. Introduction





50


26


Software Defined RadioSoftware Defined RadioImplement radio functionality as software

modules running on a general purpose basic hardware;

Adaptable and Reconfigurable;Moving as many of the radio functions from

the RF transceiver to the baseband digital chip as possible will improve the radio performance, cut the overall power and,

51

J. Mitola, “The software radio architecture,” IEEE Communications Magazine, vol. 33, nº 5, pp. 26–38, May 1995

performance, cut the overall power and, most importantly, allow reconfigurability of the radio designs for multi-band and multi-standard coexistence.

From a Physical point of view the software radio concept ismoving from the simple base band technological approach to the


real RF signal evaluation.

Antenna

BandpassFilter

RF IF Baseband

ADC/DACDSP

52

VariableFrequencyOscillator

LocalOscillator

Filter DSP

27




Antenna RF IF Baseband

ADC/DACDSP

All Digital

53

VariableFrequencyOscillator

DSP




Antenna RF IF Baseband

ADC/DACDSP

All DigitalAll Digital

54

DSP

28

SDR approaches imply the use of ADC’s and DAC’s for theconversion between domain modes.

So analog and digital figures of merit should be measured and


So analog and digital figures of merit should be measured andco-exist in this type of SDR approaches.

For instance in a receiver we have:Antenna

ANALOG DIGITAL

55

ADC/DACDSP

The Analog Part includes an Amplifierand sometimes a tuned filter.

Thus Figures of merit like:

Antenna

ANALOG

ADC/DACDSP

DIGITAL


Thus Figures of merit like:

Noise Figure

Bandwidth

Input VSWR

Gain16dB15dB

14dB

1.1dB

1.6dB

2.1dB

2.6dB

3.1dB

ISC

56

IMD, IP314dB

13dB

29

Converters like ADC’s and DAC’simpose different types of Figures of

Antenna

ANALOG

ADC/DACDSP

DIGITAL


merit like:

Static Gain and Offset

Integral and Differential Nonlinearity,Monotonicity

Hysteresis, Harmonic and Spuriousi i l i i i

57

Distortions, Total Harmonic Distortion

SFDR, SINAD, IMD

Most of these figures of merit can be measured using combinations of analogand digital instrumentation such as:

- Vector Signal Analyzers (either Analog and Digital)- Oscilloscopes


- Logic AnalyzersNevertheless these instruments are stand alone, and many of them allow powermeasurements but not phase.

900

ase

Ban

dss

ing

Uni

t

AGCControl

PowerDetector

... ...101011

58

Rx

- B P

roce

s

... ...110101

ANALOG DIGITAL

30

Most of these figures can be measured using combinations of analog and digital instrumentation


900

ase

Ban

dss

ing

Uni

t

AGCControl

PowerDetector

... ...101011

( ) reftntn VNi .22 )()( +⋅⋅⋅+

59

Rx

- B P

roce

s

... ...110101

ANALOG DIGITAL

The mixed mode signal analyzer is a viable solution for thecomplete analysis of SDR solutions.


60

Pedro Cruz, Nuno Borges Carvalho, Kate A. Remley and Kevin G. Gard, “Mixed Analog-Digital Instrumentation for Software Defined Radio Characterization”, IMS 2008

31

Synchronous input and output samplers:

• One in the analog domain

Oth t th di it l d i


• Other at the digital domain

61

Usual figures of merit could continue to bemeasured

)(ωV


)()()(

)( 11 ωωω

ωρ Dinc

ref SVV

==

∫ −+⋅⋅⋅+2/

)()( .].22[1)(

Ttj

reftntn dteV

TVNi ω

ω

62

∫−

−

−== 2/

2/

2/

).(1)()(

)( T

T

tjinc

T

inc

digL

dtetvT

TVV

Hωω

ωω

32

Using typical laboratory measurement instruments: 10MHz


0.400

Logic Analyzer

Oscilloscope

900

Rx -

Base

Ban

d P

roce

ssin

g U

nit

A GCC o nt rol

P ow erD e te c tor

... ...101011

... ...110101

Trigger

Trig

ger

63

+ 1GHz

Control ComputerGPIBGPIB


1. Introduction





64


33

65

First test: Sine wave excitation

Measure the amplitude gain and phase over frequency

Measure the AM/AM and AM/PM by characterizing the amplitude


-0.4

-0.2

0

dB]

250

300

350

e [º

] -1.5

-1

-0.5

0

[dB

] 200

250

300

350

e [º

]

Measure the AM/AM and AM/PM by characterizing the amplitudegain and phase over input power sweep

66

200 300 400 500 600 700 800 900 1000-1

-0.8

-0.6

Gai

n [

Input Frequency (KHz)200 300 400 500 600 700 800 900 1000

100

150

200

Ang

le

-10 -5 0 5 10 15

-3

-2.5

-2Gai

n

Input Power (dBm)-10 -5 0 5 10 15

50

100

150 Ang

le

34

Second test: Two-tone excitation


67

68

35

Interference at airport frequency: 118.1 MHz

Lisbon airport reports:Control tower frequency: 118.1 MHz highly interfered !

118.1

69

Co t o to e eque cy 8 g y te e edAirways controllers hear a broadcast radio station on the

airport-plane audio frequency.That frequency channel becomes unavailable!For security reasons other frequency channel starts to be

used..

What can be done?

70

36

Possible problems

Interference possible sourcespComplex analysis

– Huge amount of licensed radio communications services– Huge amount of free radio services (Wi-Fi, Bluetooth, RFID’s, ...)– Illegal use of spectrum– Other sources (microwave ovens, TV amplifiers that become

unstable, ...)

71

We should measure:– A correct characterization of the bandwidth, power, should be

measured– Obtain the maximum information about the interfered signal, what

happens as alarm reports, ‘drop calls’, etc.

Measurement procedure

Spectrum MonitoringSpectrum MonitoringInterference Characterization

–– Occupied BandwidthOccupied Bandwidth–– Interference DurationInterference Duration (permanent, intermittent, random, ...)–– Occurred periodOccurred period (weather conditions, temperature, humidity, rain,

etc.)

72

–– PowerPower (allow us to predict the distance to the source)–– Signal Demodulation Signal Demodulation (viable in analogue systems)

37

Airport problem

We already know much about the interference...Is permanent (allows the location more efficiently)

In the demodulation procedure, we can conclude that it is an analoguesignal, and a radio broadcast station, which we will call “A“A RADIO”RADIO”,which is licensed to broadcast at frequency channel 104104..33 MHzMHz

Since no transmission is been done at that channel fromth di t i f N liN li

73

the radio operator, we are in presence of a NonlinearNonlinearDistortionDistortion ProblemProblem, that is a signal that is generated ina nonlinear device, using the 104.3MHz channel as oneof the mixed signals.

Intermodulation Interference

Available information from the measurements:“A RADIO” – 104.3 MHz (identified due to the interfered signal demodulation)“B RADIO” - ? (unknown)

. Solution: Use “A Radio” channel frequency as a un modulated carrier.

. Try to demodulate the interference channel, where we can hear “B RADIO”, which is identified as the one that transmits at 90.4 MHz (and is installed in the same tower in adjacent antennas).

74

Using a third order mixing product:

2 x 104.3 2 x 104.3 –– 90.4 = 118.2 !!90.4 = 118.2 !!

38

Intermodulation Interference

75

Cause: Passive Intermodulation (PIM)

Source of the problem:Inferior Radiation Structure

Solution:Solution:Change the antenna due to oxidation

76

39

Pedro CruzJosé Pedro Borrego

5. Acknowledgements

ANACOM

77

the importance of metrology in wireless communication systems · 10 network analyzer power and...

Documents