Vous êtes sur la page 1sur 38

RF IV Waveform Measurement and Engineering

- CW Measurement System Realization -

Centre for High Frequency


Engineering
School of Engineering
Cardiff University

Contact information
Prof. Paul J Tasker
tasker@cf.ac.uk

IEEE MTT-S Distinguished


Microwave Lecturer
2008-2010

website: www.engin.cf.ac.uk/chfe

History of RF I-V Measurements

- Development of the Non-Linear Network Analyzer


Historically has had many names;

NLVNA: Non-Linear Vector Network Analyser


LSNA: Large Signal Network Analyser
ANA: Absolute Network Analyser
Vector(ial) Component Analyser
Era of the MTA (Microwave Transition Analyser)

Kompa et al (1990)
Tasker et al (1994)
Verspecht et al

Return of the DSO

Tektronix DSA
Williams et al

1990

2010

Waveform Measurement
1980

2000

First realization of calibrated


waveform measurement solutions

Time Domain Sipila et al (1988)


Frequency Domain: Lott U (1989)

First wave of Commericalization (LNSA)

Agilent & Maury Microwave Corporation

History of RF I-V Measurements

- Era of commercialization and industrial acceptance


Second Wave of Commercialization

Agilent: PNA-X
NMDG/Rohde & Schwarz
VTD (Verspecht-Teyssier-DeGroot)
Mesuro/Tektronix

Era of the MTA (Microwave Transition Analyser)

Kompa et al (1990)
Tasker et al (1994)

Agilent PNA-X: Frequency Domain

Return of the DSO

1990

Tektronix DSA

2010

Waveform Measurements
1980

Mesuro/Tektronix: Time Domain

2000

First Realization of calibrated


waveform measurements

Time Domain Sipila et al (1988)


Frequency Domain: Lott U (1989)

Key Parallel
Development

First wave of Commercialization (LNSA)

Agilent & Maury Microwave Corporation

Waveform Engineering

Objective of RF I-V Measurement Systems


- has to enable Waveform Engineering in Design

Their measurement domain is to go beyond s-parameters

RF I(I) & V(t) Waveform Measurement

Their application domain is to go beyond linear design

RF I(I) & V(t) Waveform Engineering


4

Outline:

- CW Measurement System Realization

RF I-V Measurement Solution

Architecture and Receivers


Error Models and Calibration

RF I-V Engineering Solutions

Active Open Loop Architecture


ELP Concept
Frequency
Source(s)
Two-Port Microwave
Test-set
Four Channel
Receiver

Non-Linear Vector Network Analyzer


- Basic Architecture with RF Test-set

Time domain variant requires a four channel


receiver with each channel receiving either
incident or scattered travelling voltage
signals.

Frequency domain, PNA-X, variant requires a


five channel receiver and a reference signal.

Utilized directional couplers for


detection/separation of travelling voltage
signals.

Source switch for redirection of stimulus


signal.

Microwave
Source

Alternatively utilize two sources, PNA-X or


Tektronix AWG.

4-channel Microwave Sampler

a0(t)

b0(t)

b3(t)

Couplers

a3(t)

Couplers

Bias T

Bias T
Port 1 Port 2

All instruments and components are


computer controlled allowing for automated
measurements

Input
DC

a1(t)

b2(t)

b1(t)

a2(t)

Input
DC

Measures RF an(t) and bn(t) time varying Voltage


Travelling Signal Waveforms
6

Time Domain Systems:


- Sampling Receivers

Key component is a broadband receiver

Time domain sampling based

Tektronix: Digital Serial Analyzer (DSA)

Agilent: Microwave Transition Analyzer (MTA)

Principle is based on sampling over many cycles


|A|

2fo

2fo
fo
fo

Signals must be repetitive and on a specific frequency grid

Samples RF Voltage Waveforms vn(t)

Non-Linear Vector Network Analyzer:


- Sampling based Architecture

Measurement architecture is almost


identical to conventional Network
Analyzer

RF hardware between DUT and the


sampling receivers.

4-channel Microwave Sampler

Introduces Systematic Errors

Measured an(t) and bn(t) time varying


voltage travelling signal waveforms will
be erroneous.

Microwave
Source

Error Correction Model


Vector Calibration

a0()

b0()

b3()

Couplers

a3()

Couplers

Bias T

Bias T
Port 1 Port 2

Input
DC

a1(t)

b2(t)

b1(t)

a2(t)

Input
DC

Measurements System needs to be vector calibrated


8

Non-Linear Vector Network Analyzer:


- Error Model

Error Correction Flow Graph

8 Term Error Model

Similar to that utilized by VNA

All terms required

Independent of switch match

no transformation to a
reference impedance

a0

10

Independent of direction of
energy flow

b2

b0

01

b1

a2

b3

32

11 DUT 22

00

a1

33
23

a3

Simple de-embedding algorithm

b1 = (b0-00a0)/01
a1 =((0110-0011)a0+11b0)/01
b2 = (b3-33a3)/32
a2 =((3223-3322)a3+22b3)/32

Require a calibration procedure: going beyond s-parameters


9

Ratio Calibration:

- VNA 10 Term Error Model


Follow VNA Procedure

Microwave
Source

Determine 10 (12) Term Error Model

Load, Open, Short, Thru (LOST)


Thru, Reflect, Line (TRL)
Thru, Reflect, Match (TRM)

Forward Error Model

a0

4-channel Microwave Sampler

b0()

b3()

Couplers

a3()

Couplers

Bias T

Bias T

b0

1001

Input
DC

short

open

open

load

load
thru

Input
DC

b1

Reverse Error Model

a1

Port 1 Port 2

short

b2

1032 b
3

11 DUT 22

00
a0()

a1

a2
b2

2332

11 DUT 22
b0

2301

b1

a2

b3
33

a3

. measure s-parameters

10

Ratio Calibration:

- VNA 10 Term Error Model

Microwave
Source

Equivalent to VNA

Ratio measurements as a function of


frequency
S21 Radius = 9
135

4-channel Microwave Sampler

a0()

b0()

b3()

Couplers

a3()

S12 Radius = 0.2


45

180

Couplers

Bias T

Bias T
Port 1 Port 2

Input
DC

90

a1()

b2()

b1()

a2()

-j250

-j10
-j100

-j25
Input
DC

Measured

-j50

Modelled

. measure s-parameters

11

Relating VNA and NLVNA Error Models


- step 1

Forward Error Model

a0

11

00
b0

a1 b 2

1001

11
2301

b0

b 1 a2
a1 b 2

2332

b3
33

22

b 1 a2

a0

a1 b 2
11

00

22

Reverse Error Model

b0

1032 b3

a3

1001
a0

0
b0

00
b0

10/23

b3

33

22

b 1 a2

b3

1032

23/10 a3
a1 b 2
11

2332

33

22

2301 b1 a2

b3

a3

Reformulate Error Model: Isolate correction and Impedance transformation


12

Relating VNA and NLVNA Error Models


- step 1

Transformation of Error Model

Utilize measurement of b3/a3 (3)


during forward THRU calibration
Utilize measurement of b0/a0 (0)
during reverse THRU calibration

a0

a1 b 2
11

00
b0

1001

b3

33

22

b 1 a2

b3

1032

23/10 a3

Mathematical Conversion

a0

Forward Model
22 = 22+1032/(1-33.3)
1032= 1032/(1-33.3)

Reverse Model

11 = 11+0110/(1-00.0)
0123= 0123/(1-00.0)

0
b0

00
b0

10/23

a1 b 2
11

2332

33

22

2301 b1 a2

b3

a3

Reformulate Error Model: Isolate correction and Impedance transformation


13

Relating VNA and NLVNA Error Models


- step 2

a0

a1 b 2
11

00
b0

1001

33

22

b 1 a2

b3

b3

1032

23/10

Determine
Missing Terms

a0

a3

10

11

00
a0
0
b0

00
b0

10/23

a1 b 2
11

2332

33

22

2301 b1 a2

b3

b0

a1 b 2

01

33

22

b 1 a2

b3

32

23

a3

23

a3
Un-normalize and Combined Error Model
14

Absolute Calibration:
- Determine 10 or 23

Microwave
Source

Additional calibration steps

10 un-normalization

4-channel Microwave Sampler

23 un-normalization

a0()

b0()

b3()

Couplers

a3()

Bias T

MAG:
Attach a power meter to Port 2
PHASE: Attach a phase meter to Port 2
PHASE: Inject a known signal into Port 2

Phase Meter !!!!!!!!

Requires the utilization one of the samplers

Bias T
Port 1 Port 2

Input
DC

Couplers

MAG:
Attach a power meter to Port 1
PHASE: Attach a phase meter to Port 1
PHASE: Inject a known signal into Port 1

Power meter
Phase meter

Phase signal

Input
DC

Phase Signal !!!!!!!

Requires a spectrally rich signal with a known


phase relationship

. go beyond s-parameters
15

Absolute Calibration:

- NLVNA 8 Term Error Model


a1

a0

b0(t)

b3(t)

Couplers

a3(t)

Couplers

Bias T

Bias T
Port 1 Port 2

Input
DC

10

b0

01

a1(t)
b1(t)

a2(t)

Input
DC

b2

b1

a2

b3

32

33
23

a3

Simple de-embedding algorithm

b2(t)

a1

a2

11 DUT 22

00

4-channel Microwave Sampler

a0(t)

DUT

b1

Microwave
Source

b2

b1 = (b0-00a0)/01
a1 =((0110-0011)a0+11b0)/01
b2 = (b3-33a3)/32
a2 =((3223-3322)a3+22b3)/32

Waveform Measurements
16

RF I-V Waveform Measurement System


- Review of Fundamental Architecture
RF test set to
separate incident
and reflected
voltage traveling
waves

RF Source for both


Calibration and Measurement

Microwave Transition
analyzer
Agilent

Receiver to
measured voltage
traveling waves
Key Component

Frequencies up to 67 GHz
Power levels up to 100 Watts

17

NLVNA Goes Beyond S-parameter:

10

0
-2

-4

-5

-6

-10

-8
0

180 360 540 720


Phase []

Measures magnitude and phase of all


the frequency components present in
the terminal travelling waveforms

HFET Transistor
Power Sweep @ 1.8 GHz

16

120

12

80

40

0
0

180 360 540 720


Phase []

Drain Voltage V[mA]


ds

160

Power response
Spectral distortion

Data Transformations

Frequency to Time Domain

Drain CurrentdI[mA]

Gate Voltage V[mA]


gs

Gate Current I[mA]


d

- Waveform Measurement

a1
b1

Waveforms

a & b waves into v & i waves

DUT

b2
a2

18

NLVNA Waveform Measurement:

Output power [dBm]

- Performance Extraction
30

20

20

0
-20

f0
2f0
3f0

10
0

-40

-60

2 4 6 8 10 12 14 16 18 20
Input Power [dBm]

HFET Transistor
Power Sweep @ 1.8 GHz
Phase [deg]

180
135
90
45
0
-45
-90

Data Transformation. Non-Linear


Performance Evaluation

Gain and Gain Compression AM-AM


Output Power
Phase Response AM-PM
Spectral growth

Direct Observation

174

Mode of Operation
Breakdown/Reliability

172
170

f0
2f0
3f0

168
166

2 4 6 8 10 12 14 16 18 20
Input Power [dBm]

a1
b1

DUT

b2
a2

Are Waveform Measurements Sufficient?

19

Linear versus Non-Linear Circuit Design


- the need for waveform engineering

Linear System

characterized by s-parameters

allow impedance transformation


cascading of networks

S-parameters are
also a design tool in
Linear CAD

Non-Linear System

characterized by waveforms

includes spectra growth (harmonics & inter-modulation)

cannot perform impedance transformation

Performance is influenced by measurement environment

Need to Engineer as well as Measure Waveforms


20

NLVNA needs a waveform engineering extension to


become a productive design tool

Non-Linear Vector Network Analyzer Limitations

Valid calibration despite changing


impedance states

Passive Impedance Variation

Microwave
Source

Waveform Engineering Extension

Determines non-linear behaviour only into its fixed nominal 50 ohms impedance environment
Circuit design requires knowledge of on-linear behaviours into an arbitrary impedance environment

4-channel Microwave Sampler

Manual stub tuners


Automated source- and load-pull
systems

Active Impedance Variation

Requires multiple microwave


sources.
Allows for compensation for losses

Couplers

Couplers

Bias T

Bias T
Port 1 Port 2

Input
DC

Input
DC

21

Engineering the Stimulus Voltage Waveform


- The concept of open-loop active load-pull

load-pull requirement: Modify Reflected Travelling Wave

Active System

L (ZL)

adjust
Stub
length

b
L (ZL)

Passive System
Performance limited by losses in
measurement system

Closed loop stability issues

Matched
termination

a2

adjust
Stub
position

b2

Amplify signal to overcome losses

Couplers, bias-tees, fixture

b2

circulator

a2

Inject signal a2

absorb signal b2
Matched
termination
Phase locked
signal generator

22

Engineering the Stimulus Voltage Waveform


- multi-harmonic open-loop active load-pull

fo

Multiple RF Sources

fo

Agilent
2.fo

2.fo

3.fo

3.fo

Active Harmonic Source-Pull

Microwave
Sources

Active Harmonic Load-Pull

Frequency domain
Arbitrary Waveform
Generator
Tektronix

Time domain

Digital World reaches RF

23

RF I-V Waveform Measurement & Engineering System


- Review Fundamental Architecture
fo

2.fo

RF Waveform
Engineering
Stimulus

3.fo
Active Harmonic Source-Pull

RF
arbitrary
fo
Multiple
RF Sources
Stimulus to engineer
waveform
to engineergenerator
voltage
voltage traveling
2.fo
to
engineer
voltage
traveling
waves
waves
traveling
waves
Key Component
3.fo
Key Component
Key Component
Active Harmonic Load-Pull

RF Waveform
Measurement
Receiver

Receiver to
measured voltage
traveling waves
Key Component

Frequencies
up toat67IMS
GHz
Demonstrated
Power 2008
levelswith
up to
100 Watts
Tektronix

24

RF I-V Waveform Measurement & Engineering System


- Emerging Commercial Architectures

Time Domain Based

Demonstrated at IMS
2009 with Tektronix

25

Further Considerations and Developments


- higher power and/or higher thru-put

Packaged Devices

High Power Devices

Requirement for waveform de-embedding

Requirement for impedance transformation

High Thru-put
Requirement for closed loop active load-pull

26

De-embedding Requirements:
- Packaged 20W Si LDMOS Device
Extrinsic Waveforms: Meaningless

Intrinsic Waveforms: Design Aid

10

-1

0.5 1.0 1.5 2.0 2.5


Time (ns)

Intrinsic
Generator
I
Plane

VVint
m
e
a
s

g
e
n
p
l
a
n
e

Iint
I

c
u
r
g
e
n

Extrinsic
Measurement
Plane

Intrinsic
Device
D
e
Plane
v
i
c
e
p
l
a
n
Ce
d
s

b
o
n
d

f
l
a
n
g
e

Iext
I

1500

500

f
l
n
a
g
e
2

f
l
a
n
g
e
1

V
Vext
m
e
a
s

3
2
1
0
0

0.5 1.0 1.5 2.0 2.5


Time (ns)
5W Si LDMOS

5000
4000

1000

m
e
a
s

Package & Mounting

Voltage (V)

20

Current (mA)

30

Current (mA)

Voltage (V)

Current (A)

40

4
Current (A)

60
50
40
30
20
10
0

50

Before de-embedding
After de-embedding

3000
2000
1000
0

-500

0 -1000

10

10

20

30
40
20
30
Voltage (V)
Voltage (V)

40

50

50

60

60

27

System Impedance Issue:

- Band Limited Waveform Engineering

50
30

0.5

1.0 1.5
Time (ns)

Engineered
Voltage Waveform

Voltage (V)

Current Waveform
0
-10
-20
-30
-40
-50
-60

10

4
5
Freq (GHz)

-10
-30

60
50
40
30
20
10
0
0.0

0 1 2 3 4 5 6 7 8 9
Freq (GHz)

High Power Characterization Environment

9x10

Voltage (dBmV)

0.0

Current (dBmA)

Voltage (dBmV)

60
50
40
30
20
10
0
2.0 2.5 Engineered

4
3
2
1
0.5

1.0 1.5
Time (ns)

2.0 2.5

Current (A)

12
10
8
6
4
2
0

Ropt << Zo System Impedance


Current (mA)

Voltage (V)

Ropt >> Zo System Impedance

Engineered
Voltage Waveform

60
50
40
30
20
10
0
0 1 2 3 4 5 6 7 8 9
Freq (GHz)

28

Waveform Engineering Issues


- Active load pull at high powers?

High ! = PLP PGen requires both values to be almost equal


High dissipated power PDis = PGen ! PLP requires a difference
Both requirements can be only satisfied by a rise of PLP and PGen

Example:
Load-pull of a 100 Watt
device with 1 output impedance
in a 50 system results in the
following signal levels:

PLP = 1.2 kW PGen = 1.3 kW


VSWR = 50 VMax = 1.4 kV
Prohibitive!
29

High Power Waveform Engineering:

- Solution is a low impedance measurement system

Build a low-impedance measurement system

This is the preferable option but is impractical

Use of broad impedance transformers

Significantly reduced VSWR


Maintain integrity of waveforms
Resonance free environment
over large bandwidth
Can employ well established
TRL calibration techniques.

30

High Power Waveform Engineering:

- Solution is a low impedance measurement system

31

High Power Waveform Engineering:


- Measurement of a Freescale 100 W LDMOS Device

Equivalent
sweep area in a
50 System

Gain

Pout

Efficiency

16

60

14

50

12
40

10

Gain

Impedance
sweep area in a
7.15 System

30

20

4
10

2
0

0
20

25

30

35

Drain Efficiency (%), Pout(dBm)

Z0=7.15

Only 120 Watts required to


probe the optimum load
when using 50 to 7.15
impedance transformer

40

Pin (dBm)

32

High Power Waveform Engineering:

- Critical High power measurement set-up components


High Power
Bias-tee

100 Watt LDMOS


Transistor

Klopfenstein Impedance
Transformers

High Power
Test Fixture

33

Alternative Active Load Pull Solutions


- address the issue of measurement thru-put

A traditional passive load-pull system can:

Passive Load-Pull Technique

Set loads independent of power output of DUT

Impedance Controlling
Element

Input

Have stability & measurement artefacts due to


broadband impedance variations

Output

b2
a2

DUT

Power
Meter
Phase Shifter

_ load =

Tuner

a2

Can we take the advantages


of both of these load setting
techniques?
The active open-loop
load-pull approach:

Take a considerable amount of time to characterise


and calibrate at different frequencies

b2

Active Open-Loop Load-Pull


System
Active Load Controlling
Input

Present a complex challenge to independently set


harmonics

Output

b2
DUT

a2

Source

Load =

a2
b2

Circulator

50_

Element

Sets harmonic impedances independent of each


other
Offers unconditional stability
Does not require calibration or precharacterisation
May take considerable time to iterate to each load

34

Alternative Active Load Pull Solutions


- developed Envelop Load-Pull System

Single ELP Loop


Multi-Harmonic ELP System
Attenuator

50_

DUT

b2 Directional Coupler
DUT

b2

a2

Circulator

ELP Module
Directional

Coupler
50_

b2

ELP

50_

PC

_ load

a
= 2
b2

a
=2

a2
b2

DAC a
2

Loop Amplifier
Harmonic

Load - Pull

System

b2

Demodulator Control

RF
Coupler

a2

F0

PC

I
Qb
ESG
Q (IQ Modulator)

Ib

F
Control
Unit
N

LO

Qa

Ia

Modulator
FN

ELP
Control

PC

ESG
Sets impedances rapidly, accurately and reliably
Q (IQ Modulator)
Harmonic loads are independent of each other as well as power
Requires only 3 RF measurements with which to calibrate the system for any
frequency within the bandwidth
F
Can emulate loads that are outside the Smith chart
0

35

Envelope Active Load Pull Solutions


- calibrated electronic load-pull system

Ensure impedance
setting accuracy using
non-training
load points

Applying the calibration


error model
and
1.0
re-measuring

1.0

1.0

0.5

0.5

0.5

0.0

0.0

0.0

-0.5

-0.5

-0.5

Desired Impedances
Uncalibrated Measured Impedances
-1.0

Calibrated Measured Impedances


Desired Verification Impedances

Desired Impedances
Calibrated Measured Impedances
-1.0

-1.0

-1.0

-1.0

-0.5
0.0
0.5
Before
Calibration

1.0

-1.0

After
calibration
-0.5
0.0
0.5

1.0

Calibration Advantages & Assessment


The accuracy of the loads set after
calibration is independent of the number of
data points captured, providing that there
are at least 3 points in the set.

No. of
Cal Points

-0.5

0.0

0.5

1.0

Load Setting
Verification

Percentage Difference in Loads Set


Fundamental
(F0)

Second Harmonic
(F2)

Third Harmonic
(F3)

12

0.0236 %

0.0344%

0.0335%

20

0.0237 %

0.0348%

0.0338%

36

Alternative Active Load Pull Solutions


- developed Envelope Load-Pull System
Example:
The third harmonic load was
swept around the edge of the
Smith chart with 8 equi-spaced
impedances, whilst sweeping
the fundamental load in a 4x4
grid.
1.0

Viewing the variation of the fundamental magnitude


and phase (at a desired load of 0.38 160o) as the
3rd harmonic is swept around the edge of the Smith
chart:
1.0
150

0.8

Magnitude

0.5

0.0

0.6

100

0.4

F 0 Grid Sweep Impedances


F 3 Phase Sweep Impedances

50

Phase
Magnitude

0.2
0.0

0
0

-0.5

Phase (deg)

Zopt (F0 )

2
3
4
5
Fundamental Impedance Point

37

-1.0
-1.0

-0.5

0.0

0.5

1.0

Fully Functional NLVNA: Integrated System


- Waveform Measurement and Engineering
Drain Voltage (V)

80

60
40

Development of Class J

20
0
0

200

400
600
Normalised Time (samples)

800

1000

25
0
200

Output Current[mA]

Investigating and optimizing amplifier


modes of operation

Investigating and optimizing Transistor


performance

Fan Diagrams

150
100

50

Behavioural Characterization/Modelling

0
0

10

20
30
40
Output Voltage[V]

50

Data Lookup Models

60

0.65

a1

Imaginary (W )

0.60

0.55

b1

0.50

DUT

b2
a2

0.45

0.40
-0.5

-0.4

Real (W )

measured b2

-0.3

modelled b2

-0.2

38