Vous êtes sur la page 1sur 9

Fachhochschule Frankfurt am Main

University of Applied Sciences


Faculty of Computer Science and Engineering
RF-Laboratory

Linear Network Analysis

Prof. Dr.-Ing. G. Zimmer


1 Introduction
1.1 Network Analyzer
To measure the linear properties of a network at rf and microwave frequencies today in most
cases network analyzers are used which directly allow the measurement of the scattering param-
eters of a 1- or 2-port in the frequency domain. One must distinguish between scalar network
analyzers (SNA) and vector network analyzers (VNA). While a SNA is only able to measure
the absolute value of a scattering parameter, a VNA is able to measure its absolute value and its
phase. This is only possible with the help of powerful processors which allow the correction of
systematic errors and a flexible representation of the measured data. Due to its higher flexibility
we will only discuss the VNA in the following sections. Fig. 1 shows the principle setup of a
NVA. It consists of the following building blocks:

Processor/Display
P P
P
Signal Source

forward s? reverse


s B
s LO B
B
a0 - ? j B
b0- j B
B
a3 - j ADC

b3- j

Test-Set
B B
B B Receiver

port 1 port 2

Figure 1: Principle setup of a NVA

Signal Source
To create the RF-power for the test system either open-loop voltage-controlled oscillators
(VCOs) or more expensive synthesized sweepers with low phase noise and high frequency
stability are used. To keep the phase information during the down convertion process in
the rf-mixer, additional rf-signals have to be deduced from the original test signal.

Test-Set
The main rf-building block of a network anlayzer is the so-called Test Set. A rf-switch
switches the signal coming from the signal source into two different branches. One for
the foward (S11 , S21 ) and the other for the reverse measurement (S22 , S12 ). To provide the
reference signals (forward a0 ), (reverse a3 ), usually broadband resistive power dividers
are used, while for separation of the incident and reflected waves from the device under
test (DUT) in most cases broadband directional couplers are implemented.
Receiver
To detect the signal usually a tuned receiver uses a local oscillactor (LO) to mix the rf to a
lower intermediate frequency (IF). The IF signal is bandpass filtered and analog-to-digital
converters (ADC) are used to prepare the signal for digital-signal processing.

Processor/Display
The last major block of hardware in a network analyzer is the processor/display section.
This is where the reflection and transmission data is error corrected and put in different
representations.

Today most network analyzers are realised using coaxial connectors which cover the frequency
range up to 60 GHz. Above these frequencies the microwave part (Test Set) is build using
rectangular wave guides.

One Port Calibration


In the Test-Set of a network analyzer, usually directional couplers are used to measure the reflec-
tion coefficient of the DUT. In contrast to an ideal directional coupler, real directional coupler
shows a finite directivity, which limits the smallest reflection coefficient which can be measured
without calibration. To overcome these problems the systematic errors of the measurement
setup must be evaluated and used for correction. Today this is no longer a problem due to cheap
and powerful processors. Fig. 2 a) shows in principle the setup used to measure a reflection co-
a0 b0 a0 b0

HH
HH
error
G rA G rA
real adapter
ideal
a) b)

a0 s e10 -s
@
@ @

@
e00 e11 rA
c)

b0 e
s 01 @
@s

Figure 2: One port error model

efficient. The non-ideal properties of the used directional couplers can be described, as shown
in Fig. 2 b), by a so-called error model, which is put between the ideal reflectometer and the
DUT. Fig. 2 c) shows the associated signal-flow chart [1]. During the calibration process all
three error parameters are measured, to be used later for error correction. In principle all errors
are assumed to be located in a two port, which is used as adapter, between the DUT (rA ) and
the measured reflection coefficient. The adapter is described by the three error parameters e00 ,
e10 e01 and e11 . Calculating the measured reflection coefficient rM = b0 /a0 in dependence of the
reflection coefficient of the DUT rA taking the error adapter into account, yields the following
equation:
e10 e01
rM = e00 + r (1)
1 e11 rA A
Now we will discuss the influence, which the three different parameters have on the measuring
process
e00
Accorcing to equation 1, the influence of the parameter e00 is greatest when objects have
a small reflection coefficient. In the limit rA 0, the measured reflection coefficient will
be equal to e00 . Thus it accounts for the finite directivity of the used directional couplers.
e10 e01
The influence of these parameters becomes clearest in the limit e00 = e11 = 0 and we find
that the measured reflection coefficient will be equal to e10 e01 rA . Thus it accounts for the
transmission factors of the used directional couplers and transmission lines.
e11
This parameter accounts for the source match of the measurement setup.
In the ideal case the parameters of the error adapter show the following values: e00 0, e10 e01
1 and e11 0. In this case the measured reflection coefficient would be equal to the reflection
coefficient of the DUT rA .

OSL-Calibration
The error adapter for measuring a one port consists of three independent complex error param-
eter, which are functions of the frequency. To carry out a correction for each frequency value
all three parameters have to be determined using three independent calibration standards. One
frequently used method is called OSL-calibration. The letters originate from the used standards
Open, Short, Load. In the following section we want to examine how to use the measurement
results of the calibration standards to determine the error parameters e00 , e10 e01 and e11 in a
coaxial measurement setup.

Measuring Load - rML


In a coaxial system the measuring port has to be terminated with the characterisitic
impedance of the coaxial transmission line. In case of an ideal matched load this would
result in rA = 0. With increasing frequencies this becomes more and more complicated.
This is one of the drawbacks of the OSL-calibration. Assuming the ideal case measuring
the load results directly in the error parameter e00 .

rML = e00 (2)


Measuring Open - rMO
The realisation of an ideal calibration standard Open is impossible in pratice. A real
open coaxial transmission line would always radiate a certain amount of power resulting
in |rA | < 1. Instead one can only realise a calibration standard as shown schematically in
Fig. 3, which always results in a certain capacitance CE at the end of the transmission line.
To use this as calibration standard its behaviour has to be carefully determined. Neglect-
ing losses the reflection coefficient of the standard can be written as: rA = exp( jC ). Of
course the frequency dependence of the phase angle C must now be very precise. This
results in the following equation for the measurment of the open standard (3):
e10 e01
rMO = rML + exp( jC ) (3)
1 e11 exp( jC )
-
- C
- E

Figure 3: Schematic of an open coaxial transmission line

Measuring Short - rMS


Realising the Short standard in coaxial tranmission line technique is less complicated
and we will assume that the standard behaves ideal (rA = 1). For the measured reflec-
tion coefficient we get (4):
e10 e01
rMS = rML (4)
1 + e11
Equations 3 and 4 can be used to determine the still unkown parameters e11 and e10 e01 :

rM0 exp( jC ) + rMS + rML [1 + exp( jC )]


e11 = (5)
rMO rMS

(rML rMS )[1 + exp( jC )]


e10 e01 = (rMO rML ) (6)
rMO rMS
With the help of the equations 2, 5 and 6 the error parameters shown in Fig. 2 can be determined
using the measured reflection coefficients rML , rMO and rMS for each frequency value. The
denominator of the equations 5 and 6 is determined by the term rMO rMS . If the parameters of
the error adapter are next to their ideal values, the denominator will show an absolute value next
to two and the equations 5 and 6 will result in well defined values. But if the phase angle due to
the end capacitance of the open standard becomes too large, the term rMO rMS may approach
zero, defining an upper frequency limit for the used standard. If all error parameters of the error
adapter (Fig. 2) are known for all measured frequency values, equation 1 can be rearranged to
get the corrected reflection coefficient of the DUT.
rM e00
rA = (7)
e11 (rM e00 ) + e10 e01
1.2 Modelling Reactive One Ports
The picture given in Fig. 4 shows a printed circuit board (FR4) on which all reactive one
ports are realised which will be examined in more detail during course of this laboratory work.
Usually a printed circuit board consists of a non-conducting base material with one or two

Figure 4: Picture of the test board

conducting copper layers as top and/or bottom layer. Typically the copper layers posses a
thickness of 35 m, 70 m or 140 m depending on the current density they have to carry.
Instead of the thickness sometimes also the mass of the conducting layer is given in ounce
per square feet (oz/sq.ft). Today the base material consist of glass fibre matts filled with epoxy
(FR4). Compared to older materials it shows lower leakage currents and better RF performance.
The elements named L1 to L3 realise real lumped inductors while the elements named C1 to
C3 act as real lumped capacitors. As was already shown in [3] realizing a lumped reactive
element always also results in the realisation of paracestic elements, which have to be taken
into account, if one wants to describe the frequency behaviour of the real element. Fig. 5 shows
simple equivalent circuits able to describe the behaviour of real inductors (a) or capacitors
(b). During this laboratory exercise all elements of the equivalent circuit will be deduced from
measurements. Additional to the inductors and capacitors a series resonance circuit (SSK) and
Rs Ls Rs
r r r r r r

Cp C
a) Rp L

b) Rp

r r r r r

Figure 5: Equivalent circuit of a) real inductor b) real capacitor

a parallel resonance circuit (PSK) is realised on the test board. Both circuits are combinations
of the real inductor L2 and the real capacitor C2. If one would try to carry out an analysis of
the combined circuit on the base of the equivalent circuits given in Fig. 5 one would end up in
very complicated expressions, hiding the essential behaviour of the circuit. To circumvent this
problem we will use simplified equivalent circuits for the series and parallel resonance circuits,
which are valid only next to the resonance frequency. For the complex impedance of both

RS L GP


s s s s s s

C
Figure 6: Simple equivalent circuit of a series and a parallel resonance circuit

elements we find according to [3]:


Rp
Z s ( j) = Rs (1 + jQ v) Z p ( j) = (8)
1 + jQ v

Here the quality factor Q of the resonance circuit and v = /0 0 /, called detuning factor
were introduced. If one defines the frequencies f2 and f1 to be the frequencies of 45o detuning,
one can get the following relation between the quality factor Q, the bandwith B = f2 f1 and
the resonance frequency f0 of the resonance circuit [3]:
f0
Q= (9)
B

2 Laboratory Experiments
2.1 Measuring the Inductors
With the help of the network analyzer the reflection coefficient of the inductors shall be mea-
sured in the frequency range from 1 MHz to 350 MHz. The first step is to define the sweep
range of the network analyzer. To do this proceed as follows:

Press button Start in menu Stimulus. Define the start frequency by pressing the numeric
buttons and finish by pressing the dimension button MHz

Press button Stop in menu Stimulus. Define the stop frequency by pressing the numeric
buttons and finish by pressing the dimension button MHz

Now the sweep range of the network analyzer is defined and we can proceed with the one port
calibration, performing the following procedure:
Press button Cal in menu Response. Further submenus will be displayed on the screen of
the network analyzer. To define the Cal-Kid to be used press button Cal Kit and select the
submenu Select Cal Kit. Chose the Kit 3.5mmD. Pressing two times the button Return
brings you back to the Cal menu.
Press button Calibrate Menu to open the calibration menu. Choose S11 1-Port calibra-
tion. The submenus Open, Short and Load will be displayed.
To perform the calibration one first connects the calibration standard to port 1 and after
connecting press the appropriate button. This has to be done with all three calibration
standards. To finish the calibration press the button Done. Now the error parameters are
calculated.
The calibration has to be saved by pressing the button Save/Recall in the menu Instrument
Stat and by pressing the button Save State.
After calibrating the network analyzer the different objects to be measured may be connected to
port 1. To see their frequency response the button Meas in menue Response has to be pressed.
With the help of the button Format one can chose Smith Chart and with the help of the marker
functions one can read out the values of L and Rs at the lowest frequency and f p and R p at the
resonance frequency at the intersection point with the real axis of the smith-chart.

2.2 Measuring the Capacitors


To perform the measurements on the capacitors the frequency range of the networkanalyzer
has to be changed to 10 MHz to 1 GHz and a new one port calibration has to be carried out.
The frequency response of the DUT is again displayed in the smith-chart. With the help of the
marker functions one can read out the value C of the real capacitor at the lowest frequency.
At the same frequency the value of a series resistance Rx is displayed. According to [3] the
following equation can be used to calculate the corresponding parallel resistance R p .
1
Rp =
(C)2 Rx
Furthermore the marker functions can be used to read out the values of the series resistance Rs
and the series resonance frequency fs at the intersection point with the real axis of the smith-
chart.

2.3 Measuring the Resonance Circuits


To measure the resonance circuits the sweep range of the network analzer has to be set to
the range of 50 MHz to 150MHz and a new one port calibration has to be carried out. The
response is again displayed in the smith chart. The values of the series resistance Rs of the
series resonance circuit and the parallel resistance R p of the parallel resonance circuit, as well
as the resonance frequency can be read out at the intersection point with the real axis of the
smith-chart. To determine the bandwidth, we use the points of 45o detuning. One uses the
marker functions to find approximately the frequencies of the following readouts:
Rp
Zs = Rs (1 j) Zp = (1 j)
2
With the help of both frequencies the bandwidth and hence also the Q-factor of the resonance
circuits can be determined. According to [3] the characteristic impedances and the values of the
inductor and capacitor of the resonance circuit can be calculated using the following equations:
Zc Rp
Qs = bzw. Q p =
Rs Zc
and

1
Zc = 0 L =
0C

3 Data analysis
1. Calculate the self-capacitance of the winding using the parallel resonance frequency and
the measured values of L1, L2 and L3

2. Carry out a simulation for the real inductors L1, L2 and L3 using the equivalent circuit
according to Fig. 5 a) and display the results in the smith chart.

3. Calculate the series inductance of the real capacitors C1, C2 and C3 using their series
resonance frequency

4. Carry out a simulation for the real capacitors C1, C2 and C3 using the equivalent circuit
according to Fig. 5 b) and display the results in the smith chart.

5. Carry out a simulation using the simple equivalent circuits of the parallel resonance cir-
cuits according to Fig. 6 and the detailed equivalent circuit using the equivalent circuit of
the real capacitor and the real inductor (Fig. 5). Compare the absolute value of the total
impedance of both simulations in one chart (LinMag).

References
[1] Rytting, D.: An Analysis of Vector Measurement Accuracy Enhancement Techniques,
Hewlett Packard Application Note.

[2] Thumm, M.; Wiesbeck, W.; Kern, S.: Hochfrequenzmetechnik, B. G. Teubner Verlag,
1997.

[3] Zimmer, G.: Hochfrequenztechnik, Lineare Modelle, Springer Verlag, 2000.

Vous aimerez peut-être aussi