Text preview for : an_1287-7.pdf part of HP an 1287-7 HP Publikacje an_1287-7.pdf
Back to : an_1287-7.pdf | Home
Improving Network Analyzer
Measurements of Frequency-
Translating Devices
Application Note 1287-7
LO
RF IF
RF
LO+RF LO
LO-RF
IF
IF
LO
RF
LO+RF
LO-RF
RF IF IF
LO IF
LO
RF
LO+RF
LO-RF
RF IF IF
LO IF
LO
RF IF
RF
LO+RF LO
LO-RF
IF
IF
2
Table of Contents
Page
Introduction 3
Network Analyzer Mixer Measurement Configurations 3
Scalar Network Analyzer Configuration 3
Vector Network Analyzer in Frequency Offset Mode Configuration 5
The Upconversion/Downconversion Technique 6
Conversion Loss 7
Definition and Importance of Conversion Loss 7
Measurement Considerations 9
Mismatch Errors 9
Considerations Unique to the Scalar Network Analyzer 11
Importance of Proper Filtering 11
Frequency Response Error 12
Considerations Unique to the Vector Network Analyzer 13
Importance of Proper Filtering 13
Sampling Architecture and Issues 13
R-Channel Phase-Locking Considerations 16
LO Accuracy and Stability 16
Power-Meter Calibration 18
Accuracy Comparison of the HP 8757D and a Vector Network Analyzer 22
Fixed IF Measurements 25
Relative Phase Measurements 27
Relative Phase and Magnitude Tracking 27
Group Delay 29
Important Parameters when Specifying Group Delay 29
Absolute Group Delay 31
AM Delay 31
Reflection Measurements 33
Isolation Measurements 35
Feedthrough Measurement of Converters and Tuners 36
Absolute Group Delay -- A More Accurate, Lower Ripple Technique 38
Characterizing Mixer Calibration Standards 39
Labeling Conventions 40
Measurement Assumptions 41
Test Procedures for Characterizing the Calibration Mixers 42
Mathematics to Extract Individual Mixer Data 45
Calibrating the Test System with the Calibration Mixer 46
Calibration Configuration 46
Calibration Error Terms and Equations 48
Test Procedure for Calibrating the Test System 49
First-Order Error Correction: Frequency Response 50
Second-Order Error Correction: Frequency Response and Input Match 51
Third-Order Error Correction: Frequency Response, Input and Output Match 53
Appendix A: Calibration Mixer Attributes 55
Appendix B: Tips for Making and Measuring the Calibration Mixer 56
Appendix C: Program for Fixed IF Measurements with One External LO Source 57
Appendix D: Application and Product Notes 60
3
Introduction Frequency-translation devices (FTDs) such as mixers, converters,
and tuners are critical components in most RF and microwave
communication systems. As communication systems adopt more
advanced types of modulation, FTD designs are increasingly
complex, tests are more stringent with tighter specifications, and
the need to reduce costs is more important than ever.
The measurement trade-offs for frequency-translating devices vary
widely among different industries. Measurement accuracy, speed,
cost and ease of setup are among the considerations for
determining the best test equipment. This application note explores
current test equipment solutions and techniques that can be used
to accurately characterize and test frequency-translating devices.
Frequency-translating devices present unique measurement
challenges since their input and output frequencies differ. These
require different measurement techniques than those used for a
linear device such as a filter. This note covers linear frequency-
translation measurements, such as magnitude, relative phase,
reflection and isolation. Corresponding accuracy issues are also
discussed.
To get the most from this note, you should have a basic under-
standing of frequency translation terminology, such as "RF port,"
"IF port" and "LO port." Understanding of fundamental RF and
network analyzer terms such as S-parameters, VSWR, group delay,
match, port, full two-port calibration, and test set is also expected.
For a better understanding of such terms, a list of reference
material appear in the Appendix section.
Network Analyzer Network analyzers used for testing frequency-translation devices
include scalar network analyzers, vector network analyzers with
Mixer Measurement frequency offset capability, and vector network analyzers using an
Configurations upconversion/downconversion configuration. Each solution has its
own advantages and disadvantages. This section provides a
synopsis of the three configurations so you can quickly evaluate
which is the best fit for your measurement needs. Detailed
information about each solution is discussed in later sections.
Scalar The most economical instrument for FTD tests is a scalar network
Network Analyzer analyzer. A scalar network analyzer uses diode detectors that can
Configuration detect a very wide band of frequencies. This capability enables a
scalar network analyzer to detect signals when the receiver
frequency is different from the source frequency. Magnitude-only
measurements such as conversion loss, absolute output power,
return loss and isolation can be made, as well as nonlinear
magnitude measurements such as gain compression. Group delay
information is available in some scalar network analyzers using an
AM-delay technique, which employs amplitude modulation.
AM-delay measurements are less accurate than group delay
4
Figure 1.
HP 8711C Scalar
Network Analyzer HP 8711C
Configuration RF network analyzer
10 dB Lowpass filter
External LO source
10 dB
measurements obtained with a vector network analyzer. AM delay
typically has an uncertainty of around 10 to 20 ns, whereas group
delay with a vector network analyzer has an uncertainty as good as
150 ps. Advantages of the scalar solution include low cost and good
magnitude accuracy. As shown in Figure 1, fully integrated scalar
network analyzers such as the HP 8711C or HP 8713C provide
economical RF measurements up to 3 GHz, and include AM delay
capability. The HP 8757D scalar network analyzer, shown in
Figure 2, measures up to 110 GHz, and provides very good absolute
power measurements, particularly when installed with an internal
power calibrator and used with precision detectors. In certain
cases, such as measuring FTDs with an internal filter, the
HP 8757D with internal power calibrator and precision detector
can typically make more accurate magnitude measurements than
a vector network analyzer.
Figure 2.
HP 8757D
Scalar Network
HP 8757D
Analyzer Sweeper scalar network analyzer
Configuration
Directional bridge
Lowpass filter
External LO source
Precision detector
5
Vector Network Analyzer A more versatile solution for FTD test is a vector network analyzer.
in Frequency Offset A vector network analyzer uses a tuned-receiver narrowband
Mode Configuration detector, which allows measurements of both magnitude and
relative phase. The vector network analyzer's frequency offset
mode offsets the analyzer's receiver from it's source by a given LO
frequency, and makes frequency-translation measurements
possible.
There are two common vector network analyzer configurations for
FTD measurements. The simplest configuration is shown in Figure 3,
and is practical for testing upconverters and downconverters. This
configuration allows magnitude-only measurements with a limited
dynamic range. For example, if you are interested in the magnitude
response of the FTD's passband, the HP 8753E vector network
analyzer has 35 dB of dynamic range in the R channel and provides
a quick and easy solution.
Figure 3.
Vector Network
Analyzer in
Frequency
Offset Mode FREQ OFFS
Vector Network Analyzer ON off
LO
MENU
Lowpass DOWN
RF in filter CONVERTER
UP
CONVERTER
10 dB 10 dB
RF > LO
Start: 900 MHz Start: 100 MHz RF < LO
Stop: 650 MHz Stop: 350 MHz
VIEW
Fixed LO: 1 GHz MEASURE
LO power: 13 dBm
RETURN
6
To also measure the FTD's relative phase and out-of-band response,
Figure 4 illustrates a high-dynamic range configuration. An alter-
native high-dynamic range configuration can be achieved by
splitting the analyzer's RF output power between the device under
test (DUT) and the reference mixer (This configuration is similar to
the one shown in Figure 24). In both configurations the vector
network analyzer has around 100 dB of dynamic range. A signal
into the reference R channel is always necessary for proper phase-
locking of the vector network analyzer. In addition, the
R channel provides a reference for ratioed measurements such as
relative phase or magnitude and phase tracking. Vector network
analyzers such as the HP 8720D series and the HP 8753E have
frequency offset capability to 40 GHz and 6 GHz, respectively.
Figure 4.
Vector Network CH1 CONV MEAS log MAG 10 dB/ REF 10 dB
Analyzer, High
Dynamic Range
Configuration
Vector Network Analyzer
START 640.000 000 MHz STOP 660.000 000 MHz
RF out
Lowpass filter Reference mixer
RF in
10 dB 10 dB
External LO source
Power
splitter
The Upconversion/ A vector network analyzer in normal operating mode can also be
Downconversion configured for frequency-translation measurements. This config-
Technique uration has two main advantages. First, the instrument can be
used to measure a FTD's magnitude and relative phase response
without the need for frequency offset. As shown in Figure 5, two
mixers are used to upconvert and downconvert the signals,
ensuring the same frequencies at the network analyzer's source
and receiver ports. Second, this configuration provides a potentially
more accurate method for measuring absolute group delay. You can
simply measure two mixers and halve the response, accepting the
resulting uncertainty. You can also use a more elaborate technique
that involves characterizing the amplitude and phase of a
calibration mixer and then applying external error correction for
the most accuracy. Although this technique is more accurate, it is
7
also more complicated, requiring an external controller and an
operator who is familiar with network-analyzer data transfer and
error-term manipulation. Different levels of error correction can be
applied to achieve the desired accuracy. This technique will be
discussed in more detail in the section titled, Absolute Group Delay
-- A More Accurate, Lower Ripple Technique.
Figure 5.
Upconversion/ Vector network analyzer
Downconversion
Configuration
Bandpass Bandpass
DUT filter Mixer filter
RF IF IF RF
6 dB LO 6 dB 6 dB LO 6 dB 6 dB
Power
External LO source splitter
Conversion
Loss
Definition and Importance Conversion loss, as shown in Figure 6, measures how efficiently a
of Conversion Loss mixer converts energy from one frequency to another. It is defined
as the ratio of the output power to the input power at a given LO
(local oscillator) power. A specified LO power is necessary because
while the conversion loss of a mixer is usually very flat within the
frequency span of its intended operation, the average loss will vary
Figure 6.
Conversion
Loss
RF IF Conversion loss =
mag(f IF )
20*log [ ]
mag(f RF)
LO
Power level
Conversion loss
Frequency
8
with the level of the LO, as the diode impedance changes. As shown
in Figure 7, conversion loss is usually measured versus frequency,
either the IF frequency (with a fixed LO) or the RF frequency (with
a fixed IF). The configuration for a fixed IF measurement is
different from those described up to this point. (See the Fixed IF
Measurement section.)
Figure 7.
Two Types of Conv loss vs IF freq Conv loss vs RF freq
Conversion Loss (fixed LO freq) (fixed IF freq)
Measurements
RF IF RF IF
LO LO
Loss
Loss
0 0
IF freq RF freq
Figure 8 illustrates the importance of a flat conversion-loss
response. The DUT is a standard television-channel converter. The
input signal consists of a visual carrier, audio carrier and a color
subcarrier. Since the frequency response of the converter has a
notch in the passband, the color subcarrier is suppressed and the
resulting output signal no longer carries a valid color-information
signal.
Figure 8.
TV Tuner
Conversion Converter response
Loss Example
Visual Audio
carrier Color carrier Color sub-carrier
sub-carrier attenuated
LO
Input Signal DUT Output signal
9
Measurement Conversion-loss measurements can be made with either a scalar
Considerations network analyzer or a vector network analyzer, using the
configurations shown in Figures 1 through 5. The measurement
uncertainties are different for each type of analyzer. For both types
of analyzers, the two main systematic errors are port mismatch and
frequency response. The scalar network analyzer approach requires
additional care to minimize errors due to the analyzer's broadband
detector. For some vector network analyzers, an internal process,
called sampling, and phase-lock requirements can also create
errors. Next we will examine each of these error terms and explore
techniques to minimize their effects.
Mismatch Errors Mismatch errors result when there is a connection between two
ports that have different impedances. Commonly, a device's
behavior is characterized within a Z0 environment, typically having
an impedance of 50 or 75 ohms. Although the test ports of a
network analyzer are designed to be perfect Z0 impedances, they
are not. The imperfect source and receiver ports of the network
analyzer create errors in the calibration stage. Therefore, even
before a device under test (DUT) is connected, some errors have
already been created in the calibration stage (see Figure 9). Once
the DUT is connected, the total measurement uncertainty is equal
to the sum of the calibration error plus the measurement error.
Once the DUT is connected, interaction between the DUT's ports
and the network analyzer's ports cause mismatch errors. As shown
in Figure 9, mismatch effects generate three first-order error
signals. The first is interaction between the network analyzer's
source port and the DUT's input port. The second is between the
network analyzer's receiver port and the DUT's output port.
Figure 9.
Mismatch Effects
Calibration:
Source Calibration plane
Receiver
Measurement:
Source DUT source receiver
RF IF
Receiver
LO
( )
source DUT input
( )
receiver DUT output
( )
source receiver
Total Uncertainty = Calibration Error + Measurement Error
10
The third is between the network analyzer's source port and
receiver port. For an FTD measurement, this third interaction is
usually negligible because the conversion loss and isolation of the
FTD will attenuate the reflected signals. As frequency translation
precludes conventional two-port error correction, attenuators can
be used to improve port match.
By adding a high-quality attenuator to a port, the effective port
match is improved by up to approximately twice the value of the
attenuation. A high-quality attenuator has around 32 dB of port
match. The effective match is a function of the quality of the
attenuator as well as its attenuation, as shown in Figure 10.
Figure 10.
Effective Match
as a function of
source E ff source match
Attenuator's
Match.
Source Attenuator
( )
attenuator
(source )( attenuation ) 2
2
E ff source match = (attenuator ) + (source )( attenuation )
As shown in Figure 11 and Figure 12, a well-matched attenuator
can significantly improve the effective port match. For example,
a 10-dB attenuator, with a port match of 32 dB, can transform an
original port match of 10 dB into an effective match of 25 dB.
Figure 11.
Effective Match
as a Function of 35
Attenuator's
32dB attenuator match
Effective match (dB)
Match (Fixed 30
10 dB Attenuator). 26dB attenuator match
25
21dB attenuator match
20
18dB attenuator match
15
Region when attenuator
10 no longer results in improved match
0 5 10 15 20 25 30 35
Original match (dB)
11
However, as the match of the attenuator approaches the match of
the original source, the improvement diminishes. As shown in
Figure 12, the larger the attenuation, the more nearly the resulting
match approaches that of the attenuator. However, excessive
attenuation is not desired since this will decrease the dynamic
range of the measurement system. The port match of an FTD can
be poor, typically around 14 dB. Therefore, it is recommended that
attenuators be placed at the FTD's input and output ports.
Figure 12.
Effective Match
as a Function of
Attenuation
(Attenuator 35
Match = 32 dB) Effective match (dB) 30 20 dB attenuation
25
10 dB attenuation
20
15 6 dB attenuation
10 3 dB attenuation
5
Region when attenuator
0 no longer results in improved match
0 5 10 15 20 25 30 35
Original match (dB)
Considerations Scalar network analyzers use different detection methods than
Unique to vector network analyzers that should be considered when testing
the Scalar FTDs. Scalar network analyzers use broadband diode detectors.
Network Analyzer Although capable of both narrowband and broadband detection, the
HP 8711 series, which includes the HP 8712C and HP 8714C vector
network analyzers, uses broadband detection for FTD measurements.
Therefore, if you use an HP 8712 or 8714, use the same FTD test
considerations as you would for a scalar network analyzer.
Importance of Proper Filtering
A scalar network analyzer's broadband diode detector will detect
any signal that falls within its passband. Although a broadband
diode detector is an economical way to measure FTDs, it also can
allow certain detection errors. The diode detector will detect the
desired IF signal, as well as other mixing products or spurious
signals. To minimize the detection of undesired signals, a filter
should be placed at the detector port to pass the desired IF signal
but reject all other signals. Figure 13 shows an example of the
incorrect measurements that might result when improper IF
filtering is used in a scalar network analyzer configuration.
12
Figure 13.
Conversion Loss
Response with
and without an 1:Conv Loss /M Log Mag 1.0 dB/ Ref 0.00 dB
IF Filter 2:Conv Loss /M Log Mag 1.0 dB/ Ref 0.00 dB
dB
Swept Conversion Loss Ch1:Mkr1 1000.000 MHz
-1 -6.38 dB
-2 Ch2:Mkr1 1000.000 MHz
-4.84 dB
-3
No IF filter
-4 2
-5 2
1
-6
1
-7
IF filter
-8
-9
Abs
Start 900.000 MHz Stop 1 000.000 MHz
In Figure 13, the conversion loss measurement without the IF filter
appears to be better than it really is. The lack of an IF filter
generates erroneous results. The broadband diode detector cannot
discriminate the frequency of the received signal(s) -- it measures
the composite response. If the source is set at 1 GHz, it is "assumed"
that this is the frequency of the detected signal. Any signal that
falls within the passband of the diode detector will be detected. If
the output of a DUT is composed of the desired IF signal plus the
image frequency, LO and RF feedthrough and other spurious
signals, the diode detector will detect the composite of all the signals
within its passband. This composite signal will be incorrectly
displayed as a response that occurs at 1 GHz.
Frequency Response Error
Without performing any sort of calibration on a scalar or vector
network analyzer, the frequency response of the test system cannot
be separated from the FTD's response. One way to correct these
errors is to perform a frequency-response normalization or
calibration, using a through connection in place of the DUT.
For scalar network analyzers such as the HP 8757D, which very
accurately measures absolute power, the normalization calibration
can be performed in two steps. See Figure 21. First, the absolute
RF power is measured and stored in memory. Second, the DUT is
inserted and the absolute IF power is measured. Conversion loss is
displayed using the Data/Memory format. The conversion loss value
is very accurate since the measurements of the two absolute power
levels, RF and IF, are very accurate. Ratioing two very accurate
absolute power levels removes the frequency response error. In
some cases, a scalar HP 8757D with an internal power calibrator
and precision detector can make more accurate conversion loss
measurements than a vector network analyzer. In the Accuracy
Comparison of the HP 8757D and a Vector Network Analyzer
13
section, error terms are used to illustrate how a scalar analyzer
with internal power calibrator can be more accurate than a vector
network analyzer.
For analyzers that do not precisely measure absolute power,
corrections for the frequency response error are less accurate.
The input and output of the DUT are at different frequencies, but
the normalization can only be performed over one frequency range.
The result is that part of the test system is characterized over a
different frequency range than that which is used during the actual
measurement.
There are two choices for the frequency range used for the
normalization: either the DUT's input (RF source) range, or the
DUT's output (receiver) range. The normalization should be done
to correct the portion of the test system that contributes the largest
uncertainty; for example, this would be the portion with the most
loss or frequency roll-off. Systems and components tend to have
poorer performance at the higher frequencies, therefore the
calibration should normally be performed at the higher frequencies.
In general, high-quality, low-loss cables and connectors should be
used to minimize frequency-response errors.
For higher accuracy, combine a normalization calibration with
external error-term correction. During the normalization, only one
section of the test configuration should be connected, either the
DUT's input range or the DUT's output range. For highest
accuracy, the removed section can be characterized separately.
An external computer is used to extract the removed section's
S-parameters from the network analyzer. This data is then used
to modify the network analyzer's error terms to account for the
effects of the removed section.
Considerations Unique to Now that we have covered the important measurement considerations
the Vector Network of the scalar network analyzer, let's continue with a discussion of
Analyzer the vector network analyzer. The important considerations include:
the need for proper filtering, an accurate and stable LO, and power
meter calibration for the most accurate measurements.
Importance of Proper Filtering
A vector network analyzer has a narrowband tuned receiver. Since
the received signal is heavily filtered by an internal narrowband IF
filter, broadband detection issues encountered by the scalar network
analyzer are not present. However, proper filtering is still very
important for vector network analyzers with sampler-based
receivers, such as the HP 8753E and the HP 8720D.
Sampling Architecture and Issues
A sampler-based receiver consists of a voltage-tunable oscillator
(VTO), a pulse generator, and a sampler (switch). The VTO drives
the pulse generator, which in turn drives the sampler. As a result,
with proper tuning of the VTO, this combination replicates a down-
converted input signal at the correct intermediate frequency (IF)
14
for further processing. This combination is similar to a harmonic
mixer in which the harmonics of the LO are generated in the mixer,
and the input signal can mix with any harmonic. With proper
tuning of the LO, one of the LO harmonics is offset from the input
signal to produce the correct IF signal.
Since there are many LO harmonics, any signal (desired or not)
that is one IF away from any of the LO harmonics will be
downconverted to the network analyzer's IF and detected.
To illustrate this sampler effect, let's use the HP 8753E as an
example.
The IF of the HP 8753E vector network analyzer is 1 MHz. Errors
might result because the incoming signal is not filtered until after
it is downconverted to the IF. If there is only one signal at the
receiver, this signal will mix with one LO harmonic and is properly
downconvert to 1 MHz. However, if there are multiple signals that
are 1 MHz away from any of the LO harmonics, these signals will
be downconverted to 1 MHz, which creates erroneous responses.
Figure 14 illustrates an example of this sampler effect where the
desired IF output signal of the mixer is 110 MHz. In order to
correctly detect this signal, the HP 8753E will use a VTO of
54.5 MHz, where its second harmonic (109 MHz) will properly
downconvert 110 MHz to the desired 1 MHz IF signal. In the
illustration, we show two mixer products (6 LO-2RF and 9 LO-RF)
that would also produce IFs at 1 MHz. Notice that these two spurs
occur on either side of the LO harmonics (18 VTO and 42 VTO,
respectively), but as long as they are 1 MHz away, they will be
downconverted to 1 MHz. Aside from the signals which
Figure 14.
Diagram of
Spurious Given RF = 410 MHz IF = RF