Text preview for : an_1287-4.pdf part of HP an 1287-4 HP Publikacje an_1287-4.pdf
Back to : an_1287-4.pdf | Home
Network Analyzer
Measurements: Filter and
Amplifier Examples
Application Note 1287-4
Table of Contents
Page
Introduction 2
Measuring a Filter 2
Error Correction for Accurate
Passband Measurements 3
Swept-Power Amplifier
Measurements 4
Evaluating AM-to-PM
Conversion 5
2
Introduction The network analyzer is used for a variety of device and component
characterization tasks in both laboratory and production environments.
This highly accurate instrument can evaluate both active and passive
components, as will be demonstrated in this application note for
measurements of a filter and amplifier. With the addition of time-domain
capability, a network analyzer can also gate out unwanted responses
during measurements, leaving only the desired information.
Hewlett-Packard Company offers a wide range of RF and microwave
network analyzers for measurements from DC to 110 GHz. These
instruments are available with a wide range of options and test sets to
simplify measurements in stand-alone and automatic-test-equipment
(ATE) setups.
Often, both the magnitude and phase behavior of a component can be
critical to the performance of a communications system. A vector network
analyzer can provide information on a wide range of these devices, from
active devices such as amplifiers and transistors, to passive devices such as
capacitors and filters. This application note illustrates swept-frequency
measurements on an RF filter, and swept-power measurements on a
communications-band amplifier. The amplifier is typical of those used in
Global System for Mobile Communications (GSM) service.
Measuring a Filter Complete characterization of filters is typically achieved with swept-
frequency measurements. Shown in Figure 1 are the frequency responses
of a filter. On the left and bottom we see the transmission response in
log magnitude format, and on the right we see the reflection response
(return loss).
The most commonly measured filter characteristics are insertion loss and
bandwidth, shown on the lower plot with an expanded vertical scale.
Another common measured parameter is out-of-band rejection. This is a
measure of how well a filter passes signals within its bandwidth while
simultaneously rejecting signals well outside that same bandwidth. A test
system's dynamic range generally determines how well it can evaluate
this characteristic.
Figure 1.
Testing Filters CH1S11 log MAG 5 dB/ REF 0 dB
CH1S21 log MAG 10 dB/ REF 0 dB
with Frequency
Sweeps
Cor
Stopband
69.1 dB rejection
CENTER 200.000 MHz SPAN 50.000 MHz
START .300 000 MHz
CH1
STOP S21 log MAG 1
400.000 000 MHz dB/ REF 0 dB
Return loss
Cor
1
m1: 4.000 000 GHz -0.16 dB
m2-ref: 2.145 234 GHz 0.00 dB
2
Insertion loss ref
Cor
x2 1 2
START 2 000.000 MHz STOP 6 000.000 MHz
3
The return loss plot is typical of passive reflective filters, showing high
reflection (near 0 dB) in the stopbands, and good impedance matching in
the passband. A different type of filter, known as an absorptive filter, tends
to be well matched in both the stopband and passband, providing a good
match over a broad frequency range.
Error Correction for Variation from a constant amplitude response within the filter's bandwidth
Accurate Passband results in signal distortion. Error correction is often essential for accurate
measurements of filter passbands. When a filter's passband is measured
Measurements with a network analyzer without calibration, the response may vary
considerably, depending on the network analyzer and test cables used
(Figure 2).
When the same filter is evaluated after doing a response calibration
(normalization), the test system's transmission-tracking frequency-
response error is removed from the measured response, resulting in a
much narrower amplitude-distortion window. After normalization, the
filter's displayed frequency response still shows some amplitude ripple
caused by interaction between the test system's source and load match.
This ripple even goes above the 0 dB reference line, indicating gain (which
is impossible since passive devices cannot amplify signals). This apparent
anomaly is due to mismatch measurement error. By performing a two-port
calibration prior to the filter measurement, these errors are removed.
Following vector-error correction (two-port calibration), it is apparent
that the filter's passband amplitude response varies by only