AN1526 Freescale Semiconductor / Motorola, AN1526 Datasheet - Page 5

• Manufacturer Part Number: AN1526
• Description: RF Power Device Impedances: Practical Considerations
• Manufacturer: Freescale Semiconductor / Motorola

specified on Motorola data sheets (or any other vendor’s data
sheet known to the authors) is the impedances at the
harmonic frequencies presented to the device. In most
cases, it is the first shunt capacitance combined with the
device package series inductance that determines the
second harmonic impedance present at the INTERNAL
collector–emitter terminals. The designer should simply be
aware that it is possible to have a perfect match at the
fundamental frequency and NOT get the published
performance due to improper harmonic terminating
impedances.
technique for these very low impedances. Several
techniques exist for the measurement of the source and load
impedances presented to a device. The simplest approach
is to construct an impedance measuring probe similar to
those shown in Figures 17–19. These probes are nothing
more than a blank device package appropriate for the fixture
being used, with a short piece of small semi–rigid 50 ohm
coax carefully soldered in place. The most convenient
method of calibration is to first perform a full one–port
calibration on the network analyzer using a 3.5 mm
calibration kit. Then, with the probe clamped into the tuned
test fixture and a piece of copper foil slipped under the
probe’s center tab, use a port extension to reset the phase
angle at the frequency of interest to 180 degrees. This
technique is valid for 1/2 CQ packages through 520 MHz.
Square packages such as the CS–12 can be used up to
1 GHz. An error is present due to the discontinuity between
the end of the semi–rigid coax and the package edge. This
discontinuity has been de–embedded and found to be
negligible.
Break–Apart Test Fixture
a fixture which can be “broken apart” at the reference planes.
Figure 22 shows the construction of the break–apart test
fixture. After tuning the circuit for the desired performance
the test fixture is broken apart, the bridge on which the device
sits removed, and co–axial connectors installed at the
reference planes (Figure 23). The impedances presented to
the device can then be measured using the network analyzer
calibrated with a standard ‘N’ type calibration kit, and a port
extension applied to rotate out to the reference plane.
described above, the break–apart test fixture offers the
advantages of improved repeatability and more consistent
measurement plane location. This is at the expense of more
time intensive measurements, since at each frequency point,
the fixture has to be partially disassembled and connectors
installed before impedances can be measured. The
break–apart test fixture really comes into its own when used
for in–line impedance measurements. Each half of the fixture
can be characterized as a two port using a network analyzer,
and with a knowledge of the impedances presented
externally to the test fixture, the device impedances can be
de–embedded [26]–[30]. The test fixture provides the all
important “close in” match and a measure of broadband
performance, while the external tuners provide the fine tuning
at the measurement frequencies within the operating band.
The test fixture also provides biasing for the transistor at
current levels higher than can normally be accommodated
MOTOROLA SEMICONDUCTOR APPLICATION INFORMATION
A further point of clarification is the measurement
A technique which avoids these discontinuities involves
In the off–line impedance measurement techniques
Freescale Semiconductor, Inc.
For More Information On This Product,
Go to: www.freescale.com
by external bias tees, and if correctly bypassed, improves
the stability under mismatch conditions.
Automated Tuners
wisely, they can provide in–line impedance measurement
capability, and allow rapid characterization of a device under
load–pull conditions. In–line measurement of device
impedances makes it practical to measure a sufficient
number of devices to establish impedance distributions for
a particular device type. Currently available systems are,
however, still too slow to use in a production environment
for 100% testing of RF transistors. Load–pull characterization
can be performed under custom conditions enabling an
amplifier designer to start the matching network synthesis
with impedance data representative of the final operating
conditions.
impedance matching test fixturing for three very important
reasons. a) The impedance transformation range is normally
limited to 10:1 precluding their use with most power
transistors that have relatively low input impedances outside
this range. b) Optimum performance of a power amplifier
requires careful attention to the harmonic loading, which in
many cases requires shunt capacitance close to the
package. The low input/output impedances also require low
loss return paths for the circulating ground currents. c) Bias
networks can be designed on the test fixture to minimize the
potential for instability necessary for correct operation of
automated tuner search algorithms.
Measurement Accuracy Factors
dictates that the test fixture be designed with some
mechanical tolerance to allow devices to be tested at the
maximum extremes of the device width and height over the
seating plane. The contact of the leads to the test fixture
pads is therefore variable contributing additional series
inductance to the input, output and ground leads. This results
in more inductive device impedances than expected if careful
steps are not taken to minimize this error.
Absolute Accuracy of Measurement Instrument,
Network Analyzer
measurement accuracy over the last ten years [32]–[37]. Not
too many years ago device impedances were routinely
measured with vector voltmeters, using only a single
correction term for the frequency response of the couplers.
Now, for 1 port measurements, three term error correction
is the norm, correcting for directivity, source mismatch and
frequency response. The network analyzer does still
introduce a degree of measurement uncertainty proportional
to the magnitude of the reflection co–efficient. This is more
of a problem with higher power devices where the real part
of the impedances can be below 1 ohm. The reader is
referred to the manufacturers operating manuals for the
maximum possible magnitude of the error. In the worst case,
for a transistor with a 1 ohm real part of the impedance, the
error can be as large as
Automated tuners offer a number of advantages. If used
Automated tuners need to be coupled with some type of
There is some tolerance in the package dimensions which
Great improvements have been made in network analyzer
1 ohm.
5