|
RF Power Amplifier Testing
by Ed Crean, Symtx
A tough challenge for test engineers is explored in terms of test methods,
pitfalls, and measurement errors.
For the test engineer, RF and microwave power amplifier testing imposes unique
challenges. Although there are many tests that must be performed to verify
compliance to required specifications, two common tests deserve special
attention: linearity and compression.
Linearity Testing
RF power amplifier (RFPA) nonlinearities introduce third-order intermodulation
distortion (IMD) to the amplifier output signal when driven by a multitone input
signal. In the traditional two-tone test, IMD will show up as third-order
products that occur at a frequency of 2f1 - f2 and 2f2 - f1 as shown in
Figure
1.
|
Figure 1. Two-Tone Test-Frequency Spectrum |
A practical two-tone IMD test setup is shown in Figure 2. It consists of two
signal generators and two isolators with the signals summed in a two-way
combiner and then fed into the DUT. A spectrum analyzer performs the actual
measurement.
|
Figure 2. Practical Two-Tone Test Setup |
The isolators are necessary to prevent signal energy from one source leaking
through the power combiner into the other source. The presence of the signal
from one generator at the output of a second generator will cause its output
level to modulate and create IMD.
As an alternative, a vector signal generator will greatly simplify the test
setup. Modern vector signal generators use dual arbitrary waveform generators (Arbs)
to in-phase and quadrature-phase (I/Q) modulate the synthesizer output.
Additionally, advanced vector signal generators use I/Q predistortion to
generate two-tone signals with better than 70 dB of intermodulation suppression.
Prior to measuring the DUT, it is important to verify that the IMD created by
the test system is 20 to 30 dB below the DUT specification. Since the test setup
and spectrum analyzer-generated distortion components occur at the same
frequencies as the distortion components we wish to measure, there is no way of
knowing the phase relationship between the test system and DUT signals. The
potential range of uncertainty is the following:
Uncertainty (in dB) =
20 log(1 ± delta(dB)/20)
where: delta (dB) = the difference in decibels
between the test system and
the DUT-generated distortion
products
These errors are shown in Figure 3.
|
Figure 3. Test-System Measurement Error |
To measure the test-station IMD, remove the DUT and connect the two-tone source
to your load and measure the IMD of the source over all required power levels
and frequencies. Another simple test to verify you are getting good results is
to increase the level of attenuation in the spectrum analyzer. If the intermods
levels do not change, then the intermods created by the spectrum analyzer are
not affecting the measurement.
Linearity Testing: Multitone Signals
Linear RFPAs used in communications systems require a multitone signal source to
characterize their in-band distortion performance. A multitone input signal
simulates real-world conditions where many wireless traffic channels are
combined and amplified by the output power amplifier.
Creation of a multitone test signal is nontrivial. One approach would be to
extend the block diagram of Figure 2 by adding additional signal sources. While
this method will work, it may not be practical to tie up expensive signal
sources for this one test. Another approach would be to purchase a dedicated
multitone generator.
A third approach is to use an Arb-based vector signal generator’s synthesizers
that can create several hundred signals within the Arb bandwidth. Additionally,
Arb-based synthesizers allow the phase of each CW signal to be preset or
randomized to set the composite signal crest factor.
One drawback is the maximum power the vector signal generator can deliver while
maintaining linear performance. The maximum available power typically is +15 dBm
and must be reduced by the signal pk/avg ratio of the composite signal. The pk/avg
ratio is equal to:
pk/avg = 10 log(NumberOfTones)
For example, a two-tone signal has a pk/avg ratio of 3 dB while a 16-tone signal
provides a maximum pk/avg ratio of 12 dB.
If the available signal level is not sufficient, an external linear power
amplifier must be used to increase the signal level into the DUT. In the past,
this external amplifier had to have a 100-W or greater rating so it would not
introduce intermodulation products into the test system. This no longer is the
case.
Modern Arb-based synthesizers offer predistortion correction software that
predistorts the I/Q modulation to cancel the intermodulation products created by
the external power amplifier. With this software, the external amplifier can be
much smaller and less expensive than required in the past.
Compression Point Testing
The output power of a power amplifier cannot increase indefinitely. In the
extreme, there will be a point where an increase in input power does not produce
a discernible increase in output power. The transfer characteristic for a
typical power amplifier is shown in Figure 4.
|
Figure 4. Amplifier Compression Characteristics |
Gain compression, or P1dB, is defined as the output power level where the gain
has been reduced 1 dB from its small signal gain. It also is sometimes desirable
to measure the saturated power capability of an RFPA. The saturation point (PSAT)
is defined as the point where any increase in input power does not produce a
corresponding increase in output power.
Network Analyzer Methods
Modern vector network analyzers (VNAs) have power sweep capability that allows
fast and accurate compression measurements. The VNA can produce a plot as shown
in Figure 4.
In addition, modern network analyzers allow for absolute power calibration.
Power calibration requires the use of an external power meter to calibrate the
input power sweep for absolute power. Test fixturing and cabling losses can be
calibrated during power calibration.
Some VNAs have canned routines that automatically place a marker at the P1dB
point and return the gain and input power. If the network analyzer does not have
this feature, it is easily implemented in a user’s automated test routine.
Care must be taken when using the network analyzer. It is easy to overdrive and
possibly damage the amplifier under test. Pretesting of the amplifier should be
performed to determine the expected gain so the input power sweep limits can be
set accordingly.
This method is not suitable for pulsed amplifiers. Measuring pulsed amplifiers
with a CW VNA can overheat and damage the amplifier.
Power Meter Techniques
A test to determine the compression point of an amplifier can be performed using
the test setup shown in Figure 5. This setup can be used for CW or pulsed
compression testing.
|
Figure 5. Amplifier Compression Using Power Meter |
For radar and other pulsed power amplifier applications, the compression test
must be performed using a pulsed signal source. The pulsed source will not
overheat the DUT, and since the power transistors are not heated to the extent
they would be using CW, the power amplifier can produce more power.
Either a pulsed or an average type of power head can be used. When using average
power heads, the pulse modulation duty cycle must be known. Since the power
meter measures average power, a correction must be applied to determine the peak
power:
PowerCorrection = -10 log(PW/PRI)
where: PW = pulse width
PRI = pulse repetition interval
There are some advantages with using duty cycle to calculate the pulse power.
The duty-cycle technique provides the lowest cost solution with average power
meters and sensors being less expensive than peak and average power meters and
sensors.
With some pulse modulated signals, the pulse may not be purely rectangular since
there is an associated rise and fall time as well as overshoot and ringing on
the signal. The combination of these effects creates an error in the calculated
result.
In this case, a peak power meter is required. It has the capability to make a
time-gated power measurement. With time gating, the power can be measured during
the settled portion of the pulsed waveform. Also, the peak power meter will
return the peak power, eliminating the need to correct for PW and PRI.
About the Author
Ed Crean is senior RF engineer at Symtx and has 22 years of experience in RF
design, manufacturing, and management. Before joining Symtx, he specialized in
high-power RF amplifier design and development at Nokia and Andrew Corp. Mr.
Crean holds a B.S.E.E. from the University of Buffalo and an M.S.S.E. from Texas
Tech University. Symtx, 4401 Freidrich Lane, Austin, TX 78744, 512-328-7799, e-mail:
ecrean@symtx.com |