A Spectrum Analyzer for the Radio Amateur Part 2

By Wes Hayward, W7ZOI, and Terry White, K7TAU

A Spectrum Analyzer for
the Radio Amateur—Part 2

Part 1,14 we described the Here are
some
In design and construction of a examples of
simple, yet useful spectrum procedures
analyzer. This installment pre- you can use
sents some applications and methods that to become
extend the underlying concepts. familiar with
your new
Amplifier Gain Evaluation spectrum
analyzer.
One use for a spectrum analyzer is ampli-
fier evaluation. We can illustrate this with a lyzer signal that we saw before the amplifier (The video filter was turned on and the sweep
small amplifier from the test-equipment was inserted. The gain measured 31 dB.17 In- rate reduced until the signal amplitude was
drawer—an old module that has been pressed creasing the analyzer attenuation by 10 dB stable.) The analyzer’s attenuator, set for a
into service for a variety of experiments. This (for a reference level of –10 dBm) and in- reference level of –10 dBm at the top of the
circuit, shown in Figure 11, is used for illus- creasing the generator output to –41 dBm screen, was confirmed with a calibration sig-
tration only and is not presented as an opti- produced the same gain, but a growing har- nal from the signal generator. We adjusted
mum design. It’s a project that grew from monic output. The second-harmonic re- the IF GAIN to compensate for changes in
available parts and may be familiar to some sponse was now at –43 dBc and a third har- analyzer bandwidth and for log-amplifier
readers. The circuit uses four identical monic appeared out of the noise at –60 dBc. drift.
2N5179 amplifier stages. A combination of
emitter degeneration and parallel feedback We continued the process—moving both The output of the two-tone generator sys-
provides the negative feedback needed to step attenuators produced an amplifier out- tem was adjusted to produce a spectrum ana-
stabilize gain and impedance. (Ideally, con- put of 0 dBm with second and third har- lyzer response of –10 dBm per tone. The
struction and measurement of a single stage monics at –28 and –36 dBc, respectively. IMD responses were readily seen, now 47 dB
should precede construction of the complete The next 10-dB step, however, didn’t work below the desired output tones. The output
amplifier.) as well, producing gain compression. With intercept is given by
a drive of –21 dBm, the output was only
We began the experiment by setting the +4 dBm, a gain of only 25 dB instead of the IP3out = Pout + IMDR
signal generator at a known power level, small-signal value of 31 dB. 2
–20 dBm. (If a good signal generator is un-
available, you can easily build a suitable Amplifier Intermodulation Distortion where Pout is the output power of each de-
substitute.15 We used a surplus HP8654A for sired tone (–10 dBm) and IMDR is the
most of these experiments.) With the genera- Next, we measured intermodulation dis-
tor and spectrum analyzer connected to each tortion (IMD). The setup for these experi- intermodulation distortion ratio, here 47 dB.
other through 50-Ω coax cable, we set the ments is shown in Figure 12. Two crystal- The output intercept for this amplifier mea-
generator to 14 MHz with 10-dB attenuation controlled sources18 at 14.04 and 14.32 MHz
ahead of the analyzer. The amplifier was not are combined in a 6 dB hybrid combiner (re- sured +13.5 dBm. This is well in line with
yet connected. We adjusted the analyzer’s IF turn-loss bridge),19 applied to the 15 MHz
GAIN control for a response at the top of low-pass filter and a step attenuator having expectations for such a design.
the screen. Using a resolution bandwidth of 1 dB steps. This composite signal drove the
300 kHz allows for a fast sweep without dis- amplifier, with its output routed to the spec- When performing IMD measurements,
tortion. In our case, the second harmonic was trum analyzer. For this measurement, we
only 26 dB below the peak response. How- dropped the resolution bandwidth to 30 kHz. it’s a good idea to change the signal level
ever, when we added a 15 MHz low-pass
filter, the second harmonic dropped to while noting the resultant performance.
–57 dBc.16 The third harmonic is well into
the noise. (It is not unusual for a signal gen- Dropping the drive power by 2 dB should
erator to be moderately rich in harmonics.)
cause a –2 dB response in the desired
Next, we inserted the amplifier between
the signal generator and the analyzer, keep- tones, accompanied by a 6 dB drop in distor-
ing the low-pass filter in the generator out-
put. The on-screen signal went well above tion-product tones. The output intercept
the top as soon as the amplifier was turned
on. Decreasing the signal generator output to
–51 dBm produced the same –20 dBm ana-

14Notes appear on page 40.

September 1998 37

Figure 11—Sample wideband amplifier used to illustrate amplifier measurements.

should remain unchanged. Figure 12—Equipment setup for evaluation of amplifier intermodulation distortion.
The IMD in the preceding example was
• Confirmed the lack of on-screen distor- Figure 13—Test setup for impedance-
47 dB below the desired output tones, a value tion with two tones at the reference level. match measurements with return-loss
that we obtained by simply reading it from bridge.
the face of the ’scope, possible because we • Increased the drive of each tone by
use a log amplifier that has moderate log fi- 10 dB to provide a pair of –20 dBm tones to the analyzer for full-scale response with the
delity. If the log amplifier was not as accu- the analyzer. This higher-than-reference- LOAD port of the return-loss bridge open cir-
rate as it is, we could still get good measure- level input produced distortion products cuited. Placing a 50-Ω termination momen-
ments. In this example, you would note the 66 dB below the reference level, or –96 dBm. tarily on the LOAD port, produced a 38-dB
location of the distortion products on the dis- The input signals producing this were each signal drop. This is a measure of the bridge
play. Then, using the step attenuator, de- –20 dBm, so the IMD ratio is (–20) – (–96) = directivity. A 38-dB directivity is more than
crease the desired tones until they are at the 76 dB. Following the earlier equation, the adequate for casual measurements.
noted level. The result would be –47 dBc for input intercept was +18 dBm.
the distortion level, a measurement that de- Then, we removed the termination from
pends solely on the accuracy of the attenua- • A 2-dB drive increase produced the ex-
tor. This illustrates the profound utility of a pected 6-dB distortion increase. If this had
good step attenuator, an instrument that can not occurred, distortion measurements under
be the cornerstone of an excellent basement overdrive would be suspect. The +18-dBm
RF laboratory. value seems to be a good number. This
analyzer generally seems happy with signals
During the third-order output-intercept 20 dB above the top of the screen, but not
determination just described, we assumed much more.
that the distortion was a characteristic of the
amplifier under test. This may not be true. It The intercept for the analyzer with attenu-
is important to determine the IMD character- ation in place is the measured value with no
istics of the spectrum analyzer used for the pad plus the attenuation. Hence, with 20 dB
measurements before the amplifier measure- of attenuation, the input intercept will be
ments are fully validated. Specifically, for +38 dBm, and so forth.
results to be valid, the input intercept of the
analyzer should be much greater than the Return-Loss Measurements
output intercept of the amplifier under test.
The next amplifier characteristic that we
The spectrum analyzer input intercept is measured was the input impedance match, or
easily measured with the same equipment return loss, performed with the setup shown
used to evaluate the amplifier. The two tones in Figure 13. With the signal generator set for
are applied to the analyzer input with no at- a 14-MHz output of about –30 dBm, we set
tenuation present at the analyzer front end.
Then, the input tones are adjusted for a full-
screen response. In this condition, there
should be no trace of distortion. Although
this is generally an adequate test, it does not
establish a value for the input intercept. To
do that, we must overdrive the analyzer, us-
ing signals that exceed the top of the screen.

The following steps were used to measure
the analyzer input intercept:

• We calibrated the analyzer for a reference
level of –30 dBm with a 30-kHz resolution.

38 September 1998

Figure 14—Output of a typical QRP trans- level low enough that the amplifier remains we want to experiment on the VHF and UHF
ceiver kit. The 1-W plus output at 7 MHz is linear. In this case, we saw no difference in bands, but we need to examine higher-order
the dominant signal; all others are the results with drive that was 10 dB higher. harmonics of HF gear. One method we can
spurious outputs more than 40 dB down, use with a regular receiver is a converter,
currently meeting FCC specifications. When performing return-loss measure- usually crystal controlled. The same can be
ments—and indeed, most spectrum-analyzer done with a spectrum analyzer, although
Figure 15—Output of a simple homemade measurements—it is wise to place at least crystal control is not needed. We can build a
QRP rig. The desired signal is the large 10 dB of attenuation ahead of the spectrum- simple block converter, consisting of noth-
pip at the center of the trace. Two analyzer mixer. When this attenuation is ing more than a 100 to 200 MHz VCO (just
measurable spurious responses exist, one switched in, the reference level changes from like that used in the analyzer) and a diode
to the left and one to the right of the main the top of the screen to a point down screen a ring mixer. A Mini-Circuits POS-200 VCO
signal. The response to the left is 3.5 MHz bit. Return loss is then measured as a decibel with a 3 dB pad will directly drive a Mini-
feedthrough from the VFO at –64 dBc; the difference with regard to the new reference. Circuits SBL-1 or TUF-1 mixer to produce a
response to the right of the desired signal block converter with a nominal loss of 10 dB.
is a second harmonic at –44 dBc. Antenna Measurements (One of the spectrum analyzer front-end
boards could be used, with slight modifica-
the bridge and placed it on the amplifier out- It is interesting to look at some other tion, for the block converter.)
put. A short length of cable connected the impedance values while the return-loss
bridge load port to the amplifier input with bridge is attached to the signal generator and This block converter allows analysis of
power applied to the amplifier. The result spectrum analyzer. The obvious choice is the much of the VHF spectrum. With the con-
was a response 20 dB below the top of the station antenna system, especially if it is verter VCO set at 100 MHz, frequencies from
display; 20 dB is the return loss for the am- connected through a Transmatch. Playing 100 to 170 MHz are easily studied. The 70 to
plifier input, an excellent match for a gen- with the tuning will readily demonstrate that 100-MHz image is also available—it can also
eral-purpose amplifier.20 The match im- the return-loss bridge and sensitive detec- lead to confusion, as a few minutes with a
proved slightly when the output load was tion system will allow adjustments to accu- signal generator will demonstrate. With the
removed, an unusual situation. racy unheard of with traditional diode detec- converter VCO up at 200 MHz, the 200 to
tor systems. Although such tuning accuracy 270 MHz spectrum is also available. Clearly,
The next variation measured the output is not needed in a normal antenna installa- there is nothing special about the particular
match for the amplifier. The load was trans- tion, it is interesting to see what can be mea- VCO used in the converter. All that is required
ferred to the amplifier input and the cable from sured when the need does arise. to convert other portions of the low UHF
the bridge LOAD port was moved to the ampli- spectrum to the analyzer range is a different
fier output, producing a reading 15 dB below Transmitter Evaluation VCO, and perhaps, a higher-frequency mixer.
the screen top. This match was virtually un- We will soon build similar block converters
changed when the input termination was re- Another obvious application for a spec- to allow analysis of the 432-MHz area. One
moved. The weak dependence of both ampli- trum analyzer is in transmitter evaluation. of the popular little UHF frequency counters
fier return losses is the result of a four-stage Figure 14 shows the output of a typical QRP is quite useful with these converters. The
design. A single-stage feedback amplifier will transceiver kit. The 1-W plus output at 7 MHz block converters should be well shielded
have port impedances that depend strongly on is the dominant signal, with all others being and decoupled from the power supply.
the termination at the opposite port. transmitter spurious outputs. All spurs are
more than 40 dB down, which meets current Even without the analyzer, the block con-
The –30 dBm drive from the generator FCC specifications. On the other hand, signi- verter is a useful tool. For example, it can be
provides an available power of –36 dBm from ficantly better performance is easily obtained, used with a 10 MHz LC band-pass filter and
the bridge LOAD port. This is low enough especially if the builder has the facilities to an amplifier to directly drive a 50-Ω-termi-
that the amplifier is not over-driven. The measure them. Figure 15 is a photograph of nated oscilloscope. This can serve as a sensi-
match measurements should be done at a a simpler QRP rig with two measurable spu- tive detector for the alignment of a 110-MHz
rious responses. One is the 3.5-MHz feed- filter.
through from the VFO at –64 dBc; the second
is a harmonic at –44 dBc. As useful as the block downconverter is, it
has image-response problems that can greatly
The output available from a typical QRP confuse the results. The preferred way to get
rig (and certainly higher power rigs) is to the higher frequencies is with a spectrum
enough to damage the spectrum-analyzer analyzer—just like the one we have de-
input circuits. Attenuators that we generally scribed—but having higher-frequency oscil-
build are capable of handling 0.5 to 1 W lators and front-end filters. If you’re careful,
input without damage, while commercial you can build helical resonator filters for the
attenuators are rated at from 0.5 to 2 W in- 500-MHz region, or higher, with sufficiently
put. The mixers used in this analyzer can be narrow selectivity to allow a second conver-
damaged with as little as 50 to 100 mW sig- sion to 10 MHz. A more practical route uses a
nals. Two methods can be employed to view third IF. Such a triple-conversion analyzer is
the output of a high-power transmitter with- shown in Figure 17, where a 1 to 1.8 GHz
out causing damage to the spectrum ana- VCO moves signals in the 0 to 800-MHz
lyzer. In one, the transmitter output is run spectrum to a 1 GHz IF. This signal is still
through a directional coupler with weak cou- easily amplified with available monolithic
pling to the sampling port—perhaps –20 to amplifiers. Building a 1 GHz filter should not
–30 dB. The majority of the output is dissi- be too difficult, for a narrow bandwidth is not
pated in a dummy load. The second method required. A bandwidth of 20 or 30 MHz with
uses a fixed, high-power attenuator. Figure a three-resonator filter would be adequate.
16 shows an attenuator that will handle The resulting signal is then heterodyned to a
about 20 W while providing 20-dB attenua- VHF IF (such as 110 MHz) where the remain-
tion. The design is not symmetrical. ing circuitry is now familiar.

Spectrum Analysis at Higher Clearly, there are several ways to attack
Frequencies the project. Recent technology offers some
help in the way of interesting band-pass filter
Although the 70-MHz spectrum analyzer structures, as well as high-performance,
is extremely useful, we constantly wish that low-noise VCOs. Time spent with some cata-
it covered higher frequencies. Not only do
September 1998 39

An inside view of a prototype VHF band-
pass filter. See Figure 7.

Figure 16—A 20-W, 20-dB attenuator for transmitter measurements.

Figure 17—Block
diagram for a version of
this analyzer that will
function into the VHF and
low-UHF regions.

logs and a Web browser could be very pro- ally, the casual specifications that we have the paper and the role of measurements in
ductive in this regard. Just as important is the enjoyed for many years are being replaced by
experience that the lower-frequency analyzer new ones that are more stringent—and more amateur experiments.
will provide. Not only is it the tool (perhaps realistic in safeguarding the spectral envi-
supplemented with block converters) that is ronment and reflecting the sound designs that Notes
needed to build a higher-frequency spectrum we all strive to achieve. Equipment such as 14Wes Hayward, W7ZOI, and Terry White,
analyzer, but it is the vehicle to provide the the spectrum analyzer described here can
confidence needed to tackle such a chore. provide the basic tool needed to meet this K7TAU, “A Spectrum Analyzer for the Radio
new challenge. Amateur,” Part 1, QST, Aug 1998, pp 35-43.
Summary 15The wide-range oscillator presented in Fig 68,
Acknowledgments Chapter 7 of Solid-State Design for the Radio
This analyzer has been a useful tool for Amateur (Newington: ARRL, 1997) is still in-
over 10 years now. It was a wonderful ex- Many experimenters had a hand in this tact and still often used.
perience and fun to build HF and VHF CW project and we owe them our gratitude. Jeff 16The term dBc refers to dB attenuation with
and SSB gear with the “right” test equip- Damm, WA7MLH, and Kurt Knoblock, respect to a specific carrier.
ment available. We have built special, nar- WK7Q, built versions of the analyzer and 17Formally, this is the transducer gain, or 50-Ω
row-tuning-range analyzers to examine have garnered several years of use with them. insertion gain. There are many different pa-
transmitter sideband suppression and distor- Their experiences have been of great value in rameters that are called “gain.”
tion. The equipment uses the same concepts our efforts. Barrie Gilbert of Analog Devices 18The signal sources used are updated ver-
presented. Northwest Labs suggested the AD8307 log sions of the circuits shown in Fig 66, p 168, in
amplifier and provided samples and early the Note 15 referent.
There are many ways that this instrument data needed for evaluation measurements. 19See page 154, Solid-State Design for the
can grow. One builder has already bread- Many of our colleagues within the Wireless Radio Amateur.
boarded a tracking generator. (Let’s see that Communications Division at TriQuint Semi- 20A 20-dB return loss corresponds to a voltage
in QST!—Ed.) Many builders will want to conductor have helped us with filter mea- reflection coefficient of 0.1, or an SWR of
interface the analyzer with a computer in- surements: Thanks go to George Steen and to 1.222. See Wes Hayward, W7ZOI, Introduc-
stead of an oscilloscope. A recent QST net- Don Knotts, W7HJS. Finally, special thanks tion to RF Design (Newington: ARRL, 1994),
work analyzer paper suggests circuits that go to our colleague in the receiver group at p 120.
may provide such a solution.21 TriQuint, Rick Campbell, KK7B, who pro- 21See, for example, Steven Hageman, “Build
vided numerous enlightening discussions Your Own Network Analyzer,” QST, Jan
A recent QST summary of WRC9722 out- and suggestions regarding the preparation of 1998, pp 39-45; Part 2, Feb 1998, pp 35-39.
lines new specifications regarding spurious 22Larry Price, W4RA, and Paul Rinaldo, W4RI,
emissions from amateur transmitters. Gener- “WRC97, An Amateur Radio Perspective,”
QST, Feb 1998, pp 31-34. See also Rick
40 September 1998 Campbell, KK7B, “Unwanted Emissions
Comments,” Technical Correspondence,
QST, Jun 1998, pp 61-62.

◊ Please refer to Wes H
Terry White, K7TAU,
lyzer for the Radio Am
Sep 1998, p 40, Fig 16
resistors for the 20-dB
units, not 820 Ω as or
Wes Hayward, W7ZOI

Hayward, W7ZOI, and
, “A Spectrum Ana-
mateur—Part 2, QST,
6. The six 2-W input
pad should be 620 Ω
iginally specified.—
(tnx EA2SN)