Wideband QRP VSWR Meter
Logarithmic detectors and simple coupler work from 1- 500 MHz at
very long time ago, I read an article in some amateur radio publication that
explained how performance problems in VSWR meters relate to the nature of diode
detectors at different power levels. At
low power levels, these diodes are square-law detectors, so that output voltage
varies linearly with applied power, but at high levels these diodes are
rectifiers, so that output voltage varies linearly with applied voltage or the
square-root of applied power. I wish I
could find that article again.
of us learned about logarithmic RF detectors from the 2001 QST article that featured
the Analog Devices AD8307. I immediately built several pieces of test
equipment that included the AD8307 and similar devices, as did many other folks,
as evidenced by the flurry of articles.
occurred to me in 2008, that a VSWR meter with a pair of logarithmic detectors
in place of the diode detectors would overcome the problems inherent in the
diode detectors. As a bonus, a
subtraction replaces the division in the calculation that compares the forward
indicating voltage to the reflecting indicating voltage because we are now
dealing with logarithms. I actually
wrote up an invention disclosure for such an instrument before discovering that
someone else had already patented this apparently obvious idea in 2002. Too bad.
Look for US
QST article by Kaune in 2011 featured dual logarithmic detectors in a VSWR
and led me to check whether the 2002 patent was still in force. If a patent proves to be commercially
unsuccessful, patent owners often choose not to pay the renewal fees, which
increase as the patent matures. The
first fee is due after year 7. I went to
the US Patent and Trademark Office Patent Application Information Retrieval web
learn the status of patent 6486679 and was gratified to see the message: “Patent
Expired Due to Nonpayment of Maintenance Fees Under 37 CFR 1.362”. That’s Chapter 37 of the Code of Federal
Regulations, whatever that is. This
patent expired in December of 2010, the month before the Kaune article, so we
are all free to do as we choose with this concept. I was thrilled. This led me to reconsider the approach and
what I wanted in a VSWR meter.
I wanted was a VSWR Meter suitable for QRP that gives consistent measurements
as power level varies. Most instruments
cannot handle the low power levels inherent in QRP operation, but the excellent
sensitivity and huge dynamic range of the AD8307 logarithmic detector allows
operation at low RF power levels, even with short couplers that have very small
coupling coefficients at low frequencies.
I started by adding a pair of AD8307 logarithmic detectors to an
existing coupler in an old VSWR meter and achieved an early success. It turned out that simple transmission line
couplers work up to very high frequency if they aren’t too long. The limit is ¼ wavelengths at the highest
frequency. Such couplers are what you
see in old Heathkit and citizen’s band VSWR meters. They have no ferrites to impose frequency
limitations. They just have length,
distance between coupled lines, and characteristic impedances of the main and
coupled transmission lines.
decided to use printed circuit techniques in order to make couplers with
balanced and controllable characteristics, so I could reproduce them. This turned out to introduce interesting new
problems, but it’s all about learning, isn’t it? The lesson below is that simple microstrip
couplers do not work well, because modes in the air above the board have
different propagation characteristics than modes in the dielectric within the
board. Luckily, the fix is not
The Directional Coupler
Figure 1 shows the bare printed circuit
board that hosts the logarithmic detector circuits and constitutes the forward
and reverse directional couplers. The transmitter
enters the main line in the center from an SMA connector on the left to the
antenna or other load via an SMA connector on the right. I designed the prototype to use SMA
connectors to learn what frequency response I could achieve. The results are truly remarkable!
main line conveys RF power from the transmitter on the left to the antenna on
the right. The main line couples a
sample of the forward signal (from the transmitter to the antenna) to the upper
line and then to the logarithmic detector at the upper left. A matched termination at the other end in the
upper right on the rear absorbs and does not reflect any contribution from the
opposite direction. Similarly, the main
line couples a sample of the reverse signal (from the antenna back to the
transmitter) to the lower line and then to the logarithmic detector at the
lower right. A matched termination at
the other end in the lower left on the rear absorbs and does not reflect any
contribution from the opposite direction.
line couplers should be embedded in a uniform dielectric such as air or FR-4
printed circuit material, so the Heathkit style coupler works and stripline
couplers work. Stripline is printed
conductors embedded in dielectric between two ground planes; whereas microstrip
is printed conductors above a dielectric with a ground plane below the
dielectric only. Microstrip does not
make good transmission line couplers.
explains: “The λ/4 coupled line design
is good for coaxial and stripline implementations but does not work so well in
the now popular microstrip format, although designs do exist. The reason for
this is that microstrip is not a homogeneous medium – there are two different
mediums above and below the transmission strip. This leads to transmission
modes other than the usual TEM mode found in conductive circuits.”
with minor edits, says: “Planar
structures (unless they are stripline) have notoriously bad directivity.
Directivity (Isolation minus coupling) is determined in these types of
structures by the difference between the even and odd mode phase velocities
with the best coming when these are equal. In microstrip, the odd mode is
mostly in dielectric, and the even mode is mostly in air. To equalize the phase
velocities, you need to slow down the even mode which can be accomplished using a dielectric overlay over the lines
3rd reference also
suggests the method of adding dielectric material above a microstrip coupler to
make it work, and that’s what I did. The
dielectric above the microstrip should be the same as the material below for
the even and odd modes to propagate at the same speed. We call this configuration of microstrip within
a thicker layer of dielectric “embedded microstrip.” Figure
2 shows that it looks like stripline without the top ground plane.
very critical to control the impedances of the main and coupled lines in order to
minimize degradations to directivity. The
coupled lines can be of any impedance as long as they terminate in their
characteristic impedance. The main line
must be the characteristic impedance of the transmission line between your
transmitter and antenna or tuner, usually 50Ω.
I tried to make all three be 50Ω.
found four impedance calculators for embedded microstrip transmission lines on the
web. Two agree closely, and the other
two disagree dramatically. Unfortunately,
I started with a bad one, but I got excellent results with one from the Swedish
company Multi-Teknik. I designed and had ExpressPCB fabricate the
board using 0.080 inch traces with 0.070 inch gaps between the main and coupled
lines using the ExpressPCB parameters for dielectric constant and trace thickness.
Figure 3 shows the first working QRP VSWR
Meter. Note that the dielectric overlay
converts the ordinary microstrip to embedded microstrip. SMA connectors that connect to the edge of a
printed circuit board prove to be a convenient way to attach to the center main
line trace for this prototype.
Analog Devices AD8307 data sheet
describes the operation and use of this logarithmic detector. I built a test jig following figure 37 in
that data sheet to characterize these devices to assure that forward and
reverse measurements are on the same basis.
My method sets all devices to a slope of 0.025 Volts/dB and an intercept
of -86 dBm. My settings are only as
accurate as my power measurement, but they are consistent across devices and
allow good matching. I could measure the
characteristics and compensate in software as I did in my DDS project,
but this time I measured and installed the shunt resistances necessary to
achieve the desired slope and used trim-pots to set and match the intercepts.
curved lines in Figure 4 show how well 8 sample AD8307s agree following
the calibration process and indicate that their usable dynamic range is from
+10 dBm down to -74 dBm. Measurement
results within this range are very reproducible. It is this huge dynamic range that allows the
VSWR Meter to accommodate the low coupling coefficients and corresponding low
signal levels at low frequencies as well as the high coupling coefficients and
high signal levels at high frequencies. The
straight lines in this plot are the best fits to the linear portions of the
real curved line data between 0 dBm on the high side and -60 dBm on the low
side, well within the linear dynamic range.
These straight lines serve to determine the slopes and intercepts that
best fit the data.
circuit of data sheet figure 37 also serves as the circuit for the two
logarithmic detectors on the printed card except that I add a 0.1 μF capacitor
from pin 3 to ground in case this instrument ever finds use at frequencies
below 100 kHz. Figure 5 shows the schematic of the bi-directional
coupler/logarithmic detector board. In
later versions, the ‘select in test’ resistors R5 and R12 became 100kΩ trimpots
that set the slope of the logarithmic detector response to 0.025 V/dB. Similarly, resistors R4 and R6, and trimpots
R7 and R14 became 100kΩ trimpots that set the zero power input intercept to
-86dBm. It is critical that both
logarithmic detectors have precisely the same slope and intercept values.
The Math - With thanks to
Glenn Pederson WB9QIQ for bringing the previous error to my attention
The outputs of the coupler board
are two voltages that indicate the logarithms of the forward and reflected
powers from the two couplers.
The linearized logarithmic
detector response from Figure 4 is
V = (0.025 volts/decibel x 10 log Power in
dBm) + 2.15V (1)
And the reflection coefficient
traditionally represented as the Greek character rho is a ratio of reflected
signal level compared to forward signal level.
In terms of reflected and forward power, and remembering that dividing
numbers is the same as subtracting their logarithms,
ρpower = Pr/Pf or
ρpower = 10 log Pr - 10log Pf
VSWR meters measure voltages
corresponding to the reflected and forward powers to calculate reflection
coefficient and then VSWR. The simplest
way I can explain is to start with the difference between these voltages which
is a ratio of the reflected power to the forward power, which is in turn, the
power reflection coefficient ρpower. So,
(Vr – Vf) = [0.025 x (10 log Pr) + 2.15 -
0.025 x (10 log Pf ) - 2.15] (3)
This conveniently simplifies to:
(Vr – Vf)
= (10 log Pr – 10 log Pf) / 40 (4)
(Vr – Vf) = (log Pr/ Pf) / 4 = (log ρpower)/4 or (5)
log (ρpower) = 4 x (Vr –
Take the square root of the power
reflection coefficient to get the voltage reflection coefficient ρvoltage, because we ultimately want the Voltage Standing Wave
Ratio. That’s the “V” in VSWR. Since we’re still dealing with
logarithms, we obtain the square root by dividing the logarithm of the power
reflection coefficient by 2.
log (ρvoltage) = log (ρpower) / 2 = 2 x (Vr – Vf) (7)
So far, we’ve taken the two
voltages, subtracted them, and multiplied by 2. Then we obtain ρ by raising
10 to the power log ρvoltage.
Finally we obtain the VSWR from
the formula VSWR = (1 + ρvoltage)/(1 - ρvoltage). All this is just a few lines of code in a high
level language, as we’ll see.
did all the early development work with a pair of digital multimeters and a
calculator to prove the principle. When
I finally laid out this board and ordered three from ExpressPCB for
about $63 including shipping, progress came rapidly. I had a working board and nothing else ready
and just the idea that I should use a computer and some analog to digital
this point, I first considered the Arduino.
It was sitting on top of my oscilloscope, waiting to do something
useful. I had never done anything with
it except to change the default ‘BLINK’ program that blinks the LED on and off to
instead send CQ on the LED. In short
order, I learned that the Arduino has several analog inputs and that the
programming language supports raising one number to the power of another. That was all I needed besides a display. Doug and Ben, the experts at Gateway
explained that a sample 2 line by 16 character black on green LCD display in my
junk box would serve. I am very lucky to
live near to Gateway Electronics and usually visit daily.
Arduino does an amazing job of unburdening your tasks. I expected to spend a week interfacing some
kind of display and learning to twiddle the right bits to make it work. It turns out that the Arduino directs you to
references that show you how to connect displays it understands and already has
the programming library built in.
Similarly, the analog to digital converters operate with simple program
commands. Figure 6 shows the simple code that performs the basic functions of
the VSWR Meter. This is all the code
from the original working VSWR Meter, not just an excerpt.
bits of resolution would be nice, but 9 to 10 is adequate. I reduced the analog to digital converter
reference voltage AREF from 5 V to 2.4 V to get that last bit of resolution. See the sidebar on Arduino Uno ADC
If this were a
commercial product, I might select another processor that consumes less power
or has higher A/D converter resolution, but the Arduino is excellent for rapid
prototyping. I ordered several more.
Considerations for Portable Operation
prototype in Figure 3 draws 70 mA from a 5V supply for the Arduino, the display,
and the logarithmic detectors. The
backlight for my junk-box display draws another 160 mA, but this display works
fine without the backlight in suitable ambient illumination.
NØWL suggested I optimize for battery operation by turning the Arduino off
between measurements rather than using a delay loop to set the timing. This didn’t help as much as I expected,
because there is other circuitry on the Arduino including a second processor to
handle USB communication to the PC, and this other circuitry stays active and
consumes over 30 mA.
solution is to remove the ATMega328P-PU chip from the Arduino and build
suitable circuitry around it. Figure 7 shows the prototype CPU board
is similar in layout to the Arduino. Some
folks call this a “Pseuduino.” Along the
way, I purchased more such chips from DigiKey and learned to program them with
a bootloader,. Then I needed to program them remotely from the
Arduino board. Ben, one of the experts
at Gateway Electronics, taught me how to program an ATMega328P-PU chip remotely
from an Arduino Uno with the chip removed.
I find only one reference
on-line with only verbal instruction, so I provide a second sidebar to illustrate
this method of remote programming. Elimination
of the excess circuitry reduced the current drain of the ATMega328P-P, which
still ran full time, to about 16 mA. That
means the USB interface and other circuitry drew over 30 mA.
built a first Pseuduino to replace the Arduino in the VSWR Meter and a second
for use as a fixture to aid in assembling and calibrating these instruments in
a production environment. I eventually
turned an Arduino into a Pseuduino so that I wouldn’t have to build any
more. I describe this procedure in a
third sidebar below.
found an excellent tutorial on how to put the CPU of an Arduino into various power
saving modes when it isn’t necessary to be active. The ATMega328P-PU chip draws about 2.5 mA in
the lowest power sleep mode, including 1 mA for the reference diode. With the display energized full time, CPU and
display together consume 5 mA in sleep mode, which lasts for about 90% of the
added circuitry to turn on the remaining high current circuitry when
necessary. The Logarithmic Detectors
consume about 15 mA but only need to be on for 30 ms prior to the two
measurements by the A/D converter. The
backlight is optional, but I made it light up for about 100 ms after each
display update with a repeating period of 500 ms or for 20% of the time for
battery operation. Under these
conditions, current drain is under 10 mA without the backlight and 33 mA with
the backlight pulsing or blinking after each measurement.
I found a white on blue LCD display at Gateway Electronics with a very much
lower backlight current, but which doesn’t operate at all without a
backlight. This backlight requires much
less current than the ATMega328P-PU chip, so it made sense to run the backlight
constantly rather than running the CPU to blink the backlight.
Figure 8 shows the supply current as a
function of time for the high and low backlight current displays. The bottom line in each is the zero reference.
Sensitivity is 10 mA per division in Figure 8a (1 mV per division divided by
the 0.1Ω current sense resistor). Current
rises from 8 mA in sleep to 37 mA when the logarithmic detectors turn on and 22
mA when they turn off and the CPU chip performs the measurement and
calculations. Finally the backlight
blinks on raising the peak current to 70 mA for 100 ms. Average current was under 25 mA for the display
with the high current backlight in this energy saving mode, but the blinking
was annoying. Sensitivity is 5 mA per
division in Figure 8b (500 µV per
division divided by 0.1Ω). Current rises
from 5 mA in sleep to 30mA when the logarithmic detectors turn on and the CPU
chip performs the measurement and calculations.
Final current drain with the backlight is now 7.1 mA, well suited to
battery operation. The 5 mA sleep
current includes the ATMega328P-PU chip, the display, the backlight, and current
through the reference diode that supplies AREF.
Yes, I could save another 1 mA by applying power to the reference diode
only during measurements. Perhaps I
Figure 9 shows the final prototype with
the backlit 2 line by 16 character white on blue LCD display.
Figure 10 shows the innards of the
final prototype. I have since added
nickel- metal hydride AAA cells that fill some open space in the cover and
charging circuitry. A relay disconnects
the battery when power is available from the wall-wart. This proved simpler that the alternative
semiconductor circuits I tried. The
wall-wart charges the nickel metal hydride cells through a constant current
also tried a 2 line by 40 character black on green LCD display provided by Herb
AF4JF. It has no backlight and requires
suitable ambient illumination. This
enabled sufficient characters to implement a bar graph for analog display. OK1DX used this approach,
and it looks neat.
placed an actual analog meter on a PWM output for a true analog display. This could lower the cost of parts, but
current drain is much higher, because the CPU chip must stay active and never
enter a low power sleep mode. Still,
this enables a familiar interface except that the operator no longer needs to
set the full scale reading in the forward direction before switching to obtain
a reading in the reverse direction. The
correct reading just appears at the meter with no switching and no complex cross-needle
Figure 11 shows the accuracy of this instrument
in measuring VSWR over a wide range of load resistance at a sub-QRP power level. I performed these measurements as best I
could by placing small resistors with the shortest possible leads within a coax
connector. With a 7Ω termination, the
ideal VSWR should be 7 and the instrument measures between 5 and 6 over the
entire frequency range. Similarly, with
a load resistor of 500Ω, the ideal VSWR should be 10 and the instrument
measures between 5.7 and 9.1 over the entire frequency range.
attempts to make this unit more readily manufacturable led me to simulate a
true stripline version by replacing the unclad FR-4 overlay with one that uses
a single sided copper clad FR-4 overlay.
This required adjusting the trace widths. I wasn’t as lucky this time, and the mainline
impedance measures 56Ω. This limits the minimum
measureable VSWR in a 50Ω system to about 1.2, but the exercise shows that
stripline works, and we are presently trying to find someone to manufacture
such boards inexpensively.
I received several requests to incorporate power measurement. When I finally lashed a third logarithmic
detector onto the simulated stripline board, this proved to be a surprisingly
simple upgrade. The first try at power
measurement was flat within ± 0.3 dB from 1 to 30 MHz. With some attention to equalization, the
third try was flat within ± 0.3 dB to above 150 MHz.
there any drawbacks or problems with measuring VSWR at such low power
levels? You bet! It turns out that when you connect to a real
antenna instead of a dummy termination, the reverse log detector senses signals
from the antenna that may dominate the small return from a low VSWR antenna and
erroneously raise the apparent VSWR. For
this reason, I will consider shortening future versions of the sampler to make
them less sensitive. Slightly increasing
the spacing between the main transmission line and the sampling transmission
lines will yield the same result. For
now, it is best to use this instrument well above the lowest power levels.
NØVI and I are trying to turn this into a product or an SLQS
club project, if we can produce accurate couplers at low cost. The main transmission line impedance is
critical to performance. An air gap
under the FR-4 layer would degrade performance, so we want to make the coupler
in stripline rather than embedded microstrip.
of a kit would require installation of at least four surface mount chip
resistors and six surface mount chip capacitors, more if we include power
measurement. I characterize and match
the logarithmic detectors to optimize performance and recently developed a
simpler method to perform this previously burdensome task. For these reasons, the coupler should
probably be provided as a complete calibrated assembly.
Lee’s website to
learn whether a product became available.
instrument enables fairly accurate VSWR measurements for really low power
operation over an incredible frequency range from below 160 meters to above ¾
meters. Accuracy degrades gracefully as
power decreases. Results are acceptable
at the lowest HF frequencies down to 30 mW!
With slightly more power, this instrument proves useful at 100 kHz and
directly to the SMA connectors on the prototype coupler are accurate up through
UHF. With added adapters and coaxial
cable to convert to BNC connectors, performance is still very accurate up
through VHF and more than usable to 500 MHz.
prototype works over the frequency range from 1-500 MHz at less than 100 mW and
accommodates 100 Watts up to 15 meters, 70 Watts up to 10 Meters, and less with
increasingly higher frequencies as coupler efficiency increases. Increasing the spacing between the main
transmission line and the sampling transmission lines or shortening these lines
would allow higher power at higher frequency at the expense of less sensitivity
at the lowest frequencies.
was an incredibly enjoyable and rewarding project. I have used personal computers to control
circuits in other recent projects that appeared in QEX, but this was my first
attempt at embedded programming since I spent nearly 2 years programming a DSP
in assembly language in 1992-1993. I had
the Arduino programmed and running in 2 days.
What a kick!
Dr Sam Green, WØPCE, is a retired aerospace engineer.
Sam lives in Saint Louis, Missouri. He holds degrees in Electronic
Engineering from Northwestern University and the University
of Illinois at Urbana. Sam specialized in free space optical
and fiber optical data communications and photonics. He became KN9KEQ and K9KEQ
in 1957, while a high school freshman in Skokie,
Illinois, where he was a Skokie
Six Meter Indian. Sam held a Technician class license for 36 years before
finally upgrading to Amateur Extra Class in 1993. He is a member of ARRL, a
member of the Boeing Employees Amateur Radio Society (BEARS), a member of the Saint Louis QRP Society (SLQS), and breakfasts with the Saint Louis Area
Microwave Society (SLAMS). Sam is a Registered Professional Engineer in Missouri and a life
senior member of IEEE. Sam holds seventeen patents, with one more patent
application pending. Contact Sam at firstname.lastname@example.org.
the TX, RX, RESET, and GND lines of the remote ATMega328P-PU chip and the Arduino
Uno without the ATMega328P-PU chip installed as in Figure S2. Either connect 5V
or let the circuit where the remote ATMega328P-PU chip is running provide 5V. Then simply program sketches into the
ATMega328P-PU chip as if it were still plugged into the Arduino Uno. This does not work unless the crystal is 16.0
MHz, as on the Arduino Uno. Remove the
RESET line if you remove Arduino power to allow the remote ATMega328P-PU chip
first built this instrument around an Arduino.
In order to operate efficiently from battery power for portable use, I
built a copy with minimal current drain.
I eliminated all functions other than the ATmega328P-PU and the external
reference supply. Folks call such an
Arduino copy a Pseuduino.
simplify the process of making a Pseuduino, I attempted to turn an Arduino into
started with the removal of the surface mount ATmega16U2-MU. This reduced the supply current by about 20
I removed the yellow LED next to the “L” and the green LED next to the
“ON”. This further reduced current drain
by about 8 mA.
I removed the NCP1117ST50T3G 5.0 V LDO regulator next to the external power
connector, which draws 3.4 mA, even with power supplied from the output side.
LP2985-33DBVR 3.3V LDO regulator drew less than 100 µA, so I reinstalled it. The Sparkfun hot air rework station
made removal of these extra parts relatively simple.
the ATmega328P-PU CPU, the supply current for this Pseuduino implementation is now
less than 550 µA. I replaced the
ATmega328P-PU CPU after taking the photo.
program the ATMega328P-PU Chip remotely in a Pseuduino as in the second
Sidebar, simply connect the 5 pins on the unmodified Arduino with the ATMega328P-PU
Chip removed to those same 5 pins on this Pseuduino which now hosts the remote