Rectangular Waveguide to

Coax Transition Design

Learn how to find the optimum dimensions for a waveguide to
coax transition using an empirical approach that relies on a set
of impedance measurements and a few calculations.
Paul Wade, W1GHZ

question I am frequently asked is,

Why do the antenna dimensions
in the W1GHZ Microwave Antenna
Book Online not include the probe
dimensions (for the transition from
waveguide to coaxial transmission line)?1
The answer is that the transition is part of
the transmission line, not the antenna, and
does not directly affect the performance of
an antenna. The transition may be right at
the antenna, seemingly part of it, or at the
other end of a run of waveguide transmission
line, many meters away.
The transition is an important part of most
microwave systems, however, since solidstate components are usually constructed on
microstrip transmission lines and interconnected with coax, while microwave antennas normally use waveguide techniques.
A typical transition consists of a coaxial
connector on the broad side of a rectangular
waveguide with the center conductor extended as a probe into the waveguide, with
one end of the guide ending in a short circuit, like Figure 1. Since the structure is wellknown, design may be too ambitious a term,
but the correct dimensions are far from obvious and are difficult to calculate. A number
of sets of dimensions have been published,
but there is little agreement between them,
so it is difficult to tell which are right. Also,
many of the published transitions are part of
an antenna, so the dimensions may have been
chosen to compensate for a poorly matched
antenna impedance.
Therefore, I chose an empirical approach:
making a comprehensive set of measurements
from which the optimum dimensions may be
reached. I had previously used this technique
1Notes

to determine probe dimensions for circular

waveguide made from copper water pipe.2 It
is easier to determine the dimensions for rectangular waveguide, since commercial guide
and components are readily available from
surplus sources.
Characteristics of a transition are best
viewed by measuring the complex impedance (magnitude and phase) in the
waveguide, using a waveguide slotted line.
Figure 2 shows a typical X-band slotted
line the precision impedance measurement instrument of a few years ago, and,
for waveguide, still more accurate than
most network analyzer measurements.
Since all professional microwave work
today uses automatic network analyzers
and computers, and few remember how to
use a slotted line, slotted lines are almost
given away today. I have paid as little as
$2 for one at a hamfest.
To find the optimum dimensions for a transition, I needed to make measurements over a
range of transition dimensions, so an adjust-

able transition was desirable. The probe dimensions are readily varied by unscrewing the coax
connector and trimming the probe, but the distance to the shorted end of the waveguide, or
backshort, must also be varied. I machined a
sliding plug to fit inside the guide, with alternating quarter-wave sections of high and low
impedance to form an electrical short circuit,
so that the performance of the short does not
depend on intimate contact with the waveguide
walls. Figure 3 is a photograph of adjustable
transitions for two common sizes of X-band
waveguide, WR-90 and WR-75, as well as one
for circular waveguide.
Measurement Technique
The first measurement is with a short circuit (flat metal plate) closing the end of the
slotted line. The short provides a clear standing-wave pattern with sharp nulls at halfwavelength intervals, so we can measure the
guide wavelength, and make any adjustments
to the slotted line measuring probe.
The next measurement is of the sliding tran-

appear on page 16.

161 Center Rd
Shirley, MA 01464
[email protected]

Figure 1 WR-75 waveguide to coax transition for 10 GHz.

10 Nov/Dec 2006

wade.pmd

10

10/03/2006, 2:16 PM

sition with no transition probe installed, with

the face of the sliding short at the centerline of
the coax connector. The location of the
standing-wave nulls is recorded as the phase
reference for the impedance measurements.
Finally, we are ready to measure the
transition. A probe is installed in the
transition, and a precision 50- termination
installed on the coax connector. Then the
sliding short is moved to a series of locations
and the impedance recorded at each location.
(Actually, the SWR and the location of a null
are recorded, and the complex impedance
calculated later, using either a Smith Chart
or a computer.) Then the probe is removed
and replaced with a different one, and the
measurement sequence repeated.
In my previous experiments with circular
waveguide, I had found that the diameter of
an SMA inner conductor, 1.27 mm, was in
the optimum diameter range for X-band transitions, so I chose to limit the number of
measurements by using only this diameter
for rectangular waveguide at 10 GHz. With
a single probe diameter, only the probe length
is varied, so changes are easy, starting with a
long probe and cutting off various increments.
Many commercial transitions have the
probe surrounded by a cylinder of dielectric,
so that the Teflon dielectric of the SMA connector continues for the full height of the
waveguide. Since it is difficult to shorten the
probe length inside the dielectric, my preferred transitions have the Teflon dielectric
of the SMA connector ending at the inner
wall of the waveguide and only a bare probe
extending into the guide. I did, however, want
to compare the performance of bare-probe
transitions with full-height-dielectric ones, so

I carefully sliced the protruding dielectric

from a connector, and then fit it back together
for the measurements after each adjustment
of the probe length.
Measurement Results
Without the slotted line or an expensive
network analyzer, only SWR could be easily
measured. Figure 4 shows a plot of the SWR
versus probe position and diameter in WR90 waveguide at 10.368 GHz. We see that
there is a set of dimensions that produce an
excellent SWR, and a number of other combinations that produce a usable SWR < 2:1.
What isnt clear is how we reached the excellent dimensions, or whether there are other
excellent combinations that we did not find.
A plot of complex impedance on a Smith

Chart is much more meaningful. Each curve

in Figure 5 is for a constant probe length,
with varying backshort distance. The
backshort forms a section of shorted transmission line behind the probe, creating a reactance at the probe that varies with the length
of the line to the backshort. At some distance,
this reactance is equal to the reactance of the
probe but with conjugate phase, canceling the
reactance so that only the resistive component of the impedance remains. This is where
the curves cross the horizontal centerline, or
real axis, of the Smith Chart.
The resistive component where each
curve crosses the real axis varies with probe
length we can see that the curve for a short
probe crosses the real axis to the right of
center, at a higher resistive impedance than

Figure 3 Adjustable transitions for WR-90, WR-75 and circular waveguide.

Figure 2 A surplus waveguide slotted

line.

Figure 4 WR-90 to coax transition, SWR versus backshort distance at 10.368 GHz.

Nov/Dec 2006 11

wade.pmd

11

10/02/2006, 4:58 PM

Z0. Conversely, a longer probe crosses to the

left of center, so the resistive impedance is
lower than Z0. The relationship is clear and
monotonic, so we can be confident that there
is a single optimum set of dimensions that
produces the best impedance match, which
is in the center of the Smith Chart. Once we
have a set of curves that bracket the center,
we can interpolate to estimate the optimum
probe length, then make more measurements
to locate the backshort position for a perfect
match see the data point in Figure 5
marked 0.146 .
Discussion
The probe in the waveguide acts as an
impedance transformer from the low
impedance of the coaxial line, typically
50 , to the high impedance of the waveguide,
typically greater than the impedance of free
space = 377 . Since there is no loop for
current to flow in and create a magnetic field,
the probe must be an electric field transformer.
The maximum electric field is at the center
of the wide dimension of the waveguide,
where we have placed the probe. Intuitively,
we suspect that a longer probe intersects more
of the electric field and thus couples more
tightly. We saw this in Figure 5 the longer
probe produces a lower impedance in the
waveguide, because it is coupled more tightly
to the low impedance in the coax.
Adding a dielectric around the probe
would tend to concentrate the electric field
near the probe and increase coupling, so a
shorter probe would be required. We can see
this in Figure 6 the same WR-90
waveguide as Figure 5, but with the Teflon
dielectric of the SMA connector extended to
the far wall of the guide. We see the same
family of curves, but the impedance is
matched with a shorter probe length.
There seems to be a persistent myth,
found in older microwave books, that the
backshort should be exactly /4 from the
probe. We know that a shorted quarterwavelength of transmission line acts an open
circuit, so this distance would add no
reactance at the probe; this distance would
only work for an ideal probe, with no
inductance or capacitance of its own. The
probe is a wire, however, so it must have
inductance, in addition to capacitance to the
waveguide walls. We can see in the Smith
Charts that the backshort distance required
to achieve a pure resistive impedance varies
slightly with probe length, as we would
expect if it were tuning out the reactance of
the probe. More to the point, Figure 4 shows
that with the backshort /4 from the probe,
the best SWR achieved is 2:1.
Since the backshort distance is a reactive
phenomenon, it must be frequency dependent.
We can see how critical by measuring the SWR
versus frequency. I fabricated fixed transitions

to the best dimensions, one with a bare probe

and the other with extended Teflon dielectric.
In Figure 7, we can see that the WR-90 transitions have good bandwidth, with excellent
SWR from 10 to 11 GHz, and pretty good from
9.7 to 13 GHz. The transition with the Teflon
dielectric shows slightly better bandwidth than
the one with a bare probe, but not a significant
improvement both are more than adequate
to cover the whole amateur band. Figure 8
shows the performance of the two WR-90 transitions connected together, with extremely low
insertion loss from 10 to 13 GHz.
Other Waveguides
Scaling the WR-90 transition dimensions
to other frequencies and waveguide sizes is
more difficult in rectangular waveguide than
it was with circular waveguide, since other
standard waveguides have different rectangular aspect ratios. In circular waveguide,
the characteristic impedance is a function

of the guide wavelength:

 Og 

Z 0 377 x 
 O 
 0 

while in rectangular waveguide, the characteristic impedance is modified by the aspect

ratio:
 Og  2 b
x
Z 0 377 x 
(Eq 2)
O  a
 0 
where a and b are the large and small inner
dimensions, respectively. The guide
wavelength, g, is easily measured with the
slotted line, but it can also be calculated:

Og

12

O
O 
1  
 Oc 

(Eq 3)

Figure 5 Smith Chart plot of impedance as dimensions are adjusted for WR-90 to coax
transition.

12 Nov/Dec 2006

wade.pmd

(Eq 1)

10/03/2006, 9:55 AM

where the cutoff wavelength, c = 2a, or

twice the large inner dimension of the
waveguide.
Thus, in different size waveguides, we
are transforming to different waveguide
characteristic impedances. Another good
X-band waveguide is WR-75; the aspect
ratio b/a is 0.5, while for WR-90 the aspect ratio is 0.444.
Since g is larger in the smaller guide
as well, the characteristic impedance of
WR-75 at 10.368 GHz is significantly
higher than the Z0 of WR-90.
Thus, we would expect the transition dimensions for WR-75 to also be different.
In Figure 9, the impedance curves measured
for an adjustable transition are shown; we
see the same family of curves as previously,
but a shorter probe is required to match the
higher impedance. I did not take a full set
of measurements with the extended Teflon
dielectric in WR-75, just enough to find approximate matched dimensions. In
Figure 10, we can see that this shortcut was
a mistake the best probe length would
be slightly shorter than the one measured.
This is a problem with making a lot of measurements before doing the calculations
necessary to review the results.
I fabricated two WR-75 fixed transitions, one with a bare probe and the other
with extended Teflon dielectric. Figure 11
shows that there is no appreciable difference in bandwidth both have good SWR
from 10 to 10.5 GHz but narrower
bandwidth than the WR-90 transitions.
In Figure 12, the insertion loss is very
small from 10 to 11 GHz, and low from 9
to 16 GHz.
The recommended operating range for
WR-75 is 10 to 15 GHz, while the range
for WR-90 is 8 to 12.4 GHz. It is probable
that if we were to find WR-75 transition dimensions at a frequency near the center of
the recommended range, it would have a
wider bandwidth. It would probably not be
as good in the 10 GHz amateur band, however. Many surplus WR-75 transitions suffer from this deficiency they work best
around 12 to 14 GHz so we are probably
better off fabricating our own transitions
optimized for 10 GHz.
Other Frequencies
We have determined optimum waveguide
to coax transition dimensions for 10 GHz
empirically. These dimensions are listed in
Table 1, and a sketch of a transition is shown
in Figure 13. The 10 GHz dimensions are
not directly scalable to other waveguide sizes
and frequencies, but we have demonstrated
a technique applicable to other waveguide
sizes and frequencies.
What can we do if we do not have a
waveguide slotted line suitable for the target

Figure 6 Smith Chart plot of impedance as dimensions are adjusted for WR-90 to coax
transition with full height Teflon.

Figure 7 SWR versus frequency of WR-90 to coax transition.

Nov/Dec 2006 13

wade.pmd

13

10/02/2006, 4:58 PM

frequency? Can we make a transition if all

we can measure is SWR or return loss? I think
we can come pretty close, even without a
sliding short. Since we have to drill a hole for
the coax connector, we must first estimate
the backshort distance, from the hole to the
shorted end. For the WR-90 transitions, the
backshort distance is 0.146 g, while for the
WR-75 transitions, the distance is 0.118 g.
Looking at Figures 5, 6, 9, and 10, we can
see that a compromise distance of 0.125 g,
splitting the difference, will give us an SWR
of 1.3 or better with the right probe length.
So, if we put the coax connector about oneeighth of a guide wavelength from the
shorted end and trim the probe length for
best SWR, we can probably get a pretty
decent transition in any size rectangular
waveguide.
Figure 14 is a plot of SWR versus probe
length for different backshort spacings, in
1
/16 g increments, for the WR-90 transition
with the Teflon dielectric extending the full
height of the waveguide. The 0.125 g
spacing gives the best SWR. If we start with
a long probe and trim, the SWR will
decrease slowly, then increase rapidly when
we pass the optimum length. Therefore, we
should stop shortening as soon as we see
an increase in SWR. One final note: the
SWR for the /4 spacing recommended in
the old books never gets below about 2
still usable, but not very good.
Figure 14 shows that a fairly wide range
of dimensions will result in a reasonable
SWR, less than 2 or 3, which can be easily tuned out with some screws in the
waveguide. Perhaps this is why some published transitions include tuning screws.
Without good test equipment, though, it
is just as easy to make it worse as it is to
make it better. Many microwave operators
are able to get on the air successfully with
minimal test equipment, so any necessity
for tuning creates an additional hurdle for
beginners.
Software
Sometime after I finished the WR-90
and WR-75 transitions, I started a new job.
Part of the new job involves electromagnetic simulations, and I had to learn to use
some new software. One of the simulation
programs, Ansoft HFSS, is well suited to
waveguide calculations. 3 Since I needed
some simple problems as learning exercises, I tried simulating some transitions,
using the same empirical technique, but
calculating the impedance with the software
rather than measuring it.
After simulating my WR-90 and WR-75
dimensions to make sure the two techniques
yield the same results, I moved to WR-42
waveguide at 24.192 GHz. The impedances
were plotted on a Smith Chart until they

Figure 8 Insertion loss and SWR of two WR-90 to coax transitions back-to-back.

Figure 9 Smith Chart plot of impedance as dimensions are adjusted for WR-75 to coax
transition.

14 Nov/Dec 2006

wade.pmd

14

10/02/2006, 4:59 PM

converged on optimum dimensions. Since

the software calculated the impedance directly, it was easy to plot each point immediately and move quickly toward the optimum with fewer trials.
Once I had the optimum dimensions, I
built a WR-42 transition, shown in Figure
15, and measured it with another surplus
slotted line. The results were encouraging,
with an SWR of 1.02. I quickly made several more, with similar results, so that I
could finish my 24 GHz transverter.
Other Frequencies and Waveguides
To round out the results, in off-hours
when the software was not being used, I
calculated optimum dimensions for some
of the lower microwave ham bands in various sizes of waveguides. These dimensions
are listed in Table 1. Not all the dimensions have been verified experimentally,
but I have built and tested a number of
them in different sizes many of them
are shown in Figure 16. The one at the far
right, in WR-159 waveguide for 5.76 GHz,
has not had the backshort installed yet.
The bandwidths shown in Table 1 are for
a Return Loss > 20 dB, equivalent to an
SWR < 1.2:1. Larger diameter probes generally provide a wider bandwidth. For most
amateur work, wide bandwidth is not
important, but a very narrow bandwidth
often makes the dimensions more critical.
I found that the required tolerance on
the probe and backshort dimensions is
roughly 0.1 mm (0.005 inch) at 10 GHz,
0.05 mm (0.002 inch) at 24 GHz, and proportionally larger at the lower frequencies.
Some of the probe diameters were chosen to ease fabrication. The higher frequency probes are 1.27 mm diameter
the center pin diameter of an SMA connector making fabrication a simple
matter of trimming the Teflon, then cutting and filing the probe to length. At lower
frequencies, the SMA pin diameter is too
small to make an effective probe, but AWG
no. 12 copper wire fits neatly into the solder end of a female N connector with little
discontinuity. The alternative is to slip a
larger diameter probe over the SMA pin
I tried this with WR-187 transitions at
3.456 and 5.76 GHz, and found that the
discontinuity from the step in diameters
required a change in probe dimensions to
compensate. Table 1 shows dimensions
calculated for a 2.36 mm diameter probe
and the modified dimensions for the
same probe diameter extending from an
SMA pin.
Integrated Transitions
In general, it is a good idea to keep the
transition separate from the antenna, so that
each may be tested individually and tuned

Figure 10 Smith Chart plot of impedance as dimensions are adjusted for WR-75 to
coax transition with full height Teflon.

Figure 11 SWR versus frequency of WR-75 to coax transition.

Nov/Dec 2006 15

wade.pmd

15

10/02/2006, 4:59 PM

if necessary. The length of waveguide between the transition and the antenna does
not matter if it is simply a matched transmission line. For simple, foolproof antennas like a rectangular horn, however, it may
be convenient to integrate them into one
piece, with the horn soldered to one end of a
scrap of waveguide and the transition at the
other. Figure 17 is a photograph of two
10 GHz integrated rectangular feed horns for
a common offset dish.4

Figure 12 Insertion loss and SWR of two WR-75 to coax transitions back-to-back.

Summary
By measuring impedance from the
waveguide side of a transition, we are able to
predictably adjust and optimize dimensions.
The impedance may be found using either
cheap surplus hardware or expensive software.
For common waveguide sizes, the optimum dimensions are listed in Table 1 for
most of the amateur microwave bands.
Armed with this data, only some simple metalwork is required to turn a bit of surplus
waveguide into a working transition with no
further tuning required.
Notes
1P. Wade, W1GHZ, The W1GHZ Microwave
Antenna Book Online at www.w1ghz.
org.
2P. Wade, W1GHZ, Understanding Circular
Waveguide Experimentally, QEX ,
Jan/Feb 2001, pp 37-48. (Reprinted with
ARRL permission at www.w1ghz.org/
QEX/circular_wg.pdf).
3www.ansoft.co.
4The
ARRL UHF/Microwave Projects
Manual, Volume 2, ARRL, 1997, p 1-34.
Also Microwavelengths, QST, Aug 2005,
pp 75-76.

Figure 13 Sketch of rectangular waveguide to coax transition, showing dimensions.

Figure 14 WR-90 to coax transition, SWR versus probe length at 10.368 GHz.

Paul Wade, W1GHZ, previously N1BWT and

WA2ZZF, has been licensed since 1962 and

Figure 15 WR-42 to coax transition for

24 GHz in milled aluminum block.

16 Nov/Dec 2006

wade.pmd

16

10/02/2006, 4:59 PM

Figure 17 WR-90 transition integrated

into 10 GHz feed horn for DSS offset dish.

Figure 16 Transitions in several sizes of waveguide for different microwave bands.

Table 1
Rectangular Waveguide to Coax Transitions

W1GHZ 2006

Waveguide Frequency Probe Diameter Probe Length Backshort Bandwidth

(GHz)
(mm)
(mm)
Length (mm)
WR-42
24.192
1.27
2.41
2.49
>17%
WR-75
10.368
1.27
5.49
5.26
14%
WR-90
10.368
1.27
5.89
5.46
7%
WR-112
10.368
1.27
6.5
6.6
15%
WR-112
5.76
1.27
9.8
5.8
7%
WR-137
5.76
1.27
10.5
8.5
10%
WR-159
5.76
1.27
11.17
10.0
11%
WR-159
5.76
AWG no. 12
10.9
10.0
14%
WR-187
5.76
2.36
11.3
11.0
16%
WR-187
5.76
SMA to 2.36
11.6
9.7
16%
WR-187
5.76
AWG no. 12
11.3
11.2
14%
WR-187
3.456
2.36
14.5
18.0
5%
WR-187
3.456
SMA to 2.36
15
16.5
5%
WR-187
3.456
AWG no. 12
14.9
17.4
7%
WR-229
3.456
1.27
18.2
15.0
8%
WR-229
3.456
AWG no. 12
17.7
15.1
10%
WR-229
3.456
2.36
17.4
15.06
11%
WR-229
3.456
3.175
17
15.6
11%
WR-229
3.456
4.76
16.2
16.2
14%
WR-229
3.456
6.35
15.5
16.75
17%
WR-284
3.456
6.35
17.8
17.5
23%
WR-284
3.456
AWG no. 12
19
17.5
11%
WR-284
2.304
6.35
20
28
8%
WR-284
2.304
AWG no. 12
24
24
7%
WR-284
2.4
AWG no. 12
23.7
22
8%
WR-340
2.304
6.35
25
23
11%
WR-340
2.304
AWG no. 12
27
23
9%
WR-340
2.4
AWG no. 12
26
23
9%

has never made a contact below 50 MHz. He

has been a microwave experimenter for years
and published numerous articles, is active in
the Vermont 10 GHz group and a past President of the North East Weak Signal Group.
An ongoing project is the W1GHZ Microwave Antenna Book Online at
www.w1ghz.org. In 1997 Paul was honored

Number
Tested
4
1
5
1
1
1

Mercury Computer Systems designing computer hardware. A frightening experience at

Microwave Update 2001 involving a giant
meatball resulted in an angioplasty with two
stents, and swearing off meatballs in favor
of a low-fat diet.

The Binary Clock

So simple, yet no one

figures out what it is until
you explain to them how
to read binary code.
Then they are amazed
at what a great clock it is
and how easy it is to
read the binary code.
A great gift and perfect
for the techie. The clock
includes a dimmer
switch and real binary
as well as BCD.

to be named by the Central States VHF Society as the recipient of the Chambers Award.
More recently, he was honored by the ARRL
with the 2000 Microwave Development
Award, and in 2001 with the Thomas Kirby
Eastern VHF/UHF Society Award.
A former microwave engineer and retired
ski instructor, he is currently employed by

For instructions on how to

read the clock and or
to purchase, visit us at:

www.realnerds.com
Clock pricing
starts at $18.75.

Nov/Dec 2006 17

wade.pmd

17

10/02/2006, 5:00 PM