THREE-BEAD-BALUN REFLECTION BRIDGE

250 KHz to 1.5 GHz

Sam Wetterlin
swetterlin@[Link]
9/22/08

This paper presents a reflection bridge useful from 250 KHz to the neighborhood of 1.5 GHz.
This bridge may be used in conjunction with a vector network analyzer (VNA), in which case the
use of OSL calibration makes its performance about as good as one could want. It may also be
used without calibration as a free-standing bridge, as discussed below.

The schematic of the bridge is shown in the Appendix. It consists of three parts. The first part is
an input attenuator, resistive bridge, and Mini-Circuits TC1-1-13 balun, all on a single PCB. The
second part is a homemade balun made with RG-178 coax and three small ferrite beads. This
balun is supported by a brass strip, which connects the first and second PCBs. The third part,
contained on a second PCB, is an output attenuator and output connector. Because of the
homemade balun, I refer to this as the Three-Bead-Balun Bridge. The finished bridge is shown in
Figure 1.

Figure 1—Assembled Three-Bead-Balun Bridge

The ferrite beads are wound in fiberglass tape to keep the coax
loops from shifting over time.

Figure 2 shows a wound ferrite bead before wrapping.

Figure 2-Ferrite bead with three turns of RG-178

(Passing once through the center counts as one turn)
Theory of Operation
This bridge is a standard resistive bridge, using two 49.9 ohm resistors (close enough to 50 ohms
for present purposes) as the upper legs of two voltage dividers, and using the device under test
(DUT) as one lower leg, and an external 50 ohm load as the other lower leg. (The external load is
used to preserve symmetry, which is useful at higher frequencies. Below a few hundred MHz, an
on-board 49.9 ohm resistor can be used.) The reflection from the DUT is equal to the difference
between the voltage on the DUT and on the 50-ohm reference. That difference is extracted by
means of two successive baluns that function as common mode chokes to pass only the
differential signal.

The trick to obtaining decent performance is to have an effective balun. The TC1-1-13 balun by
itself can only obtain directivity on the order of 20 db (more on directivity below), and also has
weak performance below 5 MHz. Therefore, a second balun is used to further choke the common
mode signal.

Operation
To use the bridge as a free-standing device to measure return loss, an input signal is connected to
the input and the output is connected to a spectrum analyzer. An “open” standard is connected to
the DUT port, and the output level is recorded. A “short” standard is then connected to the DUT
port, and the output level recorded. The average of the open and short output levels becomes the
reference level for that frequency. Various DUTs can then be attached to the DUT port, and the
output level recorded. The difference between the DUT output and the reference level (obtained
from averaging the open and short), expressed as a positive number in db, is the return loss of the
DUT. That return loss measures the reflection received from the DUT relative to the signal sent
to the DUT. For example, if the average of the open and short is -12 dbm, and the output with the
DUT attached is -30 dbm, then the measured return loss is 30-12=18 dbm. (Error will be
discussed later.)

The bridge may be used with a vector network analyzer, in which case the use of Open-Short-
Load (OSL) calibration will greatly improve accuracy. In addition, the VNA will be able to
measure the phase of the reflection, which enables it to determine S-Parameters.

Performance
The quality of a reflection bridge is measured primarily by its directivity. Directivity is the
amount of indicated return loss with a 50-ohm load. There should be no reflection in that
situation, so the return loss should be infinite. For use with a VNA directivities of 20 db or better
are acceptable, and anything over 30 is nearly perfect. For free-standing use without calibration,
a directivity of 30 db will allow measurement of actual return losses up to 30 db, with some
uncertainty (more on that later).

Another measure of quality is the difference between the open and short output levels. Ideally,
the magnitude of the reflection with an open and shorted DUT should be equal. A discrepancy
between them may indicate that the DUT port does not present a good 50-ohm impedance to the
DUT. When directivity is good and the open and short are close to one another, the DUT port
impedance is likely very close to 50 ohms.

Figure 3 shows the measured directivity and open/short for the Three-Bead-Balun Bridge. Note
that while we subtract the short output from the open output to determine open/short, we are
subtracting db levels so the resulting difference actually represents the ratio of the open and short

2
outputs. Hence the label “open/short”. Open/short magnitudes can be as high as 1.5 db without
presenting a serious problem.

Figure 3—Directivity and open/short

Figure 3 shows that the directivity is good from 250 KHz to somewhere near 1.9 GHz, and is
excellent from 1 MHz to 1 GHz. The open/short value exceeds 0.5 db at only two small peaks for
frequencies below 1.4 GHz.

The DUT port should ideally present a 50-ohm impedance to the DUT. If any part of the
reflection reaching the DUT port is itself reflected back to the DUT, the measurement of the
reflection will be inaccurate. When the bridge is used with a VNA, calibration will correct for
this error, unless it becomes unreasonably large. For free-standing use of the bridge, DUT port
return losses of 30 db or better are acceptable. The open/short value can be used as a guide to
whether return loss is getting out of hand, but the correlation is not perfect. Figure 4 shows the
measured return loss of the Three-Bead-Balun Bridge, using a technique described later. This
technique does not pin down the exact return loss, but does indicate the worst that the return loss
can be. In the case of Figure 4, the technique could only be used up to 300 MHz:

3
 Figure 4—Return Loss of DUT Port to 300 MHz

Figure 4 shows that the return loss is extremely good from 250 KHz to 300 MHz. Given the
open/short and directivity values, I would expect this good return loss to continue beyond 1 GHz,
possibly dipping below 30 db briefly around 450 MHz.

Measuring Return Loss Manually

A reflection bridge can be used to measure return loss manually, as described above under
“Operation”. The question arises as to how accurate such measurements are, given the lack of
OSL calibration, which a VNA would perform. First, consider how much accuracy is needed.
Return loss indicates the amount of power reflected by the DUT. Generally, if the DUT has
return loss of 30 db or better (higher), it is reflecting so little power that it does not really matter
exactly how little it is. Or in terms of impedance, return loss of 30 db or better means the
impedance of the DUT is close enough to 50 ohms that it can be considered to be 50 ohms. For
example, resistances between 47 and 53 ohms have return losses of 30 db or higher.

On the other hand, if a DUT reflects half or more of the power reaching it, we may just consider
it to be unacceptable, without having to know exactly how much it reflects. A DUT with a return
loss of 3 db or worse (lower) reflects at least half the power reaching it.

So if we know the return loss is higher than 30 db or lower than 3 db, that may be all we need to
know. And even for values in the 3-30 db range, errors of several db may not be meaningful.

The one situation where precision might be useful is to determine the exact impedance of the
DUT, for example in order to devise an impedance matching network to adjust it to 50 ohms.
However, because return loss is a scalar value, we cannot determine the exact impedance even if
4
we knew the exact return loss. Phase information would be required. Our manual use of the
bridge provides only scalar information, and is effectively a “scalar network analyzer”, as
opposed to the fancier “vector network analyzers”, which provide phase information. The
absence of phase information in return loss generally makes it pointless to try to determine return
loss precisely.

With that background, what error can we actually expect? Consider the frequency range 1 MHz
to 1 GHz. Directivity in that range equals or exceeds 30 db, so assume as the worst case that it
equals 30 db for the entire range. Return loss of the DUT port exceeds 30 db at least up to 300
MHz, and we have reason to expect it will do so up to 1 GHz, possibly with a small dip below 30
db near 450 MHz. With those numbers, the major sources of error are as twofold.

First, the open/short discrepancy creates some error. Ideally, we would use the bridge output
with an open DUT as the reference to calculate return losses for large resistances, and the output
with a shorted DUT to calculate return losses for small resistances. But we don’t know what
resistance we are measuring, so we use the average of the open and short values. That creates a
possible error equal to half the open/short difference. In our case, this difference reaches a
maximum of 0.7 db, so the error is 0.35 db. It is conceivable that the calculation for reactive
devices might involve larger error; we don’t have phase information so we can’t be precise about
that. But our open/short and directivity levels are so good and well behaved that I would not
expect there to be additional error for reactive devices. But let’s say the total error from the open/
short discrepancy may equal 0.5 db.

Second, directivity essentially indicates a bogus reflection that is present even when there is no
real reflection from the DUT. If the DUT actually produces a reflection, the bogus reflection may
add to or subtract from that reflection, depending on phasing. So the bogus reflection may cause
us to measure too much or too little reflection. With our assumed directivity of 30 db, we can
actually calculate the extent of maximum error in our measurement. I won’t describe the process
here, but Figure 5 illustrates the relationship between our return loss reading and the possible
range of values of the actual return loss.

5
 Figure 5—Possible actual return loss values for a given return loss reading.
For example, a reading of 20 db indicates the true return loss lies above 17 db
and below 23 db.

Figure 5 shows that we can determine return losses with reasonable accuracy if they are below
20 db. Once the return loss reading rises to 30 db and higher, we can say with confidence that the
return loss is at least some minimum value (24 db for a reading of 30 db), but there is no limit to
what the maximum might be. That may seem like unacceptable error, but remember that for
large return losses we frequently only need to know that the return loss exceeds some threshold,
without caring how far it exceeds the threshold. If we get a very large reading, like 60 db, we can
be sure the true return loss is at least 30 db, which may be all we are looking for.

Remember our 0.5 db uncertainty resulting from the open/loss discrepancy. For readings above
15 db, that error is dwarfed by the uncertainty created by directivity. For readings near zero, it
becomes significant. For very low readings, indicating highly reflective DUTs, there are
additional possibilities for error. It is probably good practice to proceed as follows:

1. For readings below 5 db, assume a 1 db error.

2. For readings of 5 db to 25 db, use the graph in Figure 5 to estimate the possible error, and
add 0.5 db.

6
 3. For readings above 25 db, use the graph in Figure 5 to estimate the minimum possible
return loss, and assume there is no ceiling to how high the true return loss may be.

The graph in Figure 4 was actually calculated by the method described in (3), using a bridge
whose directivity was greater than 60 db at low frequencies and dropped to 40 db at 300 MHz.
That is an active bridge using op amps, and will be described in a separate paper.

7
APPENDIX--Schematic