In the first parts of this series of articles, some experiments with prototypes
of return loss bridges are described.
With the lessons learned there, I will now build the definitive version.
I wanted the frequency range at the low end to extend to at least 100 kHz or
less.
For this purpose, cores with high permeability (μi = 5000)
are added.
And at the high end of the spectrum it would be nice if it is useable to at
least 2000 MHz.
For the higher frequencies, the cores with lower permeability are more
effective.
I tried to make optimum use of the available space in the enclosure, and use as
much as possible cores.
This design uses the following circuit diagram.
If you order the ferrite cores, order some extra pieces because these things can
break if you handle them to rough.
And some of the 74270176 cores can have too much thickness to fit into the big
cores, therefore you might need some extra pieces.
Also some extra of these very tiny RF resistors are welcome, if you drop one on the floor, it is
very hard to find :-)
In the return loss bridge, we will make a junction where five coax cables come
together.
Also the two resistors must be mounted between there.
The more compact this junction is build, the better the bridge will work at very
high frequencies.
Any extra mm length in the interconnections in the junction should be avoided.
Figure 2: two PCB boards are used in this return loss bridge.
The large one is used to attach the coax cables to.
The small board will hold the resistors.
Figure 3: back side of the large board.
Figure 4: the small board with the two resistors mounted.
The back side of this board has no copper layer.
The board has three small copper islands, which are directly soldered to the
centre conductors of the IN, the REF and the TEST coax cables.
Figure 5: the output line with connector, and the balance line, bend in there final shape,
and cut exactly to the desired length.
It was quite some work to reach the optimum shape for these coax cables, with a
lot of fitting and shifting ferrite cores on and off the cables.
The "hot side" of the cables (right hand side in this picture) must be
symmetrical, to keep optimum balance of the bridge at higher frequencies.
The cables need enough length to fit all the ferrite cores I wanted, therefore I
needed an extra bend near the "cold side" of the cable (left in this picture).
The mounting of the cables should be without mechanical stress on the solder
joints or ferrite cores.
In this stage we should give the cables exact there final shape
You can make it yourself easier if you reduce the number of 74277233 cores from
12 to 11 or 10 on each line, this will increase the minimum frequency of the bridge from 60
kHz to maybe 70 kHz or so.
Figure 6: detail of the ends of the output line and balance line.
When you heat up the copper tubing outer of the cable, you will notice that the
Teflon will expand slightly, and this will come out of the cable.
You can in this stadium already pre-heat the cable end one time, and remove this
surplus of Teflon, so it doesn't bother you later when you solder the coax in
place.
Figure 7: after the cables have there shape, they are insulated with heat shrink tubing
(black in this case), and partly with PVC tape (green / yellow in this case).
On the bend, the heat shrink tubing had too much thickness, and the ferrite
cores could not pass the bend.
Therefore the thinner PVC tape in used on the bends.
The bend at the right side of the coax is not yet insulated as can be seen in
the picture above, this is done later when most of the ferrite cores have passed
this bend.
The balance line has a cable shoe soldered to one side, this point is later
connected to the ground of the box.
Near the cable shoe and the N-connector I placed a rubber ring on the coax,
meanly to prevent breaking the ferrite cores when they are pressed to strong on
the extra thickness the coax has there.
The cables are insulated for several reasons:
- Some of the ferrite cores are electrical conductive (the 8W5000 material), the insulation prevents
any unbalance by the fact that maybe one ferrite core is making better contact
with the coax then the other.
- The insulation causes the distance between the coax and the ferrite to be
everywhere the same, in other words; the coax runs quite exactly through the
centre of the hole in the ferrite cores, this improves the balance in
capacitance between coax and ferrite cores.
- The insulation prevents rattling of the ferrite cores, when you move the
device.
Figure 8: from 0.5 mm plastic sheet, six of these rings are made.
The hole in the centre can be easily made with a hole puncher.
Figure 9: in this picture the return loss bridge is partly assembled.
The PCB board is mounted, with 3 N-connectors with short pieces of coax
connected to this PCB.
All these coax cables have a (slight) bend, so the cables can take up some thermal
expansion.
First the outer conductor of the three cables (IN, REF and TEST) are solder
together on the board, and after that the resistor board which is already in
place is soldered to the coax pins.
In this way we avoid mechanical stress to the small resistor PCB board.
The output coax and balance coax are not yet soldered to the PCB board, because
more ferrite cores have to be added.
The green plates are extra supports for the coax cables with ferrite cores, to
keep them in the correct position, and to reduce the mechanical stress on the
solder joint to the PCB.
The green material is polypropylene, and it can be fixed to the box via screws
through the bottom of the box.
On the right side you see the small ferrite cores, these cores are also
inside the bigger cores, each big core in this picture contains 3 of these
smaller cores.
The 3 cores inside are taped together with PVC tape (green / yellow, also
visible in the picture), this fills up the gap between the smaller and bigger
cores, and also gives some extra insulation.
The plastic rings prevent the smaller cores to shift sideways, from one big core
into another big core, and the plastic rings also insulate the big cores from
each other.
Before assembling the 74270176 cores, you better do a check if they are small
enough to fit into the big cores.
Figure 10: here the bridge is fully assembled.
The directivity of the bridge is carefully optimized by slightly shifting the
cores, there was some mm space to shift the cores from left to right.
The measurement gave maximum directivity with the cores in this position, which
is with the cores symmetrical on both lines.
The ferrite cores are then fixed in there position by applying polymer glue, the glue
didn't had any effect on the measured performance of the bridge.
Figure 11: detail of the junction of coax cables.
To get an idea of the scale; the coax cables are about 3.6 mm thick.
The very small resistors (green colour) are also visible.
And the tuning wire is located above the resistors.
The tuning wire gives the best balance in the bridge when the wire is pointing
straight forward, the effect is about the same as not using a tuning wire at
all.
This seems to indicate that this design of a return loss bridge is very well
in balance of it's own.
From the three coax cables on the left side in this picture, the outer conductor
is filed thinner at the end where the cables touch each other, so that the
cables could be placed closer together, but this is not very clear to see in
this picture.
Figure 12:
I accidentally drilled one hole for a connector in the wrong place in the box,
as you can see here.
Figure 13: this is how it looks with the box closed.
I now discovered my label printer has no ">" sign, I actually wanted it to print
"60 kHz ... >2000 MHz"
Measurements:
Now the return loss bridge is finished, the performance is measured.
The REF and TEST ports are terminated with 50 Ohm N type terminators from
Radiall, part number: R404131000.
Datasheet: datasheet_Radiall N
terminator.pdf
The two pieces I have, have a measured DC resistance of 49.9
Ω and 49.7 Ω.
Figure 14:
N type terminator from Radiall.
Figure 15: response of the return loss bridge over the maximum span of my spectrum analyser
(0 to 2100 MHz).
The reference trace (yellow) is flat within 3 dB.
Directivity is 40 dB at 1000 MHz, and not less then 34 dB over the entire span.
Figure 16: response from 0 to 50 MHz, directivity is more then 50 dB.
Figure 17: response at low frequencies (0 to 2 MHz).
The two markers are set to 60 kHz, at that frequency the reference trace has
dropped 3 dB.
Directivity is about 47 dB at 60 kHz.
But you can even use the bridge at (slightly) lower frequencies with reduced
performance.
Figure 18: because the input and output port of the bridge are on the same side of the
device, it can
easily be connected to the spectrum analyser with short cables.
Here I test a very simple dummy load, which I build in my younger days.
The screen shows the response from 0 to 500 MHz, at 100 MHz the return loss is
about 19 dB
