Back to the index.
<< Go to part 3 Go to part5 >>
Return loss bridge. (part 4)
On this page, some experiments are shown with my second prototype of a return loss bridge.
Prototype version 2.1
Figure 1: this picture shows return loss bridge version 2.1 .
The bridge is build on a wooden frame.
The coax is RG402 semi-rigid coax.
Both the output line and the balance line have:
20 ferrite cores part number 74270176 from Würth Elektronik, these are the small cores, datasheet: (datasheet_Wurth74270176.pdf).
The ferrite material is: 7W850 with permeability μi = 850, these cores are effective for higher frequenties.
And 5 ferrite cores part number 74277255 from Würth Elektronik, these are the bigger cores. datasheet (datasheet_Wurth74277255.pdf).
The ferrite material is: 8W5000 with permeability μi = 5000, these cores are added to get enough impedance at lower frequencies.
Here in version 2.1 I use all the cores behind each other, but it is also possible to shift
the smaller cores into the bigger ones, but more on that later.
Figure 2: detail at the output side.
Note the (white) insulation applied between the big cores and the coax cable.
The 8W5000 material of the big cores is electrical conductive, about 2 kΩ between the ends of one core.
When these cores touch the coax cable, I noticed the bridge can easily come in unbalance at lower frequencies.
Therefore I insulated these cores from the coax.
Figure 3: detail of the resistors, which are in this case standard 1/4 Watt carbon film resistors.
Two resistors of 100 Ω are parallel to get 50 Ω, the resistors are selected to have equal resistance.
In this case, I measure 49.3 Ω and 49.3 Ω.
The next thing is to measure the directivity of the bridge.
This gives a small problem for measurements below 100 kHz, because the used tracking generator has a minimum specified frequency of 100 kHz.
At 100 kHz the generator amplitude is already reduces by 1 dB, and at lower frequencies this reduction get more and more.
To overcome this problem, first the tracking generator is directly connected to the analyser input, and a "normalize" function is performed.
The analyser then compensates for all variations it measures in the tracking generator output level.
Figure 4: directivity of version 2.1 from 0 to 10 MHz.
The directivity is about 50 dB
Both marker 1 and 2 are set to 100 kHz, this seems to be about the minimum useable frequency of this bridge.
At 100 kHz, the reference trace is dropped 2 dB.
Maybe you also notice the yellow reference trace is at 5 MHz about 1 dB lower then at 0.5 MHz.
This is caused by the impedance curve of the big 8W5000 cores, which shows a dip around 5 MHz, this will be solved in a later version of this return loss bridge.
Figure 5: directivity of version 2.1 from 0 to 2000 MHz.
The directivity at higher frequencies is much better then prototype 1 (see part2).
Prototype version 2.2
In this version, 50Ω / 0.1% metal film resistors are used.
These resistors are from the PTF56 series made by Vishay.
Datasheet: datasheet_PTF56_resistor.pdf .
The measured values of the two resistors are 49.9 Ω and 49.9 Ω.
Figure 6: version 2.2 of the return loss bridge with 0.1% metal film resistors.
Figure 7: directivity of version 2.2 between 0 and 10 MHz.
The results are comparable with version 2.1
Figure 8: directivity of version 2.2 between 0 and 2000 MHz.
Prototype version 2.3
The resistors are now 1206 SMD types, two 100 Ω resistors soldered above each other to get 50 Ω.
The measured values are: 50.3 Ω and 50.2 Ω.
The resistors are placed on a small piece of PCB material.
Figure 9: version 2.3 of the return loss bridge with 1206 SMD resistors.
Figure 10: directivity of version 2.3 between 0 and 2000 MHz.
Prototype version 2.4
The resistors are now replaced by 50 Ω / 0.1% RF resistors in 0402 SMD package.
Datasheet: datasheet_0402_RF_resistor.pdf .
These resistors made by Vishay are specially designed for high frequency up to several GHz.
Figure 11: return loss bridge version 2.4 with the 0402 RF resistors, the resistors are very small.
The PCB board holding the resistors is the same as in version 2.3 .
Figure 12: directivity of version 2.4 between 0 and 2000 MHz.
Prototype version 2.5
In this version, the coax to the REF and TEST port are placed closer together.
The PCB board holding the resistors is made smaller, and the output and balance line are connected in a much more compact way, with shorter leads.
The resistors are the same as in version 2.4 .
Figure 13: the connections of the coax cables are made shorter.
Figure 14: directivity of version 2.5 between 0 and 2000 MHz.
For a good comparison, the results of version 2.4 are also shown.
Version 2.5 has a more flat reference trace, and a higher directivity, this is only the result of making some wires shorter.
Figure 15: overview of version 2.5 as measured above.
20 small 7W850 ferrite cores and 5 large 8W5000 cores are placed behind each other on the coax cables.
Prototype version 2.6
In this version, there are still 20 small ferrite cores on each coax line.
But 10 of these small cores are shifted into the big cores.
The small cores have 10.0 mm outside diameter, and the large cores have 10.2 mm inside diameter, so it should just fit.
However some cores have some tolerance in diameter, I had to do some selection to find out which cores actually fits inside each other.
Shifting the cores in each other, make the whole return loss bridge smaller, and it's easier to find a box where it all fits in.
Figure 16: shifting 2x 10 pieces of the small ferrite cores in the larger cores.
The measured directivity of version 2.6 is the same as version 2.5 .
Prototype version 2.7
Until now I used 20 small 7W850 ferrite cores, and 5 large 8W5000 cores on both coax cables.
In the next measurement the number of small cores is reduced, in an attempt to make the return loss bridge smaller.
In version 2.7 there are 10 small 7W850 cores on each coax, 5 inside a large ferrite core, and the other 5 not.
Figure 17: version 2.7
2x 5 small ferrite cores are visible, and 2x5 small ferrite cores are hidden in the big cores.
Results are shown below version 2.9.
Prototype version 2.8
Now the number of small 7W850 ferrite cores is reduced to 2x5.
This looks the same as figure 17, however now there are no small cores hidden in the big cores.
Prototype version 2.9
The number of small 7W850 cores is back to 2x10.
And also added are 2x 5 pieces of 4C65 ferrite core, these are the white cores in the next picture (datasheet 4C65 material).
The white cores are part number TN9/6/3-4C65 made by Ferroxcube, the outside diameter is 9 mm, the inside diameter 6 mm, and the thickness is 3 mm.
The 4C65 ferrite material has a permeability of 125.
Figure 18: version 2.9.
2x5 pieces of 4C65 core (white), 2x10 pieces of 7W850 ferrite cores, and 2x5 pieces of 8W5000 cores (the big ones).
Figure 19: directivity of return loss bridge version 2.7 , 2.8 and 2.9 .
In all cases the reference trace was the same (the yellow trace A).
Below 1000 MHz version 2.9 has the best directivity, and as this is the most interesting part of the spectrum for me, I continue with the ferrite configuration of version 2.9 .
After the measurements I discovered that I used different spectrum analyser
settings as in figure 14.
The input attenuator is now 10 dB (was 20 dB), and the Resolution Bandwidth (RBW) is now 300 kHz, this was 1 MHz in figure 14.
This may have some effects on the measured values.
Figure 20: detail of version 2.9 .
Each time the ferrite cores are changed for the different measurements, the coax cables have to be disconnected and soldered again in place.
Be aware that even small changes in the exact location of soldering have influence on the bridge response at high frequencies.
Note the black heat shrink tubing between the cores and the coax cable, to keep the coax in the centre of the cores, and preventing the cores to shift sideways too easy.
Prototype version 2.10
The next experiment is adding a tuning wire connected to ground.
The tuning wire can be bend towards the output coax, or more towards the balance coax, to improve the capacitive balance at this junction of cables.
Figure 21: the tuning wire is added
Figure 22: the effect of a well adjusted tuning wire, the directivity improves over almost the entire spectrum.
Prototype version 2.11
Now we will jump back to the lower part of the RF spectrum
For this version, the number of big 8W5000 cores is reduces to 2x3 pieces.
But 2x10 smaller 8W5000 cores are added.
The smaller 8W5000 cores are part number 74277233 from Würth Elektronik.
However the ferrite material is the same, the smaller cores have much more impedance around 5 MHz then the big cores.
The next picture shows the impedance per cm length of the used ferrite cores.
For the 4C65 cores I don't have impedance data, so these are not in the graph.
Figure 23: impedance per cm length for the 7W850 and 8W5000 ferrite cores.
Figure 24: return loss bridge version 2.11
The composition of the ferrite cores is:
2x5 Cores of 4C65 material
2x10 Cores of 7W850 material
2x10 Small cores of 8W5000 material.
2x3 Large cores of 8W5000 material.
Figure 25: by adding the small 8W5000 cores, the response at low frequencies improves.
The reference trace is more flat, and extends to slightly lower frequency, and the directivity becomes higher.
The directivity is already 40 dB at 100 kHz, and increases to 55 dB at 1 MHz.
This is the end of the experiments with this prototype.
<< Go to part 3 Go to part5 >>
Back to the index.