"Tuning" a home brew splitter / combiner?

Discussion in 'General Technical Questions and Answers' started by 2E0ILY, Jan 11, 2018 at 7:44 AM.

ad: L-HROutlet
ad: l-rl
ad: Subscribe
  1. 2E0ILY

    2E0ILY Ham Member QRZ Page

    I built this lumped component combiner to run two lowish power, cheap to build Class D amps combined on 137khz. (2200 meters) Whilst it worked the drain waveforms were like those into a very poorly matched antenna load, so I went with a Wilkinson using ferrite beads to make two lengths of Teflon co-ax act as 1/4 waves and a ferrite cored auto transformer. This works well but I want, for the knowledge, to see if I can make the lumped component one work properly. I believe this to be a narrow band, high Q device. Given I have a decent sig gen and scope and access to a very decent SA, can anyone explain in simple terms how I might "tune" it with regard to coil turns or cap values, on the bench, at low power as from the sig gen, to better resonate it please? Thanks. I link to the lumped device and the schematic of the Wilkinson with ferrite, albeit I now use an auto transformer unlike the schematic.

    Lumped component I want to tweak to my desired frequency is at:


    Combiner that works fine :


    Thanks for any replies.
  2. KA9JLM

    KA9JLM Ham Member QRZ Page

    Tune for Peak.

    Have Fun.
  3. 2E0ILY

    2E0ILY Ham Member QRZ Page

    What I tried was sig gen at -5dBm at 137.5kHz into the output of the combiner, one input terminated into 50 Ohms, and the SA on the other input. Looks quite broad band, not what the web page suggests. Plus it seems to need to be tuned for minimum?? Seems it's resonant at 134kHz, but not a huge change over 5kHz each side. Am I doing something wrong? Thanks :)
  4. WA7ARK

    WA7ARK Ham Member QRZ Page

    Here is a LTSpice sim of the Wilkinson Splitter. All terminations are 50Ohms. I sweep the frequency from 130kHz to 1.3MHz.
    The open-circuit voltage at the generator V1 is 1V. Note that a 50Ohm input impedance to the network at P1 occurs only at 630kHz (the design frequency) as indicated by V(P1) being exactly -6dbV.

    V(P2) = V(P3) at all frequencies. Note that the frequency response of V(P2) and V(P3) is a classical low-pass Chebychev filter, with the output dropping off rapidly beyond 630kHz. The response at the output is not a lot different at 137kHz compared to 630kHz (Note the cursors).

    If you like, I can design for 137kHz, and re-run the sim to see how changing L or C effects the response.

    Click to get bigger view:

  5. WB3BEL

    WB3BEL Ham Member QRZ Page

    One way to understand these lumped component Wilkinson combiners is to look at them like they are two impedance transformers combined together.

    The topology shown in the web page you referenced is the most common implementation.

    It is really just two Pi section impedance transformers put together.
    They are not high Q. The page you reference is saying that the inductors that were used would were high Q not the combiner network.

    If you visualize a typical low pass Pi transformer it is made of a shunt capacitor, a series inductor and another shunt capacitor.

    If you design it so that it matches a 100 Ohm source to 50 Ohm load, you can choose the Q of the impedance transformer. So there are lots of solutions to this problem. This is similar to the way you can find a lot of antenna tuner solutions by changing the variable elements. But if you choose Q = 1.414 (Square root of 2) then you get the solution in the calculator webpage you also referenced.

    So you get a solution that has 82.15 uH of inductance and 16.4 nF of capacitance. So you make two of these Pi networks and you then take the 100 Ohm source side and parallel them together so that they create a 50 Ohm source port. You can combine the two capacitors that would be on the input side together in parallel by choosing a capacitor of twice the value.

    So you have 32.8 nF of capacitance on the input and two 16.4nF capacitors on each port of the split output side. The resistor across the output ports is to balance the device. (I am using the terms input and output as would represent measuring the device as a splitter).

    So one easy way to double check this as you are going along is to build just a single Pi network and use your antenna analyzer or if you have a network analyzer, verify that with a 50 Ohm termination you see 100 Ohms resistive or close on the input side. You can also take a step back from that and measure the inductance to verify it is close to right before making a Pi network. Then combine the two Pi networks together. You can measure with an antenna analyzer or NWA or as you mention a signal generator and spectrum analyzer or oscilloscope.

    If you put a signal into the common port the signal should split equally between the two output ports and should have the same phase. You can verify this with an oscilloscope with 50 Ohm input termination in two channels if you have that. The loss should be slightly more than 3 dB between the common port and each output. You can also measure this with spectrum analyzer, but you wont be able to observe the phase relationship.

    You can measure the isolation by driving a signal into one of the output ports with the common input port terminated in 50 Ohms while observing the other output port with a 50 Ohm impedance test equipment. A spectrum analyzer or oscilloscope with 50 Ohm input selection. If you sweep the frequency you should observe a null where the isolation is greatest between the output ports. If this is greater than 20 dB or so I would not really get concerned about the exact frequency of this null unless you are certain of all your instrumentation.

    One thing to be cautious about that can create confusion, is that you might have a good impedance transformer Pi section, but when you put two together to create a combiner/splitter you put the inductors together so that they have significant mutual coupling. This can be an issue at low frequencies where the size of the coils is large and they are placed close together. If you find this to be a problem then orient them so the flux does not link.

    This kind of combiner/splitter really should not need to be tuned very much if it is constructed correctly.

    I hope this helps. Good luck with your experimentation.
    -Harry WB3BEL
    AI3V likes this.
  6. WA7ARK

    WA7ARK Ham Member QRZ Page

    Ok, here it is re-designed for 137kHz, driving P2, and properly terminating P1 and P3. Note that the "isolation" from P2 to P3 occurs only at the design frequency. It is not a broad band network.

    Again the input swr =1 only at the design frequency as shown that V(P2) is down -6db at that frequency. 64.png
  7. 2E0ILY

    2E0ILY Ham Member QRZ Page

    Some fantastic answers and examples, thanks for taking so much time to pen these! Two more questions please?

    If I measure the coils with a Philips LCR meter they show correct inductance but it doesn't measure at 137kHz. I have an AIM4170 antenna analyser, am I better checking them with that at 137kHz?

    I am testing *WITHOUT* the load balancing resistor 9it's in the box the Teflon co-ax with ferrite beads and auto transformer is in, is it necessary for the tests or just a safety feature in actual usage at high power? I will test again with the coils at 90 degrees to one another, and one horizontal and one vertical in case coupling is the issue. I was sloppy in my description of poor waveforms wth the lumped component combiner. i meant to say the waveforms on the drains f the Class D amps being combined are noisy, which experience tells me is caused by a mismatch. Do I need two LPF's, one in each leg *BEFORE* the lumped combiner? With the ferrite toroids on Teflon co-ax one it works fine with one LPF after the combiner... The antenna is a perfect match after much fiddling with matching transformer taps and the overly sensitive variometer. My Scopematch allows me to watch this in real time. So it points to the combiner offering the mismatch, not what is beyond it?

    Learning a lot here, thanks again for all your trouble :)
  8. WA7ARK

    WA7ARK Ham Member QRZ Page

    My simulation is based on a 50 Ohm sinosoidal source. Are you thinking that the two Pi sections that are in the Wilkinson combiner double as the matching network at the output of you Class D amp?
  9. 2E0ILY

    2E0ILY Ham Member QRZ Page

    I am not entirely sure what I am thinking ;) All I can say is with the ferrite bead 1/4 wave stub combiner it seems happy with the LPF after it. The lumped one will be seeing whatever the Class D amps put out, which is probably far from sinusoidal. is there a reason the ferrite combiner would be happy and the lumped one unhappy? Thanks!
  10. WA7ARK

    WA7ARK Ham Member QRZ Page

    Just to clarify. You are feeding the combiner from a Class D FET switch with no tuning network between the drain of the FET and the input port on the Wilkinson combiner?

Share This Page