Coaxial Lowpass Filter for 2.4 GHz

  • Hi,

    I designed a coaxial lowpass filter for 2.4 GHz. The cutoff frequency is 3 GHz, Order 7 according the book "Microwave Filters, Impedance-Matching Networks, and Coupling Structures" by G. Matthaei, L. Young and E.M.T. Jones. You can see the design here:



    The most difficult part is to calculate the fringe capacities at points, where the diameter of the inner conductor changes. Fortunately, in the book you can find diagrams to estimate them.


    The next part will be to energize my workshop.


    Vy 73,

    George

    DL2KP

  • Hi Matthias,

    it is similar to Kurt´s one. The difference is the cutoff frequency and I missed an addaptation to 50 Ohm for in- and output. Maybe he didn´t show it. Here now some pitures out of my workshop:






    Now I am curious about the measurement results.


    Vy 73, Georg

    DL2kp


    P.S: Matthias, did you glue teflon, already?

  • Hi,

    now I got my first measurement:


    Here is the result after assembling without any tuning:


    The attenuation is pretty high. I guess, this is due to the kind of construction. So far I din´t solder either the disks nor the connection to the connector. They are pressed on the center 4mm wire, only. Further more the cuttoff frequency is at about 2.6 GHz and was designed at 3 GHz. For my application it is ok, but I don´t know why it it so low. In the next step I will solder all parts together.


    Vy 73,

    George

    DL2KP

  • Hi Matthias,

    you are absolutely right, but it would be a wonder if the result would be perfect at the first shot. Now I am going to optimize the filter. At these frequencies it would be necessary to silver-plate the filter to get the best performance.



    Vy 73,

    Georg

    DL2KP

  • Hi,

    I got the impression that the high attenuation of my coax filter could be an issue of the calibration of my LibreVNA. It was calibrated with female Rosenberger parts from SDR-Kits. They stated to be valid up to 6 GHz. I am not sure, if it is guaranteed. Now I got the appropriate s1p files for LibreVNA standards. Unfortunately the filter is disposed already.


    Nevertheless, a new filter was designed:


    The order of this filter is 7 instead of 9 of the old one and the cut-off frequency was chosen to 3.5GHz.


    And here some pictures:





    As you can see the material is alu EN AW 6082 (AlMgSi1) which is not very good for milling. This is while the surface is not perfect.

    Aluminium has a much better conductivity than brass.


    The measured parameters are as follows:



    The attenuation in the QO100 frequency range is now -0.5dB but the imput impedance is not 50 Ohm. This is disappointing as I included the 50 Ohm section on both sides of the filter.


    I have two questions:

    - what is the reason that the S11 impadance swings over 0 dB?

    - does anyone have an idea to improve the design?


    Vy73

    George

  • Hi Matthias,

    thank you for your fast help. It is clear that my LibreVNA has a problem, or the operator in front of it or both. I have some ideas to improve the design, already. It is like artillery, the third shot will hit the target, hopefully.


    George

  • Hi,


    Eureka, the third attempt hits the target, as I promised. Analysing the previous built filters, I assumed that the reason of the high attenuation was the so called missmatch loss. Therefore, I tried another approach. Manual computation of the filter parameters according to Matthaei is time consuming and boring. In the net you will find very detailed Exel sheets by Dominique, F1FRV. It turned out that they are not applicable for 2.4GHz because of the small dimensions of the construction. Another online calculation is offered by

    http://www.changpuak.ch.

    I realized that this calculator presents the same results as I calculated. As I don´t have access and money to use HFSS Electro Field Simulator, I grabed the old free of charge available RFsim99 Simulation program. Unfortunately it does not include parameter optimization. But with the well known try and error method you can get good results together with an available auto match function. Here is the simulation circuit:



    And here the simulated results:



    At the working frequency the simulation shows, that we get an input impedance of about -67dB and an attenuation of -770ndB.

    With these new parameters a new design was created:




    In contrast to my first construction, the brass disks were soldered to the inner conductor with butan burner at about 750 deg Celsius.

    And here my new filter for QO-100:




    An attenuation of 0.1dB at the working band and an input impedance of of -27dB is an acceptable result. It is not perfect as you can see it could be better, but it is not a tunable filter. Now, I will try to build another one to get it perfect, probably.


    Best regards

    George

  • Hi,

    in addition a picture of the actual version:



    I guess that one important matter is to utilize silver braze solder to reduce attenuation, too. To keep the distance of the disks in position, you need a ficture during soldering. It is out of aluminium otherwise it would be soldered, too.



    To be honest, it is difficult to construct the filter with a accuracy within 0.01mm. I don´t have a CNC-lathe, unfortunately.


    Have a nice Easter!


    George