Please help me with rf filter

Discussion in 'The Projects Forum' started by marvelouzone, Jul 2, 2010.

  1. marvelouzone

    Thread Starter New Member

    Jul 2, 2010
    2
    0
    Thanks For taking the time to read. I'm trying to make a rff filter there are different types that could work in my situation. The first would be a high pass that blocks everything under 120Mhz, the application is cable Tv signal, so the impedance would be 75 ohm. I'm looking for attenuation in the -50db range, there are two very specific carriers i'm trying to block 75.25 in one location and 104.2 in another, which is a qpsk carrier each channel is 1.5 Mhz wide, besides those two frequencies, there's nothing on the system until about 200mhz, so a high pass filter would work in both situations or two separate notch filters could work also as the 75.25 to be blocked is at one site, and the 104.2 is at another. I'm thinking building two high pass filters that will work for both sites, would be easier rather having to build two different type of notch circuits. I tried a few different type of lc circuits, but nothing really worked as i wanted to, I'm am not new to circuitry soldering, and different type of diy electronic projects, but i am new to the rf realm of things.

    As a side note, i was thinking my lc circuits may have not been working because of bad shielding issues from the source side of the cable not being isolated from the other side of the lc circuit, but not sure.

    ps. I would like to build this as cheap as possible.

    thanks for your help
     
  2. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Have you used Elsie?
    http://www.tonnesoftware.com/elsie.html
    Their freeware version will go up to 7 poles.

    Here's something I whipped up for you:

    [​IMG]

    See the attached plot.

    It's a modified Cauer capacitor-input high pass; two zeros were deliberately placed near 75.2MHz and 104.3MHz to get band rejection in the neighborhood of -70dB. The high-pass has some slight ripple, but it's not bad. The return loss is well below -20dB in the passband, which should be more than acceptable.

    Tuning it will be a whole 'nother animal. If you have at least a 200MHz O-scope and a function generator that will sweep through that range, you're ahead of most. If you have a network analyzer with an S-parameter test set, you're cooking. A spectrum analyzer would work, but you won't get insertion loss.

    Don't know if you've built filters before, but the parasitics on the board, board to case, and interactions between the inductors will cause whacky things to happen.
     
    Last edited: Jul 2, 2010
  3. SgtWookie

    Expert

    Jul 17, 2007
    22,182
    1,728
    Here's one that's a bit better.

    Cap values are nearest 5%, inductors have been re-tuned to compensate. Return loss is much better in the passband.

    [​IMG]

    Now the problem you're going to have is trying to build/tune the thing on a PCB.

    If you're using 1.6mm thick FR4 material with one side being a ground plane, a 165 mil wide trace will have an impedance of right around 75 Ohms.

    The parasitics will be difficult to determine/compensate for unless you have a network analyzer with an S-parameter test set or spectrum analyzer.

    One thing that'll drive you bonkers is that even multilayer SMT ceramic caps don't measure the same at 200MHz as they do at DC (low frequencies) due to parasitics. The larger the value of capacitance, the worse the parasitics get.

    For example, a cap that tests as 47pF in the <1MHz range will measure perhaps 60pF to 75pF at 200MHz. If you wish to avoid the parasitics altogether, you can try using single-layer caps - better yet ceramic disc caps with VERY short leads. Radio Shack stocks an assortment; you'll mostly get very low values, which is what you need.
    [eta]
    Read this document; it explains what I'm talking about: http://www.atceramics.com/pdf/technotes/effective_capacitance_vs_fr.pdf

    Another possibility is to use lower-value multilayer SMT ceramics in conjunction with "gimmicks" - small flat pieces of tinned copper soldered to the board on one end, and hovering over a trace on the other end. You could then use these "gimmicks" to fine-tune the capacitance.

    Use Elsie to re-create my simulation, and experiment in the Plot routine seeing what happens when you increase/decrease values of various components. This will help you to determine what's off kilter in the real-world filter.

    Radio Shack also stocks magnet wire in 3-packs. The AWG-22 would be a good choice for winding your own coils. Single-layer air coils will do the trick here. Drill bit shanks work well for coil forms. If you need a hand with this, tell me what size drills you have and I'll see about calculating some inductors for you. If you don't have any drill bits, go out and buy a few. A numbered bit set (sizes #0 through #80) will give you the most choices. A 1/4" set would have enough to get by. If you're short on funds, just pick up a 1/8" or better yet a 3/32" drill.

    Learning to make/tune filters is not easy. Elsie can be a big help.

    Another tool that's very helpful is HP's "AppCad" freeware.
    Download it here:
    http://www.hp.woodshot.com/

    You can use it to figure out things like stripline/microstrip transmission line dimensions.
     
    Last edited: Jul 3, 2010
  4. marvelouzone

    Thread Starter New Member

    Jul 2, 2010
    2
    0
    thanks a million, you've been a big help, next week i will start experimenting, and i will let you know what i come up with
     
Loading...