2
\$\begingroup\$

I have been trying to make a 5th-order Chebyshev filter going from 500 MHz - 570 MHz and a 3rd-order Chebyshev filter going from 500 MHz - 540 MHz. This is my first time making a filter into a PCB.

No matter what I do, my physical implementation of it doesn't match my calculations and PSpice simulations.

I realize this circuit doesn't have the most perfect geometry and the parts have tolerances (I chose tolerances that should still work) but it should still be close enough. Why is the pass band so far off? The span on the spectrum analyzer screen is from 1 MHz to 1 GHz. The peak is seen at 350 MHz instead of 535 MHz which it should be.

I'm at my wits end. I have been stuck on this for a month now, please help.

enter image description here

Here is the board geometry. I have confirmed that the values are correct and should work even with their tolerances.

enter image description here

enter image description here

\$\endgroup\$
6
  • 4
    \$\begingroup\$ Welcome! Please post a schematic view of your PCB, what values you have tried, and also what sort of signal is going into your filter. \$\endgroup\$
    – raaymaan
    Apr 26, 2022 at 23:16
  • \$\begingroup\$ raaymaan I have added the schematic picture, i made sure my values were correct of course... just a simple sine wave is going into the circuit.. \$\endgroup\$
    – Phillip Koffman
    Apr 26, 2022 at 23:41
  • 2
    \$\begingroup\$ This is just a thought, but you may be pushing the limits of your inductors with how high your passband frequencies are. I would speculate that their SRFs are below your cutoff frequencies which may be causing the unwanted frequency shift. Hard to tell though without data sheets. \$\endgroup\$
    – Ryan
    Apr 27, 2022 at 1:01
  • 2
    \$\begingroup\$ A valuable quote for RF design: "all parasitics are low-pass." That your measured pass-band is shifted low in frequency compared to simulation strongly suggests to me that your model is missing some parasitics. I can't tell what the components in your lumped schematic is supposed to represent, but if you're not accounting for stray parasitics in your PCB, that's likely the first place to check... \$\endgroup\$
    – Shamtam
    Apr 27, 2022 at 2:53
  • 1
    \$\begingroup\$ ...try adding some extra shunt capacitances to your schematic and see how much you need to add to match your measured response, then ascertain whether that amount seems like a reasonable contribution from your PCB. \$\endgroup\$
    – Shamtam
    Apr 27, 2022 at 2:56

1 Answer 1

Reset to default
0
\$\begingroup\$

Each track has both inductance, capacitance and resistance and thus a characteristic impedance from Z^2=L/C

I believe you only have trace inductance. That will not work.

There are ways that do work. I have always relied on coaxial resonators and SAW filters or coaxial 1/4 shorted notch filters.

enter image description here REF

Note the Q's in this filter > 1000 and the quality of the return loss.

https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8502815

\$\endgroup\$
13
  • 1
    \$\begingroup\$ They should not be only inductance. Are you going to do what I suggest? analyze each trace and compare with your schematic then show result in question \$\endgroup\$
    – Tony Stewart EE75
    Apr 27, 2022 at 0:35
  • 1
    \$\begingroup\$ hmm even air is a capacitor \$\endgroup\$
    – Tony Stewart EE75
    Apr 27, 2022 at 0:37
  • 1
    \$\begingroup\$ I think you should show a better schematic and layout identifying all values, parts , trace impedance assumptions and specs \$\endgroup\$
    – Tony Stewart EE75
    Apr 27, 2022 at 0:39
  • 1
    \$\begingroup\$ Go look inside a TV tuner.. They use air coils and PTC caps. IN FM Rdaos they use trimmer ferrite coils and air coils. You still need to learn more about properties of materials. The NTC of L is compensated by choosing PTC caps instead of NP0/C0G. That's why I used ceramic resonators and SAW filters and COAX stubs for VHF/UHF and stripline for > 1GHz on low loss tangent FR4 \$\endgroup\$
    – Tony Stewart EE75
    Apr 27, 2022 at 0:47
  • 1
    \$\begingroup\$ I think you should understand what does work and why 1st \$\endgroup\$
    – Tony Stewart EE75
    Apr 27, 2022 at 1:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.