I'm busy with a direct conversion receiver for the 20m band, using a VXO and a 14.060 MHz crystal.

I designed a bandpass filter with the following components, using this design calculator. 14MHz direct coupled BPF

Someone gave me a nanoVNA to check the shape of the filter and after calibrating I got the following result.... nanoVNA screenshot 14MHz BPF The shape looks fine, but the insertion loss is -14 dB and the SWR 1.4. I wonder if this is typical and/or acceptable or if I can implement some improvements.

Here is the Smith chart (I don't have any knowledge of how to read Smith charts)....

14MHz BPF Smith chart

And lastly a photo of the filter I built....

enter image description here

Any comments and suggestions welcome.

  • 1
    $\begingroup$ Great work Hans (and good luck with the receiver!) I'm certainly no expert in filter design, but I've built a few double tuned filters like this over the years. Couple of suggestions: try simulating it in a simulator of your choice (I use ltspice, but there are others). Also, the application elsie (pun intended;-) has some very useful filter graphs that show you what to expect. It's nice to be able to tweak the values and watch the impact (you can do this in ltspice also). $\endgroup$
    – Buck8pe
    May 23 at 9:09
  • 1
    $\begingroup$ Also, I find that C5 is critical when it comes to insertion loss (you should be able to see this in a simulator). 1pF is a tiny value and I find that realistically you end up with a bigger value if you go with a ceramic and factor in parasitics. $\endgroup$
    – Buck8pe
    May 23 at 13:09
  • 1
    $\begingroup$ "...the shape looks fine..." True, it has a bandpass shape, but -14 dB gain (loss) and the single-dip S11 suggest it is under-coupled. As @Buck8pe suggests, you should increase C5 until the bandpass shape flattens slightly at its top. I think you'll find that gain improves. $\endgroup$
    – glen_geek
    May 23 at 13:35
  • $\begingroup$ These are all pretty good comments (from what I can tell, because I am well out of my depth with this question) but I bet at least some of them could be expanded into an answer -- even a partial one. $\endgroup$ May 23 at 14:35

2 Answers 2


This filter design looks good, but you should be able to tweak it for less passband loss...
A LTspice simulation of this double-tuned bandpass filter has been done with a guess for inductor Q of 156 on those T50-6 toroids. This requires adding a 0.9 ohm series resistor to each inductor.
Imperfect components do two thing:

  • Pass band loss gets worse
  • Coupling capacitor (C5) needs a slightly larger value for critical coupling.

With perfect components in a critically-coupled filter, a 50-ohm generator and 50 ohm load will yield a -3dB power gain. That's the best you can do (half the generator power is dissipated in the 50 ohm source, and the other half of the power is dissipated in the 50 ohm load). Note that in this simulation, output voltage is plotted (not power), so a perfect filter would show a -6 dB gain. voltage output from filter plotted as a function of frequency

  • With C5= 0.5 pf, this filter is under-coupled - a bit lossy
  • With C5= 1.0 pf, coupling is near the critical point
  • With C5= 2.0 pf, it is over-coupled resulting in a wider pass-band.

In this voltage plot, those lossy inductors yield a passband voltage gain near -10.4 dB instead of perfect -6 dB.

Dealing with 1 pf capacitors is a bit tricky, and trimming them is difficult. If you replace C5 with a tiny trimmer capacitor, coupling to other components can change tuning. For example, if you were to add a shield, you'd have to retune the whole filter.
You can possibly add a "gimmick" capacitor in parallel with C5 consisting of two insulated wires twisted tightly together so that there is capacitance between them - the pair perhaps two or three cm long. It is easy to cut the pair to reduce coupling.
An alternative: Replace C5 with a 3-capacitor network. Assume you have a spare 30pf trimmer. This can be used to trim the coupling between the two resonators:


simulate this circuit – Schematic created using CircuitLab


Thank you Glen for your insight, especially the under- and over-coupling parts. Your simulation seems to be spot on, because...(confession coming up)...

  1. I was actually using a 0.68p coupling capacitor (instead of the 1p in my design) which gave me the -14 dB gain as you can see in my first post, which corresponds to the 0.5p capacitor in your simulation.

  2. As an experiment I put another 0.68p capacitor in parallel with the first one, but still got -14dB gain.

  3. With a 1p+0.68p combination I saw the first traces of over-coupling.

  4. Using a 1p capacitor only I got the following readings.....

enter image description here

Still a 12dB insertion loss, but slightly better than yesterday. The return loss (yellow line) is much steeper now at -50dB, but not quite at my desired frequency.

If you state that a perfect filter would have a gain of -6dB and in practice (including losses) -10.4dB, then my filter doesn't seem so bad with -2dB below the gain in the real world.

So for now I'm happy with it, but the next step is to experiment with other construction methods to see if I can reach that -10.4 dB gain. Cheers.

  • $\begingroup$ Am unfamiliar with nanoVNA operation. Its calibration procedure quite possibly accounts for the source resistance vs. load resistance loss of -3 dB power loss. Also, your nanoVNA quite likely plots dB as power ratio, whereas LTspice plot Y-axis is actually dBV. I can do this because Rs=50, Rload=50, and Vin=1V. So don't confuse the plot's -10.4 dBV with nanoVNA -12.7 dB. Your filter may still have excess loss. I'd suggest verifying that nanoVNA is well-calibrated. Another Hans built a similar 20M filter qrp-labs.com/bpfkit.html showing -1.75 dB insertion loss. $\endgroup$
    – glen_geek
    May 24 at 12:04
  • $\begingroup$ Yes, I saw the qrp-labs bpf kits and am considering buying some to experiment with. They use inductive coupling, not the capacitive one I am using. I'm also new to the nanoVNA, so lots of things to learn and since it is only a "cheap" VNA I don't expect miracles from it. I followed the calibrating procedures found elsewhere, e.g. from W2AEW's excellent YouTube videos (see his video #315 for example) and results are consistent, so that is reassuring. He measures an LPF and the insertion loss with that is close to 0, so yes, I do still need to do a lot of experimenting and testing. Cheers! $\endgroup$
    – Hans Fong
    May 24 at 13:07

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .