# Splitting Hairs With Crystal Filters

So currently I have more crystal filters than you can poke a stick at. Continuing on from my last blog post, I have spent hours soldering and de-soldering the transformers on the CW crystal filter, both trying to improve it somewhat and to also understand what I am seeing and how it reacts to different impedance transforms. Bear with me, this could end up being long and boring.

So for those playing along at home, you might recall my rather average looking plot for the CW filter, not the best stop band, quite a lot of loss and I was never happy with it. Now the theory goes, that the input impedance of the Cohn Min Loss filter is in the ballpark of 200 ohms, so throw a 4 to 1 impedance transform at it and rock on Johnny, it will be ok. Not so with this filter.

What i did with this plot was terminate one side of the filter with 50 ohms and then ran an S11 on both ends of the filter. You can see by the plot and the one below this that maker one is around 75 ohms and it close to being within the bandpass. However, you have these other 2 peaks either side of it at around 12 ohms. I did not know what to do here, how do you raise the low impedance while lowering the high impedance? This goes against the laws of physics of course, as the impedances are going to track with what the transformers are doing.

The first thing that I tried was to use averaging. I figured that if I raised both the 12 ohm peaks to 50 ohms, the 75 ohms would go to 300 ohms, but the average of the passband would be closer to 50 ohms than not. This it turns out was a bad idea and introduced other complex impedance’s in the passband and thus greater loss and reduced stop band.

So i changed tack. Perhaps its what is within the passband that is the most critical here. That peak at close to 75 ohms might be the key as this is a rather narrow filter. So for the 100th time i pulled off the transformers and wound them this time with a turns ratio of 4 turns primary and 5 turns secondary, and you can see with the above plot marker 2 went from 75 ohms to 30 ohms, all i did was split the difference. HAHAHA. The other problem is, it gets rather hard to wind transformers for odd ball values of impedance transform.

But i think that this has put me on the right track. If i pick a frequency that is right smack in the middle of the filter and try and match for its impedance, I should be able to improve the losses somewhat while also retaining the much improved stop band. That extra -10db in the stop band is a massive improvement for no extra loss. However, if I can improve the pass band losses by a couple of db, without degrading the stop band, then that would be better still.

One last thing before i sign off, using the NanoVNA for measuring impedance, now thats a game changer.

## 2 thoughts on “Splitting Hairs With Crystal Filters”

1. Paul says:

Hi Rob, Did you use design software for the xtal filter? The Dishal software has a Chon filter designer. If I put my crystals in there the Z is about 100 Ohm (of cause in theory ..). Anyway I also build some filters with mixed result. But not the Cohn types. Have to try also.

1. Rob Powell says:

Hi Paul,

No I did not use any design software, all I did was put the crystals in an oscillator and matched them for frequency and the amplitude of the sin wave they produced. After that, it was just a matter of adding the caps which are all the same value, they determine the width of the filter. Lower value == wider filter, higher value == narrower filter. The value of the caps can range from 33 to 330pf but its dependent on a number of factors, i just ball park it and then change them as needed to get the desired filter width.

Rob.

This site uses Akismet to reduce spam. Learn how your comment data is processed.