Chebyshev Low Pass Filter 8MHz, 9 poles, 0.5dB ripple
Today I decided to add another pole to the Chebyshev filter with 8 poles I built yesterday, making it a 9 poles filter. One advantage, apart for the sharper roll-off is that I was able to use just one standard value capacitor for the five capacitors used. The ones at the edge are 680pF silver mica and the other three in the middle are 1nF styroflex.
I built the coils so that they were measuring about 1.3uH and 1.4uH (which were 25 turns and 26 turns, wire 0.6mm or AWG22, on my Philips screwdriver with a shaft diameter of 6mm).
I am considering building a band stop filter to put in series with this filter and see if I can improve the rejection over 50MHz.
In practice probably there is no need for this, but I am curious if it will work.
I did not try to adjust the inductors, they are untouched, as they turned out. I tried to avoid two coils on the same axis, to prevent coupling between the coils.
Here is the frequency response. In the meantime I learnt how to normalize the scale of the spectrum analyzer, so that the reference shows 0dB.
The response is excellent up to 50MHz, after which it gets worse.
I made some poor attempts to improve the rejection over 50MHz by using some shields between coils, but it was no better than before.
Maybe the reason of poor rejection over 50MHz resides in the characteristics of the coils. They have parasitic capacitance between turns and a resonant frequency of each own, so maybe they don’t act like inductors at higher frequencies.
I will do an experiment in the future, to measure the response of a single coil for a 200MHz frequency range.
Since this design was made for a 0.5dB ripple (which I understand allows a difference of 1dB between highest point and lowest point), the response of the pass region is better than what I obtained yesterday.
Please note the marked area of the screen, with a vertical drop is due to a fault in my spectrum analyzer.
The following day I decided to try to identify the reason why there is poor rejection (only 30dB) toward 100MHz. My bet was either on the parasitic capacitance between turns for the coils or the inductive coupling between coils, since they are close to each other. I had the idea to simulate the filter in LTSpice, but to add various capacitances in parallel with each of the coils. It became quite clear that neither the coil capacitance nor the inductive coupling were responsible for the poor rejection. It had to be another reason: the lengths of wire between the BNC connectors to the board and the unnecessary long terminals of the filter components. So I decided to shorten all of them. For good measure I tried some shielding as well, but the effect was minimal up to 100MHz.
So, now the filter board looks like this:
And the frequency response is presented below. You can notice the worst rejection up to 100MHz is better than 63dB. The roll-off of the filter is almost 80dB in 8MHz.
For curiosity, I checked the rejection up to 500MHz. It looks like this:
And a quick test of the 7MHz oscillator designed by K7HFD together with the Chebyshev filter:
The response is a pure sinewave signal, with the harmonics and non-harmonics suppressed by about 84dB.