Filter hacking for 2395 ISS TV

[g4eml]

Gareth, interesting results. I had noticed that the ends don't seem to affect the filter itself but didn't think about external interference entering that way. In that case adding metallic tape to the end plugs might be worth it.
I have such terrible mobile coverage here that I doubt I would notice anything.

The length of the ends should not be too critical.

My finding about the brass screws seemed to be quite repeatable. There was only a small difference but the brass screws always produced less loss. Toolstation do M4 * 20 brass in a bag of 100 for £5.

Colin.

[g0mjw]

If the distance from the input/output to the first/last resonator is similar to or less than the distance to the ends, then they are obviously going to need covering, but that will impact the impedance. If the distance from the input probe to the end is much more, I suspect it won't make much difference to the match, but it is still possibly a good idea.

These dimensions produce a cut-off frequency of around 4 GHz. The calculator Tom, WA1MBA developed indicated a loss at 2.39 GHz of about 560dB/metre. That's a lot but if it were to be say 5cm, it would be only 25 dB of loss - assuming you can scale it that way, which might not be the case. While 25 dB is still a lot, it isn't compared to what we are expecting from the filter, so best either make it long enough or make it a short circuit.

Aside:

In the filter I have, the distance to the end of the guide is double the distance between the resonators, so it isn't going to impact the matching, but I suspect it will be a problem for 5G/6G.

The cut off of the waveguide is ~4 GHz. As well as signals above 4 GHz getting in, the second TX harmonic will propagate in the waveguide. While hopefully, any harmonic energy won make it through the filter, (it might) it will make it out of open ends, regardless of how far the spacing is. So, if the filter is also intended to prevent TX harmonics escaping, it isn't likely to work very well at that and they will come out of the filter, not the antenna.

For 13cm, the significant harmonics are probably the 2nd and 3rd, 4.6-4.8 GHz, 6.9-7.2 GHz. 4.6-4.8 GHz is used for aviation and military applications and also targeted for 6G mobile. 6.9 - 7.2 GHz is planned for 6G mobile and Wifi. Interference to either isn't a good idea, so, a) use a low pass filter after the PA and b) it is probably best to cover the ends, especially if you plan to put the filter up on the mast.

Mike

[g0mjw]

Here are the results from my 5 resonator version in 40*30 mm tube.

Resonator spacings 49 51 51 49, all other dimensions the same as the three resonator version. Total length becomes 310mm.

IMG_20251114_211741.jpg

Steeper sides than the 3 resonator version but the through loss is close to 2dB and it is more difficult to align.

I did also try replacing the brass tuning screws with steel ones. The losses increased by about 0.1dB per screw. So if you are looking for the lowest loss I would suggest using brass screws.

Colin G4EML

I did some calculations with the help of Google AI to give me the coupling coefficients for a 5 pole with the same tube and 10mm rods. It came up with spacings of 51mm and 56mm for an interdigital Butterworth filter. Repeating that for a Chebyshev 0.1 dB ripple gave 49mm and 47mm.

I did a sanity check for the filter I have which gave results very close to what it actually is, so hopefully the calculations are correct. I might find time to build one tomorrow.

Mike

[G3GJA]

Gareth

Very basic tests across the shack didn't show the filters affecting Rx capability, even with both in series. What DID seem to affect the reduction of 5g was screening the ends of the filter as if left open, rotating the filter in azimuth caused changes to the 5g signals. It must have been sneaking in the open ends?

I found the same thing with the filters and duplexer used in GB3EY made by Erwin at ID Electronik Gmbh. I made copper friction fit end plates for the open ends of the duplexer to get the isolation up to the spec sheet supplied with the filter. I suspect leakage from the Tx was getting into the ends.

Erwin now supplies all of his filters with the ends sealed.

Clive G3GJA

[g4eml]

Just for future reference here are the drawings for my 5 pole 10 MHz wide filter. Adapted from the S53MV original design.

10MHz 5 Pole.pdf

(1.3 MiB) Downloaded 96 times

Colin G4EML

[g0mjw]

I built my 5 pole interdigital. Initially 28mm posts reduced to 24mm as it worked well but at 2.2 GHz. I am using stainless tuning screws, not ideal and the loss is high, 3dB. It is also very narrow but wasn't hard to tune. 10 MHz bandwidth, 30 MHz -40 dB bandwidth.

13cm_5_pole_filter.jpg (92.66 KiB) Viewed 329 times

- drawing in the PDF. The rods are 10mm dia 24mm long. The input and output are rods of the same size - I tapped the top of them to M3 and used some brass studding to get to the N-connector. M2.5 would fit better, but I didn't have any to hand.

Very easy to make on a lathe, you could probably do the same with a hacksaw, file and less impatience.

Here is photo - yes it is very long.

13cm_filter_image1.jpg (593.49 KiB) Viewed 316 times

End on view:

13cm_filter_image2.jpg (341.98 KiB) Viewed 316 times

Mike

[g0mjw]

Sorry for yet another post. The brass screws arrived today. I re-calibrated and re-tweaked.

Some results. Firstly, the 5 pole filter isn't symmetric, that's simply because the match varies between forward and back. Ideally you want to measure S11, S22, S21 and S12 all at the same time to tune up, but that needs a full VNA. Simply turning the filter around achieves the same, but you can't see all 4 at once, so need to iterate. The more poles there are, the more this probably matters when tuning up.

This is the response of 5pole above but this time with brass screws. I optimised it for minimum loss at 2.4 GHz, that's adjustable of course. Centring on 2.395 GHz would be what you would want for HAMTV. This filter is narrow enough for it to matter.

2400_5pole_S21.jpg (81.45 KiB) Viewed 59 times

reversed:

2400_5pole_S12.jpg (81.47 KiB) Viewed 59 times

Here is the original 3 pole:

2400_3pole_S21.jpg (81.25 KiB) Viewed 59 times

Something to note - the loss of the 5 pole is now about 1.5 dB. Brass screws make a difference vs aluminium. That's because it is so high Q.

The loss of the 3 pole is about 0.6 dB. Quite an improvement. It already has brass screws but is is also wider and has fewer poles. Conclusion, if you only need 3 poles, only use 3 poles.

The 3 pole has a wider pass band, 20 MHz vs about 10 MHz for the 5 pole. It could be made narrower, with wider spacings, but it would then be more lossy. The -20 dB bandwidth of the 3 pole is 60 MHz. By that time the 5 pole is getting on for 70 dB down. The 5 pole -20 dB bandwidth is 20 MHz. If you happen to have a strong 5G signal up to 2.39 GHz, it might need the additional filtering you get from more filter poles.

Both these filters would probably reduce the 5G interference sufficiently in most locations as long as there isn't a base station locally, which was the issue we found at CAT25.

Mike

[radiogareth]

Thanks Mike, all interesting and useful. Is there any benefit between a comb style and an interdigital? Other than easier to make the feeds on a comb design....
Ordering up more tube because I like making things
Gareth

[g0mjw]

Thanks Mike, all interesting and useful. Is there any benefit between a comb style and an interdigital? Other than easier to make the feeds on a comb design....
Ordering up more tube because I like making things
Gareth

Nothing significant for this application. There is no difference in making them, it is just which side you have clarence holes for the rods vs tapped holes for the tuning screws. Might be easier to tune or mount if all the screws are on the same side.

Mike