Filter hacking for 2395 ISS TV

[g4eml]

I have just finished building the 20MHz bandwidth S53MV design that uses 40 x 40 x 2 tube and is 280mm long.

After a quick tune on 2395 MHz the results look as expected. A nice shape and a -3dB bandwidth of about 17MHz. Insertion loss about 0.3dB. The -30dB bandwidth is about 57MHz.

I used Gareth's idea of 3D printed end plugs with copper tape on the inside faces and these seem to work fine and look good.

I will try to take some plots and pictures tomorrow.

Colin G4EML

[G4IUG]

I will try to take some plots and pictures tomorrow.

Yes please

Thanks Dave G4IUG

[g0mjw]

They were in Colin's earlier post Dave. Here is the link again http://s53mv.s5tech.net/cavity/cavity.html

Mike

[g4eml]

Sorry I got distracted by other things yesterday. Here is is a picture of the NanoVNA screen. (I can't work out how to do a screen capture!)

IMG_20251107_112335.jpg (288.03 KiB) Viewed 134 times

And here is a picture of the completed filter.

IMG_20251107_112818.jpg (103.4 KiB) Viewed 134 times

Colin G4EML

[radiogareth]

Nice job Colin
Right, back from a brief holiday in Belfast (excellent BTW) and have ordered some 40x40x2 tubing. Cant find any 40x30x2 but as Mike suggested, 'other sizes may work'. Already got 6.5,8 and 9.5 alloy bar and SMA's so i can progress the build.
Based on earlier comments about filters in series and unexpected results, it might be better to build a 10MHz wide filter rather than a pair of staggered 20MHz wide, one tuned high, one tuned low to get the reduced bandwidth of interest. "Similar loss" either way in an ideal world????
Next question...where best to place a preamp? Given the size of the 5g signal, it would seem wise to put any gain after the 5g and wifi has been curtailed a bit, but will the insertion loss be a deal-breaker? At around 0.4dB I'd guess not, but happy to be told otherwise.
Gareth

[g4hiz]

Yes, Colin's filter looks very nice!

Regarding where to put the filter. What is the impedance of your feed and the input of the LNA across the frequency band of interest, and that means not only the passband but the rejection bands also? If there is a good match, you may be lucky and retain something that looks like the VNA sweep. Otherwise, the filter shape can change somewhat. The coax connections between the feed and filter input and filter output to LNA input will also add to the mix in terms of the impedance transformations potential created by them. In the worst case, the LNA may oscillate at some strange frequency due to the impedance appearing at its input.

It also depends on what LNA you intend using. The VLNA as designed by G4DDK was from my recollection, intended for contest scenarios, meaning potentially high local signals. From the experience of using one in my own system, I have not had any issue putting it in front of the filter.

In my system, the VLNA connects to the POTY feed using an SMA barrel. I have used attenuators at the input and output of the filter to improve match and stability, especially with the following amp in mind. The LNA gain should compensate for these.

So I'd put the LNA in front of the filter.

Jen

[g0mjw]

Nice job Colin
Right, back from a brief holiday in Belfast (excellent BTW) and have ordered some 40x40x2 tubing. Cant find any 40x30x2 but as Mike suggested, 'other sizes may work'. Already got 6.5,8 and 9.5 alloy bar and SMA's so i can progress the build.
Based on earlier comments about filters in series and unexpected results, it might be better to build a 10MHz wide filter rather than a pair of staggered 20MHz wide, one tuned high, one tuned low to get the reduced bandwidth of interest. "Similar loss" either way in an ideal world????
Next question...where best to place a preamp? Given the size of the 5g signal, it would seem wise to put any gain after the 5g and wifi has been curtailed a bit, but will the insertion loss be a deal-breaker? At around 0.4dB I'd guess not, but happy to be told otherwise.
Gareth

I would use a different design with more poles. Look here https://www.w1ghz.org/QEX/Waveguide_Int ... ilters.pdf and at the calculator program linked here https://www.w1ghz.org/10g/Filter_articles.htm but it requires old windows. I think you could simply get away with a 5 or 6 pole based on the 3 pole with 2 or 3 elements added of the same design. Don't try to cascade two, it won't work.


Mike

[g4eml]

The 10MHz BW design from the S53MV article is actually shorter which seems a bit strange. The pole spacing is shorter but they are using 8mm diameter resonators instead of 10mm in 40 x 30 tube so maybe that reduces the coupling.

I may try retrofitting 8mm resonators to the current one and see what changes. ChatGPT seems to think it will reduce the bandwidth by about 30% but I am not sure if I believe it.

I may also make a 5 pole version of the current design, it will be quite long though! I did buy 2m of tube so I have a bit to play with.


Colin G4EML

[G4IUG]

Hi Mike et all

Apparently you showed your new filter yesterday ? Can you give some details if you liked it please ?

I can get intfil.exe to run on an XP Emulator but you probably have the results anyway ? And in any case no intuition of what to enter.

And finally did you build it yourself ?

Or if you were me, would you do something differently?

Thanks Dave