What type of filter should I use to fix my microwave filter problem?

[waht]

I'm working on a bandpass filter at 3.2 GHz. I built a few prototypes interdigital, pipecap cavity, and even microstrip.

Microstrip filter sort of works but it can't be tuned, so I abandoned it.

The other types also work, bandwidth and insertion loss is ok but they all exhibit garbage responses from 6 to 20 GHz, with some peaks having similar insertion than the fundemental.

So I'm looking for a way to either build a new type of filter, (mechanically tuned) or match impedance or something. This is pretty much a homebrew design.

Appreciate any info.

 

[Averagesupernova]

How did you do the cavity type filter? If you are building it how I think you are then the odd harmonics of the fundamental will have a significant response. In any case you will probably have to do multiple stages such as a low-pass after the first bandpass. Some interesting things can be achieved using transmission line stubs

 

[berkeman]

I'm no expert at these frequencies (yet), but it sure seems like the microstrip could be tuned. What properties of the microstrip determine its filter characteristics? If it's a homebrew circuit and tuning, can't you do a little shaving of the copper to tune it? What are you using in the tuning process (do you have access to a network analyzer?)?

 

[waht]

Cavity filter is basically a 1.5" copper pipe cap soldered to a FR4 pc board, with two SMA UT141 jumpers inserted into the cavity making a capacitive coupling by 1/8" probes extending from the jumpers. The cavity is tuned with a screw.

This is a common design popular among hams. Like this one

http://www.czd.org.uk/astro/radioastro/filter/index.html

I found that it works beautifully at the frequency of interest, but as you go up a couple of octaves in frequency, really bad things start to happen. Spikes up and down, as if it was acting like a high pass filter.


Same thing with my homebrew 5 element interdigital filter, that is enclosed in metal box, works great at the my frequency, but higher it looks so ugly.

Microstip filter that I did, I was close to my frequency, but to my surprise no other response was detected through 20 gigs. But I found even if you are millimeter off, it deteriorates and I am unable to optimize it. It's very difficult to tune it, even by taking bits of the trace. I think microstrips are good when designed by an em cad.

For tuning I just use a sweep oscillator and a diode detector, crude approach, but it's a homebrew so I don't complain.

I think a simple solution would be to add a low pass filter, but it's just as difficult, I think it has to do with band impedance matching, I don't know how to intergrate a stub or something.

thanks

 

[berkeman]

Cavity filter is basically a 1.5" copper pipe cap soldered to a FR4 pc board, with two SMA UT141 jumpers inserted into the cavity making a capacitive coupling by 1/8" probes extending from the jumpers. The cavity is tuned with a screw.

This is a common design popular among hams. Like this one

http://www.czd.org.uk/astro/radioastro/filter/index.html

I found that it works beautifully at the frequency of interest, but as you go up a couple of octaves in frequency, really bad things start to happen.

Good stuff, what. But you won't find a physical filter that works over several ocatves, unless you use the voltage dependent capacitance of a reverse biased diode. Have you tried adding that to your physical GHz filters?

 

[waht]

interesting point, I've never seen a diode used to aid the response of filters. I found out, that a transmission line on FR4 works as a good low pass filter through 6 or 7 GHz. So I think I will go with that.

But the root of the problem is bad impedance match I guess.