Lossy Stripline/Microstrip Filters at UHF

I've designed microstrip and stripline filters at S-Band that have excellent performance. Because the propagation mode is nearly TEM, I expected performance to be even better at UHF. That's not what I'm observing. Both stripline and microstrip are showing extremely high losses compared to the theoretical models. This is true for FR4 and a variety of high and low dielectric Rogers materials I've tried (I'd never use FR4 at S-Band, but I thought it might stand a chance for UHF). These are high order (7 element) filters. I've made similar order filters at S-Band though (at least 5 element). I still see distinct poles in the response, so I don't think its quite as simple as the elements just being de-Qed. The best theory I can come up with is that, because the resonant elements are so much longer at UHF and the 50 ohm lines are wider, my dielectric losses are increasing substantially. Is this the likely cause of my high insertion loss and mediocre response?

I'm using IFilter in AWRDE, then exporting to schematic. I'd move to a 2.5D simulator, but the schematic results don't even meet my design requirements.

When you say "extremely high losses" can you elaborate on the magnitude of losses in dB at the frequency of interest? The majority of insertion loss occurs from conductor losses, not dielectric losses. Unless you're utilizing a very lossy material with poor Df. Also, it's very important to have an accurate surface roughness model for your transmission line. In short, if you're Insertion Loss is too large, go with a shorter and wider conductor/TL. Obviously, you will have to simultaneously alter stack-up dimensions (dielectric thickness) to maintain impedance/Return Loss.

Thanks for the reply Airbag. The losses are about 3x theoretical, without even going to EM. I'm using the appropriate conductor thickness for 50 Ohms. If I make just a straight transmission line, it can be several inches long with negligible insertion loss. It's only the filter where the losses seem to present themselves. For this reason, I am suspecting that it's some effect of propagation through the dielectric, between the coupled elements maybe. I looked mostly at combline and hairpin filters, other topologies had size or resolution constraints. As I've said, I've had good results with these topologies in the past at higher frequencies, and was expecting better performance at UHF.

I've explored a resonant coax filter that looks more promising, but will be labor intensive. I'd still like to understand my problem of course.

You can try to increase or decrease the dielectric thickness.If you feel the microstrip lines are too wide, decrease the dielectric thickness.It might make your filter behave better.

You mean they are longer, right?
They should be be the same width regardless of freqency if you compare it to other designs using the same board thickness and material.

It's hard to understand directly why you see higher loss in your UHF design compared to your L band designs.
It should be as good or better at UHF, dielectric loss should not be higher at lower frequencies.

Thanks JC. I will keep that in mind.

Harry, that's correct, the lines are the same width regardless. I think that what I was trying to say is that the resonant elements of the filter are spaced wider apart compared to the higher frequency designs. My thought was that maybe this increased the losses due to dielectric losses. That doesn't make as much sense in hindsight though. I think dielectric losses are actually a factor of wavelength? That would make sense given loss tangent as the real part of permittivity.

Given that, I'm really not sure why my design was so lossy. Maybe I did need to just try different heights to change the line width. Is there something about the propagation mode I don't understand?

make sure you are using 1 ounce or thicker copper. Skin effects come into play at UHF frequencies. Also consider you might have radiation losses, where part of the structure is acting like an antenna.

other than that, post your schematic and artwork and we could comment on specifics.

I have seen rarely Microstripline structures in UHF band except some Baluns,Coupling Devices and some sort of LP or HP filter circuits.Because wavelength is long and it creates such losses ( Radiation Losses,Metallic Losses,Substrate Losses).The ideal is that to use very high di-electric coefficient substrate like Alumina ( or better) in order to shrink the structures' dimensions to prevent such losses.But it will cost you..

I think that you guys probably hit the nail on the head with radiation losses. I was trying to make a typical hair pin structure like the one here. Several elements long, so losses would definitely stack up quickly.

The only thing that perplexes me, why is radiation loss (and the other losses you two identified) more pronounced at UHF than higher frequencies? Is it because the resonant elements, length chosen based on the substrate, are more efficient radiators in air compared to high frequency designs? That would cause less energy to be contained in the microstrip, if its true.