LPKF Machine... When is it no good?

I work for a company that has done a lot of S-Band work, but not much at higher frequencies. The LPKF machine seems to produce good results at S-Band, but I tried to design some edge coupled filters at X-Band that were much lossyer than the 3D EM simulations. I suspect this is because of the surface roughness of the routing process, but I can't say this with much certainty. The fact is, our cal kits are ancient and I can't even guarantee I'm getting a good calibration at X-Band. Has anyone had luck with complex structures, like filters, at X-Band on a routing machine? What about something simple, like a transmission line, open, or short? At what frequencies do these structures become impractical on a router and how much higher in frequency can you go with a boardhouse? Thank you to anyone who can pull from their experience. -Dan

You didn't say if the center frequency and shape of your filter was correct. If the dielectric constant is different at X-band, that could account for the apparent increase in loss. Also, these filters will radiate some energy. If you can pass your filter thru a tunnel shield (<1/2 wave wide) you can prevent radiation. I've used the TTech machine for X-band designs.

Thanks for the response edf and its good to know someone has had some luck using routers at high frequency. The center frequency was close and the passband was mostly correct, but the loss was very high (passband below -20dB if I recall correctly). This particular filter was a 6 order edge coupled design and had some very tight spacing necessary for sufficient coupling. I think this is where the router had issue. I just got some boards in etched professionally on the same material. They turned out very well. Within 1dB of the expected loss (with connectors, which I did not model) and maybe 100MHz lower. The etched boards were eneg plated, which I also didn't model, so I'm thinking maybe this caused the frequency shift. I suspect lower orders and looser spacing might be acceptable on the router. I was concerned that maybe the surface roughness was just too great at X-band.

I always adjust my design parameters so that there are now gaps less than 9mil!

Then it is unlikely that roughness is the issue. Roughness causes both losses and extra surface impedance (inductive), so it will de-tune your filter in addition to losses.

I would look for some small gap somewhere in the feedlines, or bad soldering of connectors.