[Radio Wrangler]

The pattern of the values of top coupling capacitors between resonators sets the filter shape, Butterworth, Chebyshev, Gaussian, etc.

The scaling of the coupling capacitors (with respect to the resonator capacitors) in cahoots with the load and source impedances coupled into the ends of the filter - to set the Q of the first and last stage all together set the bandwidth (with respect to the centre frequency)

The tuning of the resonators sets the centre frequency.

These 'coupled resonator' filters start out with all the resonators identical and tuned to the same frequency (geometric mean of the 3dB points of the finished filter) but the piulling effects of the coupling capacitors hve to be taken into account (consider all resonators but the one you are looking at to be shorted) and trim the one you're looking at to the planned centre frequency. Do each one in turn with the others shorted. And you now have the filter resonators adjusted. With the shorts off, each resonator gets pulled off of its centre frequency, and its Q loaded to give exactly the pole-zero pattern for the wanted filter.

The nasty bit is if you start sweeping the filter, you will see as many bumps as the filter has resonators. Aha! you think, each resonator trimmer must control one bump!

No such luck, each resonator controls a pair of bumps and interacts with all the others. The shorting trick is the quick way. Try just twiddling, even with a swept display live in front of you and you'll tend to diverge from the wanted alignment and it just keeps getting worse.

Inside the ceramic filter, someone designed the coupling factors and thence the capacitors. They then offset the lengths the resonators were ground to, to correct the tuning. This is the one setting you can't fiddle with downstream. The bandwidth is designed into the filter. If you vary the source and load impedance, you'll worsen the ripple. If you vary the coupling capacitors, you can vary the bandwidth but you can't fiddle with the resonator trimming to counteract the ripple getting worse.

So, for a filter with a controlled shape, you're stuck wiith the designed-in bandwidth, and you need to be fairly close with the source and load impedance if you want the ripple in-spec.

These hints go for all coupled resonator filters. LC, Quartz, Ceramic, waveguide, helical. as well as all the alternative coupling methods.

David