Cavity Filter Episode 4

The length of the centre resonator that was too long has been trimmed and the cavity flipped upside-down in order to solder the top end plate in place. I later found out that I had to remove the upper lid again since I apparently had produced a short circuit at one of the inner hairpin couplers while soldering it in place

Since  the whole cavity filter is made out of copper, the tubes were conducting the heat so well - the bottom joints were melting (meaning solder running down inside the cavities). I had to remove a lot of excess solder and redo the bottom joints once the metal cooled.

The picture above shows the cavity filter with the top end plate in place and the tuning screws inserted.

The centre frequency of the filter is at 1026 MHz and I tried to tune the screws in such a manner, that the left cavity corresponds to the left peak, the next cavity to the next peak and so on.

The cavity is supposed to have its centre frequency at 1013.3 MHz, so that's where the tuning begins.

Now, gently move the humps so that the power level has its maximum at the centre. This may take some time and it's a good idea to bring some patience.

Once satisfied with the tuning, simply lock the tuning screws in place with the nuts. I decided to seal them with hot glue although I will probably tune the filter again once it is placed in the signal path. The network analyser I used to tune the filter may have fairly accurate impedances of 50 ohm, but the mixer stages in the signal path of my analyser most certainly will not.

A first glance at the 3 dB bandwidth: 1.10 MHz with an insertion loss of 8.50 dB. Since the loss is rather high and the filter skirts steep, I would assume the cavities being under-coupled. I tried to have the hairpins  very close to the cavity walls so that less energy is coupled.

Now for the important numbers, the rejection at 1024 MHz (the second local oscillator) is 95.4 dB and at 1034.7 MHz (the in-band image frequency) it's 106 dB, which is what I have been hoping for. Very nice!
.

The group delay is not as good as when the cavity was critically coupled but still much better than if it was over-coupled. It apparently makes the filter a little easier to tune as the elements interact less than at over-coupling. The resulting slightly higher loss is not a big problem in the spectrum analyser but it may be in low noise receivers.

The shape factor is equal to the 60 dB bandwidth divided by the 3 dB bandwidth, i.e. 7.76 MHz / 1.1 MHz = 7.05 which is really really good. Shape factors of 11 are very common and it may get as low as 3-4 with digital filters.

My filter is ready and the results are:

Centre frequency: 1013.3 MHz

Insertion loss: -8.57 dB

Bandwidth(3 dB): 1.10 MHz

Bandwidth(6 dB): not tested

Bandwidth(30 dB): 3.25 MHz

Bandwidth(60 dB): 7.76 MHz

Shape factor: 7.05

Rejection(1024 MHz): 95.4 dB

Rejection(1034.7 MHz): -105.4 dB