The cheap pa0nhc 2mtrs-repeater duplexfilter.
Questions and possible answers about details of the construction.

 Some general thoughts:

In short some important points:

Mechanics: Solid copper is very expensive. Material alone is costing abt Fl 350! Then plugs etc. It is difficult to get it in shape and soldered/welded together. The same as for aluminium. It is only cheaper. But HF is only running at the surface (skin) of the copper. So why not use more easy to work with, cheaper glass-fiber copperclad? It is easier and more stiff than thin coppersheet.

Make the filter as stiff as possible. This enhances mechanical long-term stability. In this construction all eigth cavities are mechanically one-piece. No separate frame is then needed.
Especially the "top"plate, with the central conductor an coupling-links in it, should be sturdy. This plate is abt. 3mm thick.

How important is the couplingfactor of the coupling-links and how to adjust them?

I did a lot of experimenting. Correct amount of coupling prooved to be crucial for good filter-performance of this type of bandpass- and notch-filter. This was mentioned nowere in the literature i read. Also not how and when to determan and adjust it.

Adjustable coupling is difficult to construct and adjust. Turnable plates, secure in position when locking them, secure in contact, are difficult to make. If once determened, a factory could machine fixed couplings, having the exact right amound of coupling. But how to make this at home?

My couplings do not rotate. I change the effective length, by pulling them out of or pushing them in to the cavity. Then locking them by soldering the "cold" end directly to the cavity. Secure contact and position is garanteed. They only need adjustment ONCE, during the construction of each individual cavity. So why bothering with elaborate constructions?

How long can cavities be for good performance? 

Shortened cavities with capacitive top-loading have often lower Q and stability. So use full 1/4 wave or 1/2 wave cavities, with low top-capacitance. There must be enough room between the bottom-side of the central conductor and the cavity-bottom. This ensures low temperature-coefficient of the cavity.

What size must a cavity have? 

In the article of G8AMG was stated, that when the cavity is seen as a coaxial cable, the characteristic impedance of that "cable" must be 78 ohms for optimum Q of the cavity.
Wider cavities have bigger central conductors. They are more stiff. The cavity is then less stiff. Current-density is lower. So for high transmitter-power-levels they are maybe better, as temperature-effects are smaller. But if this reduces insertion-loss?
I wanted to use readely avalable materials. 22mm water-copperpipe can be bought everywhere, and is not to expensive. Maybe some short cheap pieces could be bought from a contractor, for old-metal prices. Copperpipe of abt 80mm dia. is very expensive and difficult to obtain. So make cavities square. A square cavity has then to have a width of 74 mm. Not to big, so cheaper.

How needed is silver- or goldplating?

Silverplating adds little to low insertion-loss of the cavities, i think. It merely acts as a anti-oxidant. So why expensive and difficult silverplating? Take measures to prevent oxidation of the copper-surfaces. Cheaper and effective.

Tuning-mechanism.

Mostly there is some possibility of finetuning in cavities. This often complicates the construction of the tuning-mechanism.
I use course-tuning of a simple, but stable construction. After all, course-tuning is only done once, during construction of the filter. It then can be forgotten, if stable.
As it is difficult to install and adjust a filter, when (fine-)tuning is in the bottom of each cavity, capacitive finetuning is done here with a bolt on the side of the filter, near the top of the central conductor.

Cables, are they critical?

The lengths of the cavity-interconnecting cables depends on the actual type of notch-cavities. In this design (ARRL) this length must be totally an 1/4 electrical wavelength. This includes connectors (when applied) and the connecting pigtails to the links.

The length of the cables, which connect the outputs of the Tx-filterhalf and the Rx-filterhalve to the antennacable, must be EXACTLY 1/2 electrical wave and 1/4 electrical wave, INCLUDING plugs an connecting wires to the output-links. This is critical. The performance of the completed filter depends on it.

So why use plugs for connecting cavities to the antenna and eachother? They cost money, construction-effort, and couse tolerances on effective cable-lengths and possebely bad contacts when oxidating. I soldered 50 ohms PTFE cable directly and shortly at the required points. It is easy to accurately determen the effective cable length. Braid can be soldered directly to mass, without melting the cable. But remember that PTFE has a different shortening factor than normally used cable.

What do this cables do?

At the transmitter-notch-frequency, the L-notch-cavities in the receiver-chain, have a LOW impedance with a high VSWR to 50 ohms. The 1/4 wave line between the last L-notch-cavity and the antennaconnector, transforms this to a very high impedance. This 1/4 wave cable in combination with the cavity-properties, acts like an open switch for the transmittersignal in the direction of the receiver. Thereby disconnecting the receiver at the transmitter-carrier-frequency from the antenna-connector.

At the receiver-notch-frequency, the C-notch-cavities in the transmitter-chain, have a HIGH impedance with a high VSWR to 50 ohms. The 1/2 wave line between the last C-notch-cavity and the antenna-connector, does not transform this. This 1/2 wave cable in combination with the cavity-properties, act again like a open switch for the receiversignal in the direction of the transmitter. Thereby disconnecting the transmitter at the receiver-frequency from the antenna-connector. Received signals only can go to the receiver. Transmitternoise at the receiver-frequency is isolated from the antennaconnector.

Interconnecting cables between cavities have to be exactly 1/4 electrical wave long, including plugs and connecting wire-pieces. They transform the very low notch-frequency-impedance of the L-notch-cavities to very high. Again this cables act like open switches at the notch-frequency. The same applies for the C-notch-cavities. but form high- to low impedance.

As the VSWR of the cavities at the pass-frequencies is low, the 50 ohms cables have no influence then. They then just act as a interconnection.

Plugs.

Use only plugs of best quality with PTFE insulation. N-connectors are more sturdy than BNC, water-tight and can withstand some strain. So better standarize all connectors to N.

Selectivity. 

Notch-cavities have little selectivity. Some types give no more then 6 dB of suppression of signals exept the notch. So adding a extra bandpass (notchless-) cavity, could be very usefull to suppress wideband transmitter noise, and strong signals from transmitters in the neighbourhood, preventing desensibilisation and the "reception" of strange interference.

Transmitter-sideband-noise can be mixed with transmitter-harmonics, generating a noise-signal on the repeater-input-frequency. This can happen in the transmitter, circulator, bad contacts in the antenna, bad contacts in surrounding metal structures, and the receiver.

So make the transmittersignal clean. Make the receiver-oscillators noisefree too. Best use well designed cristal-oscillators. Use the right amound of drive in every stage of the transmitter and receiver. Make receiver-selectivity good. Add transmitter- and receiver-selectivity by adding filterselectivity.

Is a circulator needed?

Only use a circulator as a compensation for a duplexfilter with insufficient suppression of the transmitter-carrier.
When used, first separately adjust the circulator for BEST VSWR at all ports on the repeater-frequency. Then the port-separation it also optimal.
When using a circulator, the antenna-VSWR must also be PERFECT. All reflected transmitter-power returning from the antenna is directed by the circulator to the receiver! Using a circulator with 30dB port-isolation and an antenna with a reflection-damping of only 10 dB, results in only 10 dB isolation between TX- and RX-antenna connectors of the filter, and not 30 dB.

Remember: circulators can make intermod, crossmodulation and reciprocal mixing (sidebandnoise on mirror freq.)!

So do not use a circulator. Make your duplexfilter, transmitter and receiver good.

See also: the article of G8AMG about cavity filters.