Component values
C1 80pF Transmitting Mica 350Vac   L1   5 Turns 7mm ID 1mm dia wire
C2 80pF Transmitting Mica 350Vac    L2   6 Turns 9mm ID 1.5mm dia
C3   27pF Tubular ceramic   L3   5 Turns 7mm ID 1mm dia
C4  6.8pF Tubular ceramic   L4   5.5 Turns 7mm ID 1mm dia
C5  2.7pF Tubular ceramic   L5  5.5 Turns 7mm ID 1mm dia
C6  6.8pF Tubular ceramic                                                   .

Alignment:-
The easiest way to set the filters up without a network analyser, is to connect a good 50 Ohm load to the antenna port and adjust the spacing between the turns of the coils to achieve a good VSWR.
I would recommend using low power and buffering the output of the transmitter with a suitably rated attenuator, of 6 to 10dB attenuation,  placed right on the TX output. This will ensure that the match (return loss) presented to the transmitter cannot get worse than twice the loss of the attenuator.

Performance:-
The prototype unit has the following characteristics
Insertion loss to antenna  0.5 dB for both ports  Return loss 17dB in amateur bands
Cross loss 144 MHz appearing at   50 MHz port -65 dB
Cross loss   50 MHz appearing at 144 MHz port -75 dB

Notes:-
L1 is primarily to improve the input match.  C3 L3 form a phase shift network and are in the circuit to ensure that the impedance of the 50 MHz filter as seen at 144 MHz at the antenna port is on open circuit.
The original photo on thise site was of the prototype which has been in use for many years. I have since beefed it up and the result is as seen. I have run this with 50W at 2m and180W at 6m (though not simultaneously)
This technique allows filters to work up to over 500 MHz, and a number of 2m 70 cm diplexers are in use in Scotland. When I can recover one to photograph, I will include the details here.

See information on Starpoint combining
 
 

Back to 6m page