A home-made coaxial notch filter for the 70cm band.
Rev: 2004 09 13 / layout 20150121.
(system like Kathrein type SP filter).


This article is translated by pa0nhc from Dutch to English after an article from VERON radio club, division Amersfoort.

Materials for one cavity (see construction-drawing):

Pos.No.                     Description
1 BNC connector, flange type, 2 pieces.
2 Screw M3x5, 8 pieces.
3 Copper plate 100x100x1.5 mm
4 Brass plate 40x32x1 2 pieces.
5 Copper pipe, 22mm dia, 145mm long
6 Brass pipe, 20mm dia, 45mm long
7 Brass disk, 20mm dia, 5mm thick.
8 Coil, silver plated wire 1mm dia, 2turns,
(see drawing).

9 Coax RG58U, 125mm long
10 BNC receptacle, 2 pieces
11 Johnson screw trimmer 3-13pF
12 Copper pipe 80mm dia, 220mm long.
13 Copper plate 100x200x1.5mm
14 Brass square, 6x6mm
15 Cylinder head screw M3x10 + lock washer, 24 pieces.
16 Halve clamp coupling for 6mm pype
17 Steel rod, dia 6mm, 220mm long.
18 Steel knob, 20mm dia, 10mm long.
19 Half nut M6

How to build:

    Drill a caliper for pos.3:
Use two steel plates 100x200x2mm, place them on top of each other, and drill in the four corners a hole of 3mm. Screw them together.
Draw all holes and drill them with 2.4mm dia in both plates. Check all dimensions, steel is cheaper than copper!

Mark one plate with "1", this plate will have only holes with dia 2.4mm.

Enlarge the holes in the other plate with a drill of 3mm dia. Do this cross-wise. Mark this plate with the 3mm holes as "2".

    Top plate pos.3:
Use for each set of two cavities 2 copper plates.

Drill holes in the four corners like you did in the steel plates. Mark the top-side of the copper-plates. Put the steel plate marked "1" (2.4mm holes) on top of all copper plates and screw
them all together.

Drill all holes 2.4mm. This way all holes are in exact the same pattern / distance.

Then drill all holes in the copper plates to the correct diameter.

Drill the four holes for the BNC receptacles on the under-side of the plate to match the head of the screw. The screw head must become flush with the copper-surface (see drawing).

Drill the top-side of the central hole for the BNC-receptacle so, that the BNC-plug sits flush on the top side of the plate. Polish the top-plate 3.

    Brass ring:
Clamp the square brass rod pos.14 on a pipe of 70-76mm dia. as follows: Heath the brass-rod dark red, and turn it around the pipe like a spiral. The internal dia now is a bit smaller then
the external dia of pipe pos.12.

Pull the spiral around pipe pos.12, determine the correct length and saw it.

Flatten this ring very good, and solder it to pos.12. Then file or turn it flat on a lathe.

    Holes in ring pos.14:
Draw the holes on the center of the brass ring. Check with the drill-caliper.

Drill four holes 2.4mm and tap with M3. Put the drill-caliper on the brass ring and screw them together. Now dril and tap the rest of the holes.

    Tuning pipe pos.6:
The internal dia of pos.5 is smaller then the outer dia of pos.6.Make them fit as follows:

Put graphite-grease on another piece of pipe with a dia of 20mm.
Heath the 22mm copper pipe red hot over a length of 6cm and press the 20mm pipe in it.
After cooling-down, tap the 22mm copper pipe with a piece of hard-wood until the 20mm pipe can be pulled out. This way pos.6 fits exactly in pos.5.

Saw a few slits in pos. 5 and tap it stiff around pos.6. This way you get very cheap "finger stock" HI.

Polish the outside, and inside where pos.6 comes, well.

    Assembling the parts pos.3/4, 5, and 16:
Use a piece triplex - wood.
Heat a piece of 22mm copper pipe red hot and press it on the wood, burning a ring in it.
Drill a hole of 6mm in the inner of the burned ring (Rem: it must be outside the center of the ring, see top plate drawing).

For the following use two threaded rods M6, length 200mm: Put one rod through pos.16, then pos.3/14, then pos.5, then trough the piece triplex-wood with the burned ring.

After aligning tighten it all with nuts. Position pos.5 correctly on pos.3.

Solder it together wit tin solder containing 3% silver. This type of tin solder is stronger and easier silver-plated.

Polish everything well.

Polish pos.4 and pos.6. The better the polishing was done, the easier silver-plating goes.

Silver plate all polished parts with silver-fluid from a bottle. Rub-on the silver well with a piece of cloth.

This silver-plating is necessary, as the cavities detune after a while due to oxidation (i think).

Screw the steel plate marked "2" on the ring-side of pos.12. Solder pos.12 on pos.13. Clean pos. 12 inside and lacquer it.

Assemble it all.

If the the notch on de rejection-frequency is not deep enough, then change the diameter of coil pos.8 a little.

Per cavity the insertion-loss should be abt. 0.3dB. A notch-deepness of 27dB is then possible.

The designers telephone-No is: +31 (0)33 726837.



    Remarks (pa0nhc):

This filter can be used in repeater duplex filters or to resolve receiver overload problems due to very strong signals near amateur frequencies. This pass-reject cavity resonator is usable for 1,6 MHz "shift" (the difference between pass- and reject frequencies). The reject frequency can be below or above the pass frequency. Adjustment of the reject frequency has little influence on the pass frequency. Greater shifts wiil produce deeper notches. On the pass frequency the insertion loss is low and SWR near 1:1,1. At the reject frequency the expected notch deepness is 27dB @ shift=1,6MHz, not as good as a real (bigger) Kathrein SP cavity, but this is mainly due to the smaller dimensions of the construction here.

    Number of needed cavities for a repeater duplex filter:
For at least 80dB insulation between a repeater TX and RX (minimum value if one antenna is to be used) you need 2x3 cavities.

With 2x2 cavities from the above stated construction you only get about 64dB insulation, so you have to use a circulator between the filter output cables and the antenna connection to get about 85 dB insulation between transmitter and receiver. But, a circulator can generate inter modulation and harmonics, resulting in sideband noise of the own transmitter being mixed to the repeater input frequency, and loss of sensitivity when the repeater transmitter is switched on (the repeater can get the "hiccups" when a weak signal is received). Amplified TX harmonics will be radiated too. A circulator also introduces more insertion loss than a good notch cavity.

Remember: a circulator can give about 20db extra insulation only if:
- firstly the magnetic circuit of the circulator is retuned to the operating frequency
- secondly the trimmer-C's are retuned for exact 50 ohms impedance (max insulation between ports)
- AND thirdly the antenna-SWR is very low (<= 1:1,2). If the reflection damping of the antenna circuit is only 10dB (SWR 1:1,5 or so), you only get no more than 10dB extra insulation. This could happen during icy conditions resulting in receiver-desense (hiccups).

So, better use two extra cavities and NO circulator. Isolation then is optimal and constant.

In some cases these two extra cavities could be better of the band pass-type in order to make good selectivity. They then prevent IM in the receiver, make it possible to connect an external low noise preamp, and suppresses wideband transmitter sideband noise. A good (big) band pass cavity can provide 15dB insulation @ 1dB IL.

Prevent cracking noises. Cavities should NOT touch each other or other metal parts. Place them in an insulating construction.

If possible, line up the cavities in a straight line from TX to ANT to RX. If this is not possible, keep some distance (5cm) between TX and RX cavities, so no stray coupling can occur between them due to radiation-leakage through the contact surfaces between the lids and the cavity bodies. Turn lid screws TIGHTLY and do NOT mechanically connect cavities together using metal strips (!! my experience with professional Kathrein SP cavities !!).

At the pass-frequency, individual cavities should present exactly 50 ohms impedance. At the notch frequency, the impedance is also resistive but very low (in the order of 1,5 ohms) due to the series resonance to mass of the internal coil-trimmer combination. This impedance is the lower, the better series-circuit-Q is. So thick silvered copper for coil and strip, and a loss-free trimmer is a must.

    What cable lengths, connectors?
Interconnecting cables should have such a length, that cavities do not influence each other on the notch-frequencies. If the interconnection cables between adjacent cavities and the antenna T-connector are effectively exactly 1/4 electrical wavelength, these cables will transform the low notch impedance to a very high impedance at the connection of notch circuit in the adjacent cavity. Maximal signal reflection at the notch frequency will occur, increasing isolation, which is exactly what we want! These cables then work like a frequency-dependant insulator/conductor (notch/pass).

So, an exact 1/4 lambda RF transmission path length between the series resonating circuits of the adjacent cavities is needed. This should be realized between the CENTERS of the internal 50 ohms strips inside each cavity. But also from the center of such a strip and to the center of the T-connector at the antenna-output of the filter assembly.

Cable lengths between TX, RX and the filter inputs are not critical, if the filter-VSWR, rx-VSWR and transmitter-output-tuning into 50 ohms are OK. Try to make total length of these cables (including plugs and internal wiring!) a whole number of 1/2 electrical wavelengths, to avoid higher mismatch due to cable transformation of connected impedances.

Use only double shielded (PFTE) cable and first quality (PFTE) connectors ! According to my experience, BNC connectors and  SO239/PL259 connectors are unreliable, espacially at lower levels than -80dB. Better use connectors and plugs which will be screwed , and of which the mass connections make contact tightly and very secure. This will avoid surface currents on the outside of the coax screening, bad isolation and an unreliable center contact if mechanical stress on cabling occurs.

    Total performance:
After adjustment of each individual cavity, and connecting together with exactly dimensioned "1/4 wave" long cables, the performance of the total combination should be checked. When measuring under -80dB, beware of possible stray coupling due to the use of single screened cables, even due to HF currents running over the outsides of connectors and cables if BNC is used! Do NOT retune singly tuned but now interconnected cavities, if you are not absolutely sure this cannot happen! Measuring under -90dB can be tricky.

When checking the total combination, on the pass-frequency the insertion losses of the individual cavities and cables just should add up (totally <= 2dB), and VSWR should relain very close to 1:1,0.

At the notch frequency the total insulation should be the sum of that of individual cavities, and 5dB extra for each 1/4 lambda interconnection cable: for instance 3 x 27 dB + 3 x 5 dB = 96 dB.



    How to determine cables lengths:
Connectors, T-pieces and receptacles are included in transmission paths and add to the effective RF path lengths. So, the mechanical length of the interconnection cables should be calculated by keeping in mind the already present extra external lengths of receptacles, the 1/2 length of the internal connection between the receptacles inside each cavity, and the 1/2 length of the antenna T-piece. The mechanical cable lengths (measurable with a roller measure) will be shorter than (1/4 lambda x velocity factor). How much shorter depends on the internal construction details of the connected equipment.

Mechanical cable length = (1/4 lambda - RF-path length of all receptacles etc.) x (cable velocity factor)

    HF-path length of all receptacles etc:
Measure the path lengths of all SCREENING LENGTHS as if they be connected together. I made a sketch of it. Mark type of insulation. The insulation has an electrical lengthening effect over the mechanical length. So, calculate all effective RF- path lengths by deviding mechanical lengts through the appropriate velocity factor. At positions where the insulation material is air (N-connector, cavity), the velocity factor is 1.0. If the insulation is solid PTFE, the velocity factor is ca. 0,8. Then summate all path lengths to find total RF-path length of all connection materials.

    Mechanical cable length:
Then calculate the "RF"-cable length, and multiply it with cable-velocity factor to get the wanted mechanical cable length. Velocity factor varies between 0,68 and 0,85 for different cable types.