Passive multi-port band-pass filter:

What follows are tables of suggested component values for different types of multi-band combiner/filters.  These filters are intended for specific frequency operation, but are generally likely to be wide enough to allow the complete coverage of the relevant amateur band with good results.

The data below includes options that could include adjacent WARC bands (or 60 meters) as desired - but 12 meters, due to its relative "closeness" to 10 and 15 meters - is not included as this would require insertion loss - using practical construction methods - well over 2 dB.  

You may pick and choose the values from the table below to construct filters for the bands as desired.  Two tables are included below:  The upper table has values calculated to attain 25dB of attenuation at the center frequency of the noted adjacent bands and the lower table has values calculated for 30 dB.  As expected, two parameters lead to higher insertion losses:

• The need for higher attenuation on an adjacent band.  The 30dB filters will be lossier than the 25 dB filters.
• The need to attenuate a frequency that is closer to the desired frequency.  For example, if you need to operate 17 meters, both the 20 meter and 15 meter filters needed to attenuation 17 meter signals will be narrower and lossier, so if you do not need 17 meters, to not select the 20 and 15 meters that are designed to reject it.
• If you wish to reduce, as much as possible, off-frequency signals for a receiver to prevent overload - and can tolerate a bit of extra loss, use the "30 dB" filters and select those filters with the highest insertion losses as they will also be the narrowest.

Filter topology:

These filters are series-input N=3, meaning that there are three pairs of L/C components to provide the band-pass filtering:  The series L/C on the inputs and outputs allows them to be connected in parallel with minimal interaction.  A complexity of N=3 was chosen to reduce component count, but this also means that the "Q" of the sections is a bit on the high side to attain the desired attenuation so careful attention must be paid to the layout (e.g. good, solid ground plane) and quality of the components (e.g. silver mica or NP0/C0G ceramic capacitors).  

The diagram below shows the nomenclature for the individual filter sections, and how several filters may be connected on a common bus to split and filter signals for receive - or to combine several transmitters on different frequencies.

The top section ("Band 1") represents a single section, and as the additional sections are added, they would be connected to the "Common port" as depicted with the common bus along the left side:  The Line is showed disconnected in the figure below to reinforce the fact that they should be tested individually as they are built and connected together ONLY once this is done.



For a "Window" filter:

A "Window" filter is that in which one wishes to pass only certain portions of frequency bands from a common source - and then combine the filtered signal back to a common bus, simply connect together the "output" lines (to the right of C3 in each section).

An example where this might be used is for a broadband receiver (e.g. KiwiSDR, RX-888, etc.) may be subject to overload from the entirety of signals impinging on it - but only certain frequency ranges (e.g. amateur bands) are of interest.

This same method could also be used to filter/connect several different antennas (e.g. those designed for different frequencies) to a common bus/receiver.

---

For nominal 25dB isolation to adjacent band:

Description	L1	C1	L2	C2	L3	C3	Approx. Calculated Insertion Loss
10M (no 12M) CF=28.1MHz	730nH	31pF	48nH	562pF	729nH	31pF	1.0dB
15M with 17M  CF=21.1MHz	2.0uH	24pF	25nH	2110pF	1.7uH	29pF	0.8dB
15M (no 17M)  CF=21.1MHz	1.74uH	35pF	51nH	1196pF	1.5uH	39pF	0.7dB
17M (with 20M and 15M) CF=18.1MHz	2.4uH	34pF	27nH	2926pF	4.1uH	20pF	0.9dB
20M with 30M and 17M  CF=14.1MHz	3.2uH	42pF	88nH	1599pF	3.1uH	45pF	0.7dB
20M (no 17M or 30M)  CF=14.1MHz	2.1uH	67pF	165nH	904pF	2.0uH	69pF	0.3dB
20M with 30M (no 17M)  CF=14.1MHz	2.4uH	49pF	110nH	1047pF	2.4uH	50pF	0.5dB
30M (with 40M and 20M)  CF=10.125MHz	3.0uH	85pF	96nH	2641pF	2.6uH	97pF	0.4dB
40M with 80M and 30M (no 60M)  CF=7.1MHz	3.6uH	170pF	201nH	2915pF	3.1uH	205pF	0.4dB
40M with 80M (no 30M or 60M)  CF=7.1MHz	2.4uH	194pF	310nH	1454pF	2.1uH	210pF	0.3dB
40M with 60M and 30M  CF=7.1MHz	3.8uH	117pF	103nH	4612pF	3.9uH	114pF	0.6dB
80M with 40M (no 60M or 160M)  CF=3.55MHz	4.4uH	587pF	1.1uH	2452pF	3.7uH	750pF	0.2dB
80M with 40M and 160M (no 60M)  CF=3.55MHz	5.3uH	350pF	618nH	2907pF	3.7uH	477pF	0.25dB
80M with 160M and 60M  CF=3.55MHz	6.1uH	432pF	520nH	4682pF	7.0uH	366pF	0.35dB
160M with 80M  CF=1.82MHz	7.9uH	1327pF	2.1uH	5095pF	8.2uH	1246pF	0.3dB

For nominal 30dB isolation to adjacent band:

Description	L1	C1	L2	C2	L3	C3	Approx. Calculated Insertion Loss
10M (no 12M) CF=28.1MHz	1.5uH	20pF	34nH	890pF	1.5uH	20pF	1.2dB
15M with 17M  CF=21.1 MHz	3.0uH	17pF	22nH	2396pF	1.9uH	24pF	1.5dB
15M (no 17M)  CF=21.1 MHz	2.2uH	27pF	46nH	1327pF	1.9uH	31pF	0.5dB
17M (with 20M and 15M) CF=18.1MHz	2.5uH	37pF	23nH	3495pF	4.2uH	20pF	1.2dB
20M with 30M and 17M  CF=14.1MHz	3.0uH	46pF	49nH	2744pF	3.4uH	40pF	0.7dB
20M (no 17M or 30M)  CF=14.1MHz	2.1uH	63pF	112nH	1268pF	2.4uH	57pF	0.4dB
20M with 30M (no 17M)  CF=14.1MHz	2.6uH	47pF	77nH	1536pF	2.7uH	44pF	0.5dB
30M (with 40M and 20M)  CF=10.125MHz	4.2uH	60pF	97nH	2632pF	3.5uH	74pF	0.5dB
40M with 80M and 30M (no 60M)  CF=7.1 MHz	5.3uH	129pF	349nH	1900pF	4.9uH	146pF	0.45dB
40M with 80M (no 30M or 60M)  CF=7.1MHz	3.2uH	159pF	444nH	1121pF	3.4uH	149pF	0.35dB
40M with 30M and 60M  CF=7.1MHz	4.5uH	99pF	100nH	4761pF	6.0uH	75pF	0.9dB
80M with 40M (no 60M or 160M) CF=3.55MHz	4.4uH	684pF	933nH	2974pF	5.2uH	539pF	0.25dB
80M with 40M and 160M (no 60M)  CF=3.55MHz	4.5uH	432pF	565nH	3460pF	6.0uH	329pF	0.3dB
80M with 160M and 60M  CF=3.55MHz	7.2uH	344pF	416nH	5599pF	10.3uH	222pF	0.5dB
160M with 80M  CF=1.82MHz	10.4uH	879pF	1.71uH	5653pF	11.1uH	818pF	0.25dB

Construction notes - PLEASE READ!

• Values calculated using ELSIE and optimization.
• Insertion loss calculated assuming capacitor Q of 1000 and inductor Q of 100:  "Real-world" values will naturally vary from predictions, so expect slightly higher insertion loss on the center frequency in some cases.
• Use one or more conventional capacitance values in parallel to attain values shown.
• Use only Silver Mica or NP0/C0G capacitors.
• For C2, use shortest lead length possible
• For C2, it is recommended that at least two parallel capacitors of similar value be used to make up the bulk of the capacitance to spread losses and improve dissipation, using additional, smaller capacitors to get closer.
• For C2, the optimal capacitors are SMD silver mica;  Next choice is dipped silver mica with very short leads;  Next choice is multi-layer SMD NP0/C0G capacitors and the last choice is NP0/C0G capacitors with very short leads.
• For the lower bands (160-40 meters) it may be possible to use high-current film-type plastic capacitors with good results:  Do NOT use "safety" capacitors (e.g. X1, X2 types) as these can have higher ESR.  Again, parallel capacitors to minimize losses.
• In most cases, L2 will be wound with heavy copper wire (e.g. 16 AWG or heavier).  Values of up to about 600nH are practical with this method - but above that, a large-ish toroidal inductor (T68-2 or T80-2) wound with the heaviest wire for the number of turns.  For receive-only applications or where slightly higher losses may be tolerated, T50-2 will work.
• In most cases, L1 and L2 would be wound using T68-2 or T80-2 toroids using the heaviest wire for the number of turns.  For the 80 and 160 meter filters, T68-1 or T80-1 may be considered.  For receive-only applications or where slightly higher losses may be tolerated, T50 sized toroids will work.
• Refer to the turns calculator at toroids.info for the number of turns and the length of wire needed.  Please note that toroids are imprecise and that inductance can easily vary by 20% from the calculated value, so be prepared to adjust the number of turns.  Typically, the number of turns is over-calculated slightly and 5-10% of the turns will need to be removed.
• It is strongly recommended that inductance values be measured after the toroid is wound:  This may be easily done using a calibrated NanoVNA - typically at around 100 kHz where stray reactance will have less effect:  This is usually be done by connected the inductor across the "Output" port (e.g. the port used when one is measuring only return loss or VSWR) and setting frequency/cursor to 100 kHz.
• Important not when using amplifiers:  These filters - like almost all filters - will present a matched 50 ohm load (when terminated with 50 ohms) ONLY at their tuned frequency - but will present severe mismatch at all other frequencies.  This can cause some amplifiers to become unstable, requiring either a small amount of attenuation (2 dB or so is usually sufficient) or a "diplexer" type of circuit (a parallel L/C circuit with a series resistors connected across the port with the amplifer) to provide termination at frequencies away from the pass response.

---

The picture below shows one possible construction method:



About the filter in picture:

• "Manhattan" style construction is used with the components being soldered directly to the ground plane.  This board is mounted to the lid of a die-cast alumimum box.
• "ME-Squares" from "qrpme" (link) are used to isolate junctions as needed.
• Individual sections are isolated from each other by using 20-30mm high partitions.
• RG-316 type coaxial cable is used for each section.
• For this filter, a combination of SMD and leaded NP0/C0G capacitors are used with #12 AWG tin-plated copper wire being used for the low-value inductors.
• The common bus is seen along the bottom:  Orange "kapton" tape is used to prevent the wire from shorting to the ground plane.
• The toroids used are T68-2, secured in place - after tuning - with dabs of RTV (silicone) adhesive to prevent their moving around/detuning:  The use of "hot (thermoset) glue" is not recommended.

Comments on tuning the filters  You had better read this, too!

For tuning the filters, it is recommended that they be tuned individually as follows:

• It is STRONGLY suggested that a calibrated NANOVNA or similar device be used rather than an antenna analyzer:  A "through" S21 measurement is extremely important and antenna analyzers with a single port simply cannot do this.
• You WILL NOT get very good results unless you can and do tune each section properly!
• Each filter section should, at first, be tuned individually.  It is suggested that you build and tune one section at a time -starting with the lowest frequency - so that you get a "feel" for building and tuning the filters.
• By testing each filter section individually, you will avoid interaction and confusion should - prior to tuning - a filter be far off frequency, particularly if it overlaps or is near another filter's frequency prior to tuning.
• The highest-frequency filters will be more difficult to tune, so "getting the hang of it" with the lower frequency filters is strongly recommended!
• On the VNA, set THREE markers:
• One marker on the desired center frequency.  Here, you should monitor both insertion loss and return loss (or VSWR).
• Another marker on the operating frequency in the next-lower band to be used.  This would be monitored while tuning to attain the desired attenuation.  This would not apply when tuning the filter for the lowest-used frequency, of course.
• The  third marker on the operating frequency in the next-higher band to be used.  This would be monitored while tuning to attain the desired attenuation.  Again, this would no apply when tuning the filter for the highest-used frequency.
• Note that these filters are optimized for minimum insertion loss at the desired frequency and the target insertion loss (25 or 30 dB, depending on the values) at the adjacent-band frequency.
• For this reason, these filters may not be symmetrical in terms of pass response.  In other words, do NOT try to tune the filter so that the desired frequency is at the center of the passband, but rather look ONLY at the three markers as described above.

More to be added!

CT20230405



Northern Utah WebSDR landing page