How to Build a Butterworth Bandpass Filter for Ham Radio

Your neighbor calls to say the TV went haywire the moment you keyed up on 40 meters. Your receiver falls apart every time a local AM station fires up. Both of these are real problems, and both have the same fix: a bandpass filter in the right place. Put one between your transmitter and the antenna and the harmonics that were hitting the broadcast band disappear. Put one ahead of the receiver and the strong out-of-band stations stop clobbering your front end. Two problems, one relatively simple build.

This guide covers the whole process: why the Butterworth topology makes sense for ham radio, how to pick the right components for your power level, how to wind the inductors and put the board together, and how to verify the result with a VNA. The math is here if you want to dig in, but you can also go straight to the calculator, grab your component values, and come back to the theory when you are curious.

What is a bandpass filter and why does your station need one?

A bandpass filter passes signals within a specific frequency range and attenuates everything outside it. A low-pass filter cuts everything above a cutoff frequency. A high-pass filter cuts everything below one. A bandpass filter does both at once, creating a window centered on your operating frequency. Signals inside that window get through. Everything else gets rolled off, and depending on how far outside the band you go, the attenuation can be dramatic.

In a ham station you need this for two completely different reasons, depending on which direction the signal is moving.

On transmit: harmonics and spurious emissions

Every transmitter puts out some energy at frequencies it was not supposed to. The second harmonic of a 7 MHz signal lands at 14 MHz, right in the middle of 20 meters. The third harmonic hits 21 MHz on 15 meters. A modern transceiver is reasonably clean on its own, but run it through a linear amplifier and harmonic levels go up because amplifiers are not perfectly linear. A bandpass filter between the radio and the antenna removes those harmonics before they reach the coax. The FCC requires spurious emissions to be at least 43 dB below your carrier if you are running over 5 watts. A well-built bandpass filter will get you 40 to 60 dB of rejection at the second harmonic, which is well past that requirement.

On receive: IMD, receiver overload, and front-end damage

Your antenna does not know what band you care about. It brings in everything, including whatever strong signals happen to exist near your operating frequency. A commercial AM broadcast station a few megahertz away can drive your receiver's front end into compression, generate intermodulation products that show up as fake signals in your passband, and raise the noise floor across the board. At high enough levels, the front end can be permanently damaged. A bandpass filter in the receive path cuts most of that energy before it reaches the radio. This matters most for software-defined radios, which have no preselector filtering built in at all. Hook an RTL-SDR or similar device up to an outdoor antenna in a populated area and watch the waterfall fill with garbage. Add a bandpass filter and it clears up noticeably.

Why Butterworth? Understanding filter responses

Filter design is a tradeoff. You can have a very flat passband with a gentle rolloff, or you can have a steeper rolloff with ripple in the passband, or you can push the rolloff even steeper by accepting ripple in both the passband and the stopband. Engineers have worked out the math for several different approaches, each of which optimizes for something different.

Type	Passband behavior	Stopband steepness	Phase response	Best use case
Butterworth	Maximally flat, no ripple	Moderate	Good	General purpose ham use, most forgiving to component tolerances
Chebyshev Type I	Equiripple in passband	Steeper than Butterworth	Poorer	When you need more stopband attenuation and can tolerate passband ripple
Elliptic (Cauer)	Equiripple in passband	Steepest for given order	Worst	When every dB of stopband rejection matters and passband ripple is acceptable
Bessel	Gentle rolloff	Poorest	Best (linear phase)	When phase linearity matters more than selectivity

Butterworth is the right choice for most ham radio bandpass filter work. The flat passband means your entire operating band has about the same insertion loss, with no sweet spot in the center and drooping performance near the edges. It is also the most forgiving of component tolerances. A 5% capacitor error in a Butterworth design costs you a small amount of bandwidth accuracy. That same error in a tight Chebyshev design can shift the passband ripple amplitude significantly. And a two-resonator Butterworth built with quality components will typically give you 40 dB or more of rejection at the second harmonic, which is all you really need for most ham applications.

The double-tuned topology: how this filter works

The filter in this guide uses two identical LC resonators connected by a small coupling capacitor. Each resonator attenuates signals outside the passband on its own. The coupling between them sets the shape of the combined response. Get the coupling right and you get the Butterworth response. Get it wrong and you end up with either a double-humped passband or a narrow, lossy one. Understanding what each part does makes troubleshooting a lot easier.

The resonators (L1/C1 and L2/C2)

Each resonator is a series inductor L feeding into a shunt capacitor C that connects to ground. At the filter's center frequency f0, the reactance of L equals the reactance of C and the combination resonates. Above and below that frequency the impedance shifts and the signal gets increasingly attenuated. The sharpness of this resonance is the loaded Q, written Ql:

Higher Ql means a narrower passband and sharper skirts. It also means higher voltage stress on the capacitors, which matters a lot at transmit power levels, and more sensitivity to component tolerances. The 30m allocation is only 50 kHz wide, which pushes Ql up around 200. That is why 30m filters are genuinely difficult to align compared to 40m or 20m builds. For a first project, 40m or 20m is the easier target.

The coupling capacitor (Cc)

A small series capacitor connects the two resonators. The ratio of Cc to C1 and C2 sets the coupling coefficient k, and that coefficient determines whether you get a Butterworth response or something else. The Butterworth condition is:

Too much coupling and the two resonators fight each other, producing a double-humped passband with a dip in the center. Too little and the passband collapses into something narrow and lossy. The Butterworth value of k lands exactly at critical coupling, which is the flattest response you can get from this topology.

The complete signal path

RF enters the input port from a 50 ohm source, travels through L1 in series, arrives at Node 1 where C1 shunts some energy to ground, crosses through Cc to Node 2 where C2 does the same thing, then passes through L2 in series to the 50 ohm output. The circuit is completely symmetrical, so it works identically in both directions. You can use it on transmit, on receive, or in a switched T/R circuit.

Power handling: the most important thing to get right

More homebrew filters fail from wrong component ratings than from any other cause. The mistake is treating the filter like a passive piece of wire and choosing capacitors based on their DC rating or whatever happened to be in the parts bin. Inside a resonant RF circuit, the voltages and currents at the internal nodes are considerably higher than what you measure at the ports, and the reason is resonant voltage multiplication.

Calculating voltage stress on the capacitors

In a 50 ohm system, the peak RF voltage at the input port is:

The reason the stress on C1 and C2 is so much higher than the port voltage is that energy is circulating between L and C at resonance. The reactive current through the capacitor multiplies the effective voltage by a factor related to Ql. On 40m with Ql around 24, a 100W input with 100V peak at the port puts approximately 600V peak across the resonating capacitors. A cheap ceramic disc rated for 50V will fail immediately, and it will usually fail loudly.

Wire gauge and current rating

The inductors carry both the signal current and the reactive circulating current at the resonant nodes. Use wire that is rated for the job:

Power level	Port current (rms)	Minimum wire gauge	Recommended
QRP (5W or less)	0.32 A	28 AWG	26 AWG for margin
100W	1.4 A	22 AWG	20 AWG for margin
500W	3.2 A	17 AWG	16 AWG, use T106 or larger core
1500W	5.5 A	14 AWG	QRO filter design, copper tube winding

Capacitor types by power level

Capacitor type	Max power	Voltage range	Notes
NP0/C0G ceramic (SMD or leaded)	Up to 100W	Up to 500V	Best choice for QRP through 100W builds. Excellent stability and low loss.
Silver mica	Up to 500W	Up to 1000V	The classic RF capacitor. Superb stability and Q. Harder to find in uncommon values. Good default for medium-power HF.
Polystyrene	Up to 200W	Up to 600V	Excellent Q and temperature stability. Keep it away from heat during soldering. Good for receive-only builds.
ATC (American Technical Ceramics)	Up to 1500W and beyond	Up to 2000V and beyond	Purpose-built RF capacitors. Expensive, but the right answer for QRO builds.
Vacuum variable	Legal limit and beyond	Thousands of volts	Used in legal-limit amplifier output circuits and commercial broadcast equipment.

Component selection: toroid cores

The inductor is the part that determines how well your filter performs. Commercially wound inductors are rarely available in the exact values needed for HF bandpass work, so winding your own toroids is the standard approach. Micrometals cores are what most builders use. They are consistent between batches, the A_L values are well documented, and you can buy them from normal distributors in the US. Cores from unknown Chinese suppliers often have unspecified tolerances and are sometimes mislabeled. Buy genuine Micrometals.

Micrometals cores are identified by size and mix. The size number is roughly the outer diameter in hundredths of an inch, so a T68 is about 0.68 inches across. The mix number describes the ferrite material and determines loss characteristics versus frequency.

Mix	Color	Best bands	AL (T50)	AL (T68)	Notes
Mix 2	Red	160m through 40m (1 to 10 MHz)	49 nH/t²	57 nH/t²	Highest permeability in the HF range. Go-to for lower frequencies where you need higher inductance values.
Mix 6	Yellow	20m through 10m (10 to 50 MHz)	40 nH/t²	47 nH/t²	Lower loss at higher frequencies. Required for 20m and above. Lower AL means you need more turns for the same inductance.
Mix 15	Red/white	160m through 80m	135 nH/t²	N/A	High inductance per turn. Useful for 160m builds where you need a lot of inductance.
Mix 17	Blue/yellow	50 MHz through 200 MHz	N/A	N/A	VHF mix. Use for 6m and 2m filter construction.

How to calculate turns for a target inductance

Once you know the A_L value for your core, turns count is just:

Round to the nearest whole turn, then measure the actual inductance with an LCR meter. The A_L tolerance on most Micrometals cores runs about plus or minus 10 percent, which is why measuring before you solder anything is not optional. If you are low, add a turn and re-measure. If you are high, remove one. You can also fine-tune by spreading or squeezing the winding slightly on the core.

Core size selection guidelines

Band	Core	Reason
160m	T106-2	Large core needed to hit the required inductance without too many turns. 22 AWG wire fits comfortably.
80m, 60m, 40m	T68-2	Good balance of inductance per turn, winding space, and unloaded Q for these frequencies.
30m, 20m	T50-2	Sufficient inductance per turn. Compact. Holds up well at these frequencies.
17m, 15m, 12m, 10m	T50-6	Mix-6 required for acceptable loss above 14 MHz. T50 size gives enough winding space for the turn counts needed.
6m, 2m	T37-6	Small core, Mix-6. Very few turns at VHF so physical layout and lead length matter a lot.

Tools and test equipment you will need

Tool	Required?	Purpose
Fine-tip soldering iron, 20 to 40W, temperature controlled	Required	Soldering small RF components. A chunky iron will overheat NP0 caps and damage winding insulation.
LCR meter (DE-5000 or Atlas LCR45 are popular choices)	Strongly recommended	Verify actual inductance of wound toroids. Without one you are guessing at alignment.
VNA (nanoVNA, TinySA Ultra, etc.)	Strongly recommended	Measure insertion loss, bandwidth, and frequency response of the finished filter.
Small drill and copper-clad board	Required	For Manhattan-style construction or PCB work.
Wire stripper and needle-nose pliers	Required	Stripping magnet wire and bending leads.
Heat-shrink tubing and flux	Required	Clean up your solder joints. Flux makes a real difference on small RF work.
Toroid winding dowel or short piece of PVC pipe	Optional	Helps keep turns even on small cores. A short piece of 1/4 inch dowel works well for T50 cores.
Calipers or millimeter ruler	Optional	Useful for checking physical spacing in Manhattan construction.

Step-by-step construction guide

The steps below assume Manhattan-style construction on a copper-clad board, which is the fastest approach for a one-off build. PCB construction works just as well. The layout rules are the same either way.

Testing and alignment with a VNA

A nanoVNA costs about 60 dollars and turns filter testing from guesswork into actual measurement. The S21 transmission plot will show you insertion loss, passband shape, bandwidth, and harmonic rejection all at once. It is genuinely one of the most satisfying things to look at when a filter comes together correctly.

What good numbers look like

Parameter	Target value	Notes
Insertion loss at f0	Less than 1.0 dB	Loss is mostly from winding resistance. Heavier wire and high-Q cores help.
3 dB bandwidth	Within 10% of calculated BW	Tracks directly with how accurate your component values are.
Attenuation at 2nd harmonic	More than 40 dB	Measure at 2 x f0. Should be deeply buried.
Return loss (S11) in passband	More than 15 dB, SWR below 1.5:1	A well-matched filter presents close to 50 ohms in the passband.
Passband ripple	Less than 0.5 dB	Butterworth should be very flat. Ripple means the two resonators are not matched.

Alignment with a trimmer at Cc

1. Set up the VNA for S21 and sweep from about half to double your target center frequency.
2. Look at the passband shape. Two humps with a dip in the middle means Cc is too large and the filter is over-coupled. Reduce it. A narrow peaked passband with high insertion loss means Cc is too small. Increase it.
3. Adjust the trimmer in small steps. You are looking for a single smooth rounded peak centered on f0 with equal slopes on both sides. That is the Butterworth response.
4. If the center frequency is off, adjust C1 and C2 for large corrections, or carefully squeeze or spread the toroid turns for small ones. Spreading turns slightly raises f0. Squeezing them slightly lowers it.
5. Once you are happy with the alignment, note the trimmer setting. If you want to replace the trimmer with a fixed cap, measure the trimmer with your LCR meter at that setting and substitute the closest standard value. Trim the exact value by adding a small parallel cap if needed.

Troubleshooting

Where to install the filter in your station

The location determines what problem the filter actually solves. A filter in the wrong place does nothing useful.

Installation point	Solves	Notes
Between radio and antenna, transmit path	Harmonic emissions, TVI, audio interference complaints	Must be rated for your transmit power. Most common installation.
Between amplifier and antenna	Amplifier-generated harmonics	Must be rated for full amplifier output. Use silver mica or ATC caps at legal limit power levels.
Between antenna and receiver input	IMD, front-end overload, raised noise floor	Receive-only so voltage rating is not a concern. Focus on keeping insertion loss low.
Between antenna and SDR	SDR overload from broadcast stations	Critical for any wideband SDR. Even a simple BPF makes a noticeable difference on the waterfall.
In a T/R relay circuit	Both transmit harmonics and receive IMD at the same time	Filter handles TX and RX. Must be rated for TX power.

Scaling up: switchable filter banks

Once you have built one filter and understand how it works, the obvious next step is a full bank, one filter per band, switched automatically when you change frequencies. That is the setup in most serious SO2R contest stations. The concept is simple: eight to twelve filters on a relay board, switched by the band data output from the radio.

W3NQN (Nick Tusa) published a well-regarded set of bandpass filter designs for all the HF bands in QST and the ARRL Handbook. These have been widely reproduced and kit versions are available from several vendors. The topology is essentially what this guide describes, with values optimized for each specific band. If you want a complete switchable filter bank for a contest station, the W3NQN designs are the accepted starting point.

Most modern transceivers provide band data on an accessory connector. Icom uses a BCD code on the ACC port. Yaesu and Kenwood have similar outputs. You can also read band data from the CAT interface with a small microcontroller like an Arduino or ESP32 if your radio does not have a dedicated band bus. Standard RF relays rated for your power level handle the switching without trouble.