Why a Band-Pass Filter Can Pass Its Test and the Station Still Fail

A band-pass filter can meet every published small-signal number and still be insufficient for a multi-operator station. That is not a generic filter failure. It means the component was selected, placed or tested against only part of the aggressor-to-victim problem.

“Standard Band-Pass Filter” Is Not One Architecture

Filter position and construction determine the job it can perform. At least four architectures appear in contest stations:

Ed Wetherhold's W3NQN filters are historically important but must be cited within their scope. The original ARRL design was intended between roughly 150 W transceivers and 1.5 kW amplifiers, with a 200 W ICAS component design. Its goals were to reduce transceiver harmonics and phase noise before amplification and to reduce receiver overload. It was not a 1.5 kW post-amplifier filter.

Current commercial specifications also show why no universal adjective such as “standard” is adequate. The 4O3A Series XL page describes five-pole post-amplifier filters rated at 4,500 W ICAS, with band-dependent insertion loss, match and adjacent-band rejection; it also says duty cycle, ambient conditions, cooling and VSWR affect heat dissipation. VA6AM's 15 m design publishes different limits for PEP, RTTY and fan airflow, and conditions them on antenna VSWR below 1.5. These are vendor specifications for particular assemblies, not transferable ratings for every BPF.

Separate the Interference Mechanisms

1. The other transmitter's fundamental blocks the receiver

The aggressor carrier couples into the victim antenna at the aggressor frequency. The aggressor's own BPF must pass that carrier, so the useful filter rejection normally comes from the victim-side filter or preselector:

The blocker target is not the receiver noise floor and not a universal damage level. Establish it by injecting a calibrated weak wanted signal and increasing the blocker at the actual frequency offset until an agreed criterion is reached. The ARRL Laboratory blocking-gain-compression procedure uses a 1 dB reduction in weak-signal audio; a station may add reciprocal-mixing, ADC-overload and recovery criteria.

Those −20 and −30 dBm values are examples, not recommendations. A receiver's impairment can occur at a lower or higher level depending on preselection, preamplifier and attenuator state, ADC headroom, local-oscillator phase noise, offset and the number of simultaneous blockers.

2. Transmitter noise falls inside the victim band

If the transmitter or amplifier produces broadband noise at the victim frequency, the victim filter must pass it. Suppression must occur on the source side, normally after the last stage that generates significant noise:

Here LTX is a measured noise density in dBc/Hz relative to the carrier and B is the victim receiver's equivalent noise bandwidth. If the source instrument reports absolute dBm/Hz instead, use that quantity directly and do not add carrier power again.

3. A harmonic or discrete spur lands in the victim passband

A spectral line needs an integrated-power budget, not the 10 log(B) term used for noise density. The second harmonic of 7.05 MHz falls at 14.10 MHz, and the second harmonic of 14.10 MHz falls at 28.20 MHz. A 20 m victim filter cannot reject a 40 m transmitter's second harmonic at 14.10 MHz because that is precisely what the 20 m filter must pass. The 40 m source chain must suppress it.

By contrast, harmonics of a 14.05 MHz transmitter occur near 28.10 and 42.15 MHz, not in the 17 m band. A 20 m to 17 m problem instead points toward fundamental blocking, close filter skirts, transmitter noise at 18 MHz, reciprocal mixing or an unintended coupling path.

4. The receiver or site creates new products

Two strong signals can generate intermodulation in the receiver front end. Corroded contacts, dissimilar-metal junctions, loose connectors and illuminated structures can create passive intermodulation outside the receiver. These products may appear only when a particular combination of transmitters is keyed. A single-transmitter test cannot clear a multi-multi station.

5. RF bypasses the differential filter path

Exterior coax current, control cables, Ethernet, cabinet seams, relay wiring and common bonds can form parallel paths around the intended two-port network. The original RF.Guru line-isolator page reports more than 45 dB common-mode reduction in a two-current-clamp fixture and less than 0.30 dB differential insertion loss. That is useful fixture characterization, not a promise of 45 dB receiver improvement in every station. Installed reduction depends on the common-mode source and load impedances, placement and parallel return paths.

Rejection Is a Frequency-Specific Transfer Measurement

“60 dB rejection” is incomplete without the filter model and serial number, direction, exact aggressor and victim frequencies, source and load impedances, power, temperature, fixture, calibration plane, bandwidth and measurement floor. A harmonic notch at one marker does not establish minimum rejection across an amateur band.

Two nominal 40 dB filters also do not automatically produce 80 dB in service. Decibel attenuation adds only for the same signal through the same cascade under the conditions represented by the measurements. It can stop adding because:

• the stopband source or load is not the 50-ohm termination used for the published trace;
• the filters interact through their complex S11, S22 and connecting-line phase;
• direct input-to-output enclosure or cable coupling sets a leakage floor;
• a switch, triplexer, dummy load or relay has less isolation;
• common-mode current or radiation bypasses the cascade; or
• the VNA's residual crosstalk and test-lead coupling set the displayed floor.

Measure the complete installed chain, retain the full two-port S-parameters and test both directions. Keysight's network-analyzer guidance emphasizes that source power, IF bandwidth, averaging, calibration and test-set isolation determine dynamic range; de-embedding requires valid network data and reference planes.

Insertion Loss Is Not Automatically All Heat

For a matched output and normalized 50-ohm S-parameters, incident power divides into transmitted, reflected and absorbed power:

The familiar insertion-loss calculation gives power not delivered to the matched output. It is an upper bound on filter heating until input reflection is also known:

For example, 20 dB input return loss reflects 1% of incident power. At 1.5 kW that is 15 W. If insertion loss is 0.10 dB under the same matched-output condition, approximately 19.1 W is absorbed rather than the full 34.1 W not transmitted. With a better input match, more of the missing 34.1 W is absorption. Either way, the heat is not distributed uniformly: capacitor ESR, coil resistance, joints, coupling conductors and enclosure currents can create local hot spots.

Duty cycle and waveform decide average heating

A power label must distinguish incident PEP, steady-carrier or average power, permitted on/off cycle, maximum transmission time, cooling airflow, ambient temperature and recovery time. “ICAS” alone is not a reproducible duty cycle. A filter that survives intermittent speech may exceed a component temperature limit during RTTY, FT8 or sustained CW.

Vendor data illustrate the dependency. VA6AM publishes different 15 m limits with no fan, 50 CFM and 100 CFM, plus a VSWR condition. The current 4O3A page supplies fans and a VSWR derating chart while warning that duty cycle and ambient conditions matter. Use the exact manufacturer's envelope for the exact unit, then verify it under the intended operating mode.

Line Power Is Not Internal Component Stress

At 1.5 kW incident into a matched 50-ohm line, the travelling-wave values are approximately:

With 2:1 load SWR and the same 1.5 kW forward power, |Γ| = 1/3; a plain line can reach about 365 V RMS at a voltage antinode and 7.30 A RMS at a current antinode. The filter's internal resonators can experience still different voltages and currents because stored energy, coupling and loaded Q redistribute stress.

Do not multiply the 50-ohm line voltage by one generic unloaded-Q number. Derive component stress from the actual coupled network or measure it with a validated model and suitable probes. KYOCERA AVX's high-RF-power capacitor guidance requires voltage, current, power dissipation, thermal resistance and temperature rise to be checked together. Cornell Dubilier likewise defines ESR loss, RMS and peak current, voltage and thermal resistance as distinct capacitor limits.

Mismatch can detune the cascade, raise internal voltage or current and move dissipation among components. A full-power rating therefore requires a maximum load SWR, mismatch phase or test method, frequency range, duty cycle, airflow and temperature. A low-power VNA trace alone cannot demonstrate survival under heating, ferrite or core nonlinearity, contact arcing or thermal drift.

Wrong-Band Transmission and Hot Switching Need Hardware Prevention

Driving a single-band filter on the wrong band can present a severe mismatch and high internal stress. Whether equipment is damaged depends on forward power, amplifier foldback speed, filter response, relay state, line phase and component margins. The W3NQN article explicitly warns that its single-band filters must be switched correctly; that warning should become an interlock, not an operator memory exercise.

A robust controller should:

1. select the filter and antenna from validated band or frequency data;
2. use break-before-make routing and allow relay bounce to settle;
3. verify the commanded state with independent feedback where practical;
4. release transmitter or amplifier inhibit only after the RF path is valid;
5. wait for RF to decay before releasing or changing relays;
6. block prohibited band pairs and conflicting antenna assignments; and
7. fail inhibited on missing band data, controller disagreement or communications loss.

The microHAM Station Master Deluxe manual calls transmitter inhibit the best way to prevent hot switching. A controller that merely follows CAT data without controlling PTT cannot guarantee that the correct filter is settled before RF arrives.

Commission the Station, Not the Brochure

1. Build a directional band-pair matrix. List every aggressor fundamental, harmonic and victim band, because A-to-B need not equal B-to-A after switches and amplifiers.
2. Measure antenna coupling. Record calibrated end-to-end S21 at aggressor and victim frequencies for relevant headings, polarizations and switch states.
3. Characterize each two-port. Calibrate at the filter connectors and retain S11, S21, S12 and S22, exact settings, uncertainty and the measured floor.
4. Measure the installed cascade. Include relays, triplexers, jumpers, entry panels, inactive-port terminations and control wiring.
5. Separate carrier, noise and spur tests. Use a narrow source for blocking; measure transmitter noise density and discrete products after the PA through a protected, calibrated coupler or attenuator.
6. Test receiver degradation. Inject a calibrated weak wanted signal and increase aggressor power in controlled steps while monitoring sensitivity, reciprocal-mixing noise, ADC clipping and recovery.
7. Key combinations. Test every permitted simultaneous-transmitter state to reveal receiver and passive intermodulation.
8. Run a protected thermal and mismatch soak. Begin at low power, verify linear scaling, then test the declared waveform, duty, SWR, airflow and ambient limits while monitoring local component temperatures and filter drift.
9. Fault-test the interlock. Remove band data, request an illegal combination, interrupt communications and verify that RF remains inhibited.

Bottom Line

Band-pass filters do not fail as a class. A particular component can be under-specified, mistuned, overheated or damaged, but a compliant filter can also be blamed for interference that it was never positioned to remove. The correct question is whether the complete station meets a measured aggressor-to-victim requirement for every band pair and operating state.

Specify exact-frequency rejection, passband match and loss, waveform, average and peak power, duty cycle, cooling, permitted mismatch, component stress, switching state and measurement floor. Then verify carrier blocking, transmitter noise, harmonics, common mode, multi-signal products, thermal behavior and interlocks separately.

Questions, antenna-factor records or height trials to share? Contact RF.Guru.