1.5 kW Multi-2 Band-Pass Filters: What Five Poles Do—and Do Not Prove

A high-power band-pass filter can be one of the most valuable parts of a Multi-2 station. Its useful specifications are rejection at each exact aggressor–victim frequency pair, passband S-parameters before and after heat soak, a defined power and mismatch rating, controlled leakage paths and a switching system that cannot apply RF to the wrong filter.

First Define What “1.5 kW” Means

The arithmetic in this article uses 1,500 W incident forward power at the filter input, a nominal 50 Ω system and a steady carrier unless another condition is stated. That is a demanding thermal test condition. It must not be silently substituted for 1.5 kW PEP SSB, whose time-averaged power depends on speech processing and operating pattern.

Nor is 1.5 kW a universal legal entitlement. The current Belgian BIPT page still points to its 24 May 2019 decision: a class-A certificate permits 1,500 W in many contest-band segments, but other bands and segments have lower limits. In the United States, 47 CFR § 97.313 generally caps amateur transmitter power at 1.5 kW PEP and imposes lower limits in specified cases; § 97.3 defines PEP. The operator must use the limit and power definition that apply to the station, licence, band and location.

“ICAS” is not enough by itself. A manufacturer must state the on/off cycle, maximum key-down time and recovery conditions it intends. RTTY, FT8, FM, a steady test carrier and contest SSB impose different average heat even when their PEP readings are identical.

Five Resonators Buy Design Freedom—not a Guaranteed Number

More resonators give a designer more degrees of freedom for passband shape and stopband slope. Useful rejection still depends on the response family and coupling matrix, finite-frequency transmission zeros, resonator unloaded Q, parasitic self-resonances, enclosure partitioning, connector and control-line leakage, component tolerances and temperature.

A three- or four-resonator elliptic response can beat a generic five-resonator filter at a selected stop frequency. A well-executed five-resonator design may instead provide a wider and more repeatable stopband. Therefore compare the minimum measured rejection across every permitted transmit segment, not the pole count or the deepest marker on one typical trace.

The classic W3NQN work is valuable design history but must be applied within its scope. The 1998 filters were intended between roughly 150 W transceivers and 1,500 W amplifiers and were designed for about 200 W ICAS; the paper explicitly distinguished them from 1,500 W output-filter components. The 1999 receiver-filter paper reported 60–80 dB at the centres of adjacent bands for low-power receiver preselectors. Neither paper, by itself, certifies a modern post-PA five-pole filter at 1.5 kW.

Filter Placement Decides Which Interference It Can Remove

For the other station’s fundamental, the victim-side band-pass filter normally supplies the relevant filter rejection; the transmitting station’s filter passes its own fundamental. At the victim frequency, the source-side post-PA filter is the element that can suppress PA noise or a discrete harmonic before it is radiated. A receiver filter cannot remove interference that lies inside its own passband.

A passive stopband is commonly reflective. The rejected harmonic energy is not necessarily converted to heat inside the filter; much of it can return toward the amplifier. That changed harmonic termination may affect PA voltage, efficiency, stability or distortion. Validate the amplifier and filter together instead of assuming that more stopband reflection can only improve the transmitter.

Keep Three Interference Budgets Separate

1. The other transmitter’s fundamental as a blocker

C_ANT is end-to-end antenna-path coupling loss and A_victim is rejection in the victim branch at the aggressor frequency. This out-of-band carrier can cause gain compression, reciprocal mixing, ADC clipping, false responses, receiver-generated intermodulation or damage.

2. Broadband transmitter noise inside the receive channel

Use the measured transmitter-plus-PA noise density at the relevant offset and operating power, then integrate over the receiver’s equivalent noise bandwidth. A generic dBc/Hz value scaled from carrier power can be misleading if the PA noise does not scale linearly or the spectrum is not flat across the receive bandwidth.

3. A discrete harmonic or spur in the victim band

A discrete harmonic is an integrated spectral line, not a noise density. Do not add 10 log(B) to it. Harmonic-related band pairs deserve explicit tests—for example 80/40, 40/20 and 20/10—together with a wider sweep for parasitic filter passbands.

A Corrected 1.5 kW Blocker Example

A 1,500 W carrier is:

Assume 60 dB of measured antenna-path isolation and treat 45, 55 and 70 dB as victim-side rejection at the aggressor fundamental. These are illustrative values, not a claim for a particular product.

*The IARU Region 1 recommendation uses S9 = −73 dBm below 30 MHz and 6 dB per S-unit. This is only a familiar level comparison. An actual S-meter, receiver blocking threshold and absolute-maximum input rating are three different things.

The original 70 dB row omitted the −68.24 dBm cell and shifted the remaining entries into the wrong columns; the table above corrects that error. The 15 dB improvement from 55 to 70 dB reduces blocker power by 31.6:1, but only a receiver test can show whether it improves usable sensitivity.

Required Isolation Comes From a Measured Receiver Criterion

None of those blocker targets is universal. Establish the target with the exact receiver, preamp/attenuator state, preselector, bandwidth and AGC configuration. The ARRL laboratory method raises a controlled interferer while monitoring a calibrated wanted signal and identifies the level producing 1 dB gain compression; it also warns not to exceed the receiver manufacturer’s input limit. ETSI EN 301 783 uses a wanted-plus-unwanted-signal method for commercial amateur equipment, but passing a product conformance test does not guarantee a compact Multi-2 station will meet its own weak-signal objective.

Reducing power helps only by the decibel change in transmitter power. Going from 1.5 kW to 400 W changes 61.76 dBm to 56.02 dBm, an improvement of 5.74 dB. It cannot erase a 15–25 dB rejection deficit. Conversely, an isolation calculation that produces a very large permissible wattage is not a legal limit, thermal rating or switch rating.

Insertion Loss Is Not Automatically Heat

The earlier version equated all power not transmitted through S21 with heat. That is an upper bound only: some power may be reflected at the input. For a passive two-port with port 2 matched and negligible radiation or extra ports:

The last expression requires S11 and S21 from the same frequency, temperature, power state and reference planes. With a mismatched load, use the complete two-port and load reflection coefficient; a single insertion-loss number is insufficient.

*A 20 dB return loss means 1% of incident power is reflected in this matched-output example. The values are arithmetic illustrations, not temperature predictions.

Even tens of watts can be concentrated in capacitor ESR, coil resistance, joints or one poorly coupled resonator. Temperature rise depends on thermal resistance, airflow, mounting, ambient temperature and time. A cool enclosure wall does not prove that internal capacitors, solder joints or coil supports are within their ratings.

Duty cycle and heat soak

If on-state loss remains linear, long-term average dissipation may roughly follow RF duty factor. Component peak voltage and current do not. Short bursts can still arc a capacitor or connector, while a high-duty digital mode can reveal a thermal limit that intermittent SSB never reaches. State the waveform and cycle, monitor the likely internal hot spots with appropriate isolated instrumentation and continue until thermal equilibrium or the manufacturer’s defined intermittent cycle is complete.

Matched-Line Numbers Are Only the Starting Point

At 1,500 W forward power in a matched 50 Ω line:

• VRMS = √(PZ0) = 273.9 V and Vpeak = 387.3 V;
• IRMS = √(P/Z0) = 5.48 A and Ipeak = 7.75 A.

For the same forward power on an ideal lossless line with load reflection magnitude |Γ|, line maxima become V+ (1 + |Γ|) and I+ (1 + |Γ|). For the same net transported power, the familiar maxima scale as √SWR. These are different test conditions.

A real filter adds its own impedance transformation. Its input reflection under a mismatched load is:

Therefore load-SWR magnitude alone does not locate the worst component stress; reflection phase, intervening line length and frequency matter. A serious mismatch qualification sweeps the permitted reflection-coefficient circle, includes tuning and fault states and checks both electrical stress and temperature.

Resonator unloaded Q influences loss and selectivity, but internal capacitor voltage or coil current is not obtained by simply multiplying the 50 Ω line value by Q. Those stresses depend on topology, impedance level and coupling. Derive them from the actual network model and confirm them with rated probes, couplers or validated high-power tests.

Why 70 dB on a Plot May Not Be 70 dB in the Station

Seventy decibels is a 10,000,000:1 power ratio. At that point the result may be limited by something other than the intended differential filter path:

• input-to-output electric or magnetic coupling inside the enclosure;
• seams, ventilation, connectors and relay-control wiring;
• common-mode current on feed lines and control cables;
• switch, triplexer, dummy-load or interlock-box isolation;
• a test fixture or instrument leakage floor; or
• a different antenna-coupling path from the one used in planning.

Keysight defines network-analyzer dynamic range from usable source or receiver level to the receiver noise floor and shows that IF bandwidth, averaging and test-set configuration change the floor. A valid rejection trace needs margin below the claimed response, not merely a line that has flattened at the bottom of the display.

Small-signal VNA acceptance

1. Perform a full two-port calibration at the filter connectors and record the calibration method, cables, adapters, source power, IF bandwidth, averaging and reference planes.
2. Measure S11, S21, S12 and S22 over every passband, every simultaneous band pair and the relevant harmonic range. Save the raw Touchstone file, not only screenshots.
3. Use segmented sweeps if necessary: low enough power for linear passband measurement, and adequate source power plus narrow IF bandwidth in deep stopbands.
4. Establish the measurement floor with a shielded isolation check and repeat after reconnecting or reversing the DUT. A trace that follows the floor is a lower-bound claim, not a measured filter response.
5. Repeat after thermal stabilization. Record passband loss, return loss, centre shift and minimum stopband rejection across the complete operating segments.

Switching and Interlocks Are Part of the Filter Rating

A permanently installed one-filter-per-band architecture removes some switching risk. A switched filter bank must prevent a transmitter from keying into the wrong, open or moving path. Relay carry power and hot-switch power are different specifications, and relay operating time does not necessarily include bounce, amplifier tail RF or network latency.

A defensible sequence is:

1. Assert transmitter and amplifier inhibit. Do not rely only on CAT band data, VOX or an application message.
2. Confirm RF has fallen below the switch maker’s permitted level. Include amplifier turn-off tail and stored energy.
3. Change the path break-before-make. Reject impossible combinations and prevent two transmitters from sharing a forbidden branch or antenna.
4. Wait for the worst specified operate, release and bounce time plus engineering margin.
5. Verify the commanded state where practical. A coil command is not proof that RF contacts moved.
6. Release inhibit only after every filter, antenna switch, triplexer and load state is valid. On disagreement, time-out or loss of controller communications, fail closed with RF inhibited.

Current station-control manuals reflect this principle. The microHAM TRIO manual retains amplifier keying after radio PTT release and uses an inhibit signal during switching; its timing is product-specific, not a universal 30 ms rule. 4O3A’s current two-port antenna-switch literature adds dedicated PTT-in/PTT-out ports for hot-switch prevention. Validate the actual end-to-end sequence with an RF detector and oscilloscope before enabling the amplifier.

Commission the Multi-2 Station as a System

1. Build a band-pair matrix. Include both directions, band edges, harmonic relationships, antenna headings, switch states and every permitted simultaneous combination.
2. Measure installed coupling at low power. Record transmitter-output-to-receiver-input S21 with the real antennas, filters, switches, triplexers, loads, cables and bonds.
3. Measure source cleanliness separately. At controlled power, distinguish broadband noise density from discrete harmonics and spurs while keeping the analyzer protected and below compression.
4. Test receiver degradation. Hold a calibrated weak wanted signal at a defined level and raise the blocker cautiously. Record the point for 1 dB gain loss, SINAD or decode degradation, reciprocal-mixing noise, ADC overload and false responses. Never exceed the receiver manual’s absolute input limit.
5. Raise transmitter power progressively. Confirm that interference scales as expected. Unexpected nonlinear growth indicates receiver, transmitter, connector or passive-junction intermodulation.
6. Run the thermal and mismatch envelope. Use the intended waveform, average power, key-down cycle, ambient, cooling and worst credible SWR magnitude and phase.
7. Verify interlocks under faults. Test wrong-band commands, controller reboot, lost network, stuck relay feedback, rapid band changes and PTT during transition at low power.
8. Recheck hot performance. After soak, repeat passband SWR/loss, rejection and receiver-degradation tests and archive the data by filter serial number.

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