Collins Mechanical Filter Testing

How to Verify Bandwidth, Shape Factor & Insertion Loss — With and Without a Spectrum Analyser

Two complete test methods: Method A using a spectrum analyser with tracking generator, and Method B using only a signal generator and oscilloscope — covering impedance matching, the point-by-point sweep technique, and interpretation of results for F455 and F500 series filters

Why Test Mechanical Filters?

Collins mechanical filters are among the most reliable components in vintage Collins equipment — they have no active elements to age and no capacitors to leak. However, they can be damaged by physical shock, by DC voltage applied to the transducer windings (a failure mode addressed by 75A-4 SB-1), or by substitution of filters with different impedance characteristics. A filter that has been dropped may have shifted centre frequency, increased passband ripple, or degraded skirt selectivity. A filter removed from a different receiver model may have incompatible source/load impedance requirements that produce severe passband ripple even though the filter itself is undamaged.[1]

Testing a mechanical filter allows you to verify three critical parameters: bandwidth (6 dB and 60 dB), shape factor (the ratio of 60 dB bandwidth to 6 dB bandwidth — a measure of skirt steepness), and insertion loss (the signal attenuation through the filter at the centre of the passband). These three numbers completely characterise a filter’s health and suitability for a given application.

Critical: Impedance Matching

Mechanical filters are designed for specific source and load impedances. If the filter is not terminated in its design impedance, the passband shape will be distorted — producing ripple, asymmetry, or apparent bandwidth changes that are measurement artefacts, not filter defects. The impedance requirement must be satisfied for any measurement to be valid.[2]

Collins mechanical filters of different eras have different impedance requirements. Early filters (F455B, F500B series used in the 51J-4 and 75A-4) typically have impedances from a few hundred ohms to several kilohms. Later filters (F455FA series used in the 32S-3/KWM-2, F455N series used in the R-390A) may be as high as 100 kΩ. The impedance is determined by the transducer winding and is specified in Collins filter catalogues and the CCA Tech Bulletin Issue 2 (526-9605 specification).[3]

⚠ Most test equipment has a 50Ω output impedance. Collins mechanical filters are NOT 50Ω devices. You must build an impedance matching network between the signal generator output and the filter input, and between the filter output and the measurement instrument input. Failure to do this will produce grossly misleading results — passband ripple of 10–15 dB in a filter that is actually working correctly.[1]

Building the Impedance Matching Network

The matching network consists of a resistive pad that transforms the 50Ω generator impedance to the filter’s design impedance. For a filter with input impedance Z_in and a generator output of 50Ω:

Series Resistor: R_series = (Z_in − 50) / 2 ohms, placed between the generator output and the filter input.

Shunt Resistor: A shunt resistor from the generator output terminal to ground provides the correct parallel combination so the generator sees 50Ω. For Z_in = 500Ω, R_series ≈ 225Ω, and R_shunt ≈ 53Ω.

Output Matching: An identical matching network is required at the filter output, transforming the filter’s output impedance to the measurement instrument’s input impedance.

For high-impedance filters (F455N series at ~100 kΩ), a simple resistive pad introduces too much loss. In this case, use a tuned L/C matching network or accept the impedance mismatch and interpret the results accordingly — the centre frequency and insertion loss will be correct, but the passband ripple may be higher than the filter’s true specification.[2]

Practical Shortcut: For a quick go/no-go test of a Collins filter intended for use in the original receiver, the simplest approach is to install the filter in the receiver and perform the receiver IF alignment. If the filter produces a clean passband with the expected bandwidth (verified by tuning through a known signal), it is almost certainly healthy. The bench test methods below are for stand-alone filter evaluation outside the receiver.

What to Measure — The Three Key Parameters

Parameter	Definition	What It Tells You	Healthy Value
6 dB Bandwidth	Width of the passband at the points where response is 6 dB below the peak	Usable passband for signal reception — should match the filter type number (e.g., F455FA-21 = 2.1 kHz)	Within ±10% of specification
60 dB Bandwidth	Width of the response at 60 dB below the peak	Stopband rejection width — determines how well the filter rejects adjacent signals	Within specification (filter catalogue)
Shape Factor	Ratio of 60 dB bandwidth to 6 dB bandwidth	Skirt steepness — lower is better. Typical Collins mech. filters achieve 1.5:1 to 2.5:1	Per filter type — typically <3:1
Insertion Loss	Signal attenuation at the centre of the passband compared to a direct through connection	How much signal the filter absorbs — excessive loss indicates damaged transducers	Typically 2–6 dB depending on type
Passband Ripple	Variation in response level across the passband	Flatness of the passband — excessive ripple may indicate impedance mismatch OR internal damage	Typically <2 dB within the 6 dB BW

Method A — With Spectrum Analyser & Tracking Generator

This is the gold-standard measurement. A spectrum analyser with a tracking generator provides a swept CW signal synchronised with the analyser’s display, producing a direct plot of the filter’s frequency response on screen. Modern instruments (tinySA Ultra, Siglent SSA3000X, Rigol DSA800 series) are affordable and have more than adequate resolution for Collins filter measurements at 455 kHz or 500 kHz.[4]

Equipment Required

Spectrum analyser with tracking generator (output and input both typically 50Ω); impedance matching pads for the filter under test (input and output); short BNC patch cables; known-good through connection (a BNC barrel or short cable) for the reference sweep.

Procedure

1Calibrate the through path: Connect the tracking generator output through the input matching pad, a short through cable, and the output matching pad to the spectrum analyser input. This establishes the 0 dB reference level. Store this trace as the reference (or use the analyser’s “normalise” function if available).

2Set the span: For a Collins 455 kHz filter, set the centre frequency to 455.000 kHz and the span to 10–20 kHz (enough to see both passband and stopband). For a 500 kHz filter (51J-4), use 500.000 kHz centre. Set the RBW (resolution bandwidth) to the minimum available — typically 100 Hz or less for meaningful measurements on narrow CW filters.

3Insert the filter: Replace the through cable with the filter under test (between the matching pads). The analyser display will now show the filter’s frequency response as a direct comparison to the reference level.

4Read the parameters: Use the analyser’s marker functions to find the peak response (this is the centre frequency, and the distance below the reference is the insertion loss). Place markers at the −6 dB points on each skirt to read the 6 dB bandwidth. Place markers at the −60 dB points to read the 60 dB bandwidth. Calculate shape factor = (60 dB BW) ÷ (6 dB BW).

5Check passband ripple: Zoom the span to just the passband (e.g., 5 kHz span for a 2.1 kHz filter). Examine the passband flatness. Any peak-to-trough variation greater than 2 dB indicates either an impedance mismatch in the test setup or internal filter damage.

Method B — Without a Spectrum Analyser (Signal Generator & Oscilloscope)

This method uses a signal generator and an oscilloscope to perform a point-by-point frequency sweep. It requires more patience but produces equally valid results. A modern oscilloscope with a built-in Bode plot function (Siglent SDS series, Tektronix 2/4/5/6 Series MSO) can automate the sweep and produce a frequency response plot directly on screen.[5]

Equipment Required

Signal generator capable of 455 kHz (or 500 kHz) with fine frequency resolution (10 Hz steps or better); dual-channel oscilloscope (or single-channel with sequential measurement); impedance matching pads; BNC cables; graph paper or spreadsheet for recording data points.

Manual Point-by-Point Procedure

1Connect the test setup: Signal generator → input matching pad → filter under test → output matching pad → oscilloscope Channel 2. Optionally, connect a direct feed from the generator (via a BNC tee) to oscilloscope Channel 1 as a reference level.

2Set the generator: Set the signal generator output to a constant level (e.g., 100 mV p-p into 50Ω). This level must remain constant throughout the entire sweep — do not adjust it once the sweep begins.

3Measure the reference level: Set the generator to the expected centre frequency (455.000 kHz or 500.000 kHz). Read the oscilloscope Channel 2 amplitude. This is your 0 dB reference for the filter under test. Record it.

4Sweep the passband: Starting from 2–3 kHz below the centre frequency, step the generator frequency in increments of 50–100 Hz. At each step, record the frequency and the oscilloscope Channel 2 amplitude. Continue through the centre frequency and out to 2–3 kHz above centre. For the full 60 dB measurement, extend the sweep to 5–10 kHz either side of centre, using larger steps (200–500 Hz) in the stopband where the response changes slowly.

5Calculate dB at each point: For each data point, calculate the response relative to the reference: dB = 20 × log₁₀(V_point / V_reference). This gives a negative number for all points except the peak.

6Plot the response: Plot frequency (x-axis) vs. dB (y-axis). The result is the filter’s frequency response curve. Read the 6 dB and 60 dB bandwidths directly from the plot. Calculate shape factor and insertion loss as in Method A.

Sweep Rate Warning: When using a swept (not stepped) signal generator, the sweep rate must be extremely slow for mechanical filters. Mechanical resonators have a finite settling time — if the sweep passes through a resonance too quickly, the filter output will not reach its true steady-state amplitude, and the apparent bandwidth will be wider (and the apparent insertion loss lower) than the true values. A sweep rate of no more than 100 Hz per second is recommended for 0.5 kHz CW filters; wider filters can tolerate faster sweeps.[2]

Automated Bode Plot Method (Modern Oscilloscope)

Many modern oscilloscopes with built-in function generators offer an automated Bode plot feature that steps the generator through a frequency range, measures the amplitude at each step, and plots the result directly on screen. This automates the manual point-by-point procedure above.[5]

1Connect the oscilloscope’s internal generator (or a synchronised external generator) through the matching pads and filter as described above. The generator output goes to Channel 1 (reference) via a BNC tee and to the filter input. The filter output goes to Channel 2 (through signal).

2Configure the Bode plot function: set the start frequency to (centre − 5 kHz), the stop frequency to (centre + 5 kHz), the amplitude to a constant level suitable for the filter’s transducer (typically 50–200 mV), and the points per decade to the maximum available (10+ points per decade for adequate resolution on a 2.1 kHz filter).

3Run the sweep. The oscilloscope will step through the frequency range, measure the amplitude ratio (Channel 2 / Channel 1) at each point, and display the frequency response curve directly. Use cursors to read the −6 dB and −60 dB points.

Important: Set the oscilloscope’s load impedance correctly — if your scope has switchable 50Ω / 1 MΩ input, use the impedance that matches your matching network design. If using 1 MΩ inputs (most common on general-purpose oscilloscopes), include the termination resistor at the end of the output matching pad.

Interpreting the Results

Observation	Probable Cause	Action
Bandwidth and shape factor match spec; insertion loss <6 dB; ripple <2 dB	Filter is healthy	No action needed — filter is good for service
Excessive passband ripple (>3 dB) but centre frequency and bandwidth appear correct	Impedance mismatch	Check and recalculate matching pads; verify source and load impedance values. If tested in-circuit, check the tube stage source/load impedance
Centre frequency shifted >200 Hz from nominal	Physical damage (shock)	Filter may have been dropped. Slight shift (<100 Hz) can be compensated by IF alignment; larger shifts indicate permanent damage
Bandwidth significantly wider than spec	Coupling rod damage or transducer failure	Filter is likely damaged internally. Replace.
Insertion loss >10 dB with correct impedance match	Transducer winding damage (possibly from DC voltage)	Check for evidence of DC voltage damage (75A-4 SB-1 failure mode). Filter may be repairable by specialist.
Multiple sharp peaks and nulls within the passband	Wrong filter type for the circuit (F500F in 51J-4 expecting F500B impedance)	Verify the filter type number matches the equipment’s specification. Different case types have different impedances.[1]
No output at all	Open transducer winding or broken internal connection	Measure DC resistance across the filter pins (should be low ohms, not open). If open, filter is dead.

Common Collins Filter Impedance Reference

Filter Series	Equipment	IF Frequency	Nominal Impedance
F455FA	32S-3/3A, KWM-2/2A (SSB generation)	455 kHz	Per catalogue — typically mid-impedance
F455J	75A-4 (plug-in, 9-pin miniature socket)	455 kHz	Per catalogue — typically mid-impedance
F455N	R-390A (FL501–FL505)	455 kHz	~100 kΩ in/out
F500B	51J-4	500 kHz	Per catalogue — check individual filter spec
F455K	75S-3B/3C (accessory CW/narrow)	455 kHz	Per catalogue

Cross-Reference: See the separate Collins IF Selectivity Filter Master Reference on this site for the complete filter type number decoder, case style identification, part number cross-reference, and per-filter specifications across all Collins equipment.

References & Citations

Antique Radio Forums. Collins 51J-4 Troubleshooting Help Needed. F500F vs F500B impedance mismatch causing 15 dB passband ripple; 600Ω resistive padding workaround; filter bandwidth verification in-circuit. April 2008. antiqueradios.com — 51J-4 Filter
R-390A.net / Chuck Rippel. IF Deck Alignment — Mechanical Filter Testing. Collins filter impedance matching for test equipment; resistive pad calculations for 50Ω generator to filter impedance; sweep rate requirements; F455N series 100 kΩ impedance specification. r-390a.net — IF Deck Filters (PDF)
Collins Collectors Association — CCA Tech Bulletin Issue 2: Collins Mechanical Filters Specification 526-9605. Complete filter parameter reference including impedance, bandwidth, shape factor, insertion loss, and case style specifications. October 2024. collinsradio.org — CCA TB-2 (PDF)
OE3HBW. Collins 455 kHz Mechanical Filter — F455-21Y Testing. Bench test of surplus F455-21Y (526-9337-00) from 75S-1; physical description, magnetostrictive transducer principle, frequency response measurement. oe3hbw.eu — F455 Test
Siglent. Bode Plot of a Filter Using an Oscilloscope and Function Generator. Automated Bode plot technique: dual-channel reference/through measurement, impedance matching with BNC terminators, sweep parameter configuration, CSV data export for offline analysis. Siglent — Bode Plot Application Note
WA3KEY. Rockwell/Collins Mechanical Filters. Complete filter cross-reference: type number to Collins P/N, equipment application, bandwidth, and case style for all Collins receivers and transmitters. wa3key.com — Mechanical Filters
Kolb, JL. Collins Mechanical Filter Identification. Filter type number decoding system: centre frequency, case type letter, bandwidth code, symmetrical vs. SSB designation (Z suffix), Collins P/N structure (526-9xxx vs. 526-8xxx custom orders). jlkolb.cts.com — MF Identification
EDN / Arthur Pini. Measure Frequency Response on an Oscilloscope. Swept sine, white noise, and impulse/step methods for frequency response measurement; FFT max-hold technique; sweep rate considerations. December 2015. EDN — Frequency Response
K6JRF. Collins Mechanical Filter Testing for FT1000D. Bench testing Collins F455FA-31 for SSB bandwidth optimisation; frequency response plots; filter impedance matching in non-Collins applications. k6jrf.com — Filter Testing

Credits & Acknowledgments

Collins Collectors Association (CCA) — For CCA Tech Bulletin Issue 2 documenting the Collins 526-9605 mechanical filter specification, and for maintaining the filter cross-reference data through the CCA technical library.

Chuck Rippel — For the R-390A IF deck alignment guide that includes the mechanical filter impedance matching calculations for test equipment, which applies to all Collins filter testing.

WA3KEY — For maintaining the comprehensive Collins/Rockwell mechanical filter cross-reference table that maps type numbers, part numbers, and equipment applications.

JL Kolb — For the filter identification guide that decodes the Collins type number system into centre frequency, case type, and bandwidth.

OE3HBW — For the bench test documentation of a surplus F455-21Y filter demonstrating practical measurement technique.