Mixed Ferrite Cores in One Coax Choke: What Actually Adds

Different ferrite materials provide different complex permeability and loss versus frequency. Placing unlike closed toroids together and passing the same coax turns through all of them can deliberately sum complementary common-mode impedances. That is a legitimate topology. It becomes poor engineering only when the builder assumes the sum, glue joint, power rating or long-term reliability without measuring and qualifying the finished assembly.

Each Complete Toroid Keeps Its Own Magnetic Circuit

Consider two complete toroids stacked face to face, with each coax turn passing through both apertures. The common-mode current produces magnetomotive force in each core. Flux in toroid A circulates around toroid A; flux in toroid B circulates around toroid B. It does not need to cross the adhesive layer between the flat faces.

That distinction matters. The glue line between closed toroids is not inherently a series air gap in either toroid’s magnetic path. By contrast, when two E-, U- or pot-core halves form one magnetic circuit, contamination, adhesive thickness and clamping at the mating faces can change the effective gap and inductance. Advice for bonded core halves cannot be transferred unchanged to stacked rings.

The low-signal lumped model for the stacked toroids is:

This approximation is useful when the same common-mode current links both cores, the assembly is electrically compact and the winding remains in its lumped region. The finished impedance is the response of that core sum together with winding capacitances, leakage, cable and fixture—not those parasitics added as a simple series number. It is not permission to add catalogue |Z| values: complex quantities add vectorially, and product curves may use different geometries and fixtures.

One Winding Is a Design Constraint, Not an Automatic Defect

With the same N turns through each toroid, both materials receive the same ampere-turns. Fair-Rite’s technical catalogue relates suppression-core impedance to complex material permeability, core geometry and approximately N² in the lumped model. That shared turn count may be a useful broadband compromise when the measured material responses complement each other.

It may also be the wrong compromise. More turns usually increase low-frequency inductive impedance, but they lengthen the winding and increase turn-to-turn and turn-to-environment capacitance. Self-resonance moves, high-frequency impedance can collapse or change phase, and the cable bend radius may become unacceptable. A material that needed only one or two passes may be pulled beyond its useful resonant region by the turns required for another material.

The correct conclusion is conditional:

• Use a mixed stack when one turn count and geometry produce the required R+jX, loss and temperature across the complete band set.
• Use separate stages when each material needs different turns, spacing, cable or placement, or when independent replacement and inspection matter.
• Use identical cores when more ferrite volume of one characterized response is the actual requirement—not merely because identical material sounds more predictable.

None of those topologies wins without installed data.

Why Catalogue Impedances Do Not Simply Add

A scalar impedance curve hides phase. At one frequency material A might be mainly resistive while material B is mainly inductive; at another, winding capacitance may make one contribution capacitive. Adding 1 kΩ and 2 kΩ magnitudes does not necessarily make a 3 kΩ choke.

Fair-Rite’s suppression-ferrite guidance says source/load impedance, required attenuation, frequency, field strength, temperature, complex permeability and geometry are part of selection. TDK’s EMI application notes likewise describe normalized curves as guide values that depend on core and winding geometry.

Even individual-core measurements need careful fixtures. Fair-Rite’s test-wire study shows that conductor position and vector compensation can change measured suppression-core impedance. For a multi-turn coax choke, fixture residuals, reference planes, winding placement and common-mode launch symmetry become more important as impedance rises.

There is no universal multi-kilohm pass mark

A “2 kΩ” threshold is not a general bandwidth or safety specification. Current reduction depends on the common-mode source, return path and choke:

That simplified series model shows why the same choke can be ample in one installation and ineffective in another. Define the required reduction in current, pattern disturbance, received noise, equipment upset or accessible RF potential. Then measure enough R+jX bandwidth and thermal margin to meet that outcome.

Separate Stages Do Not Automatically Add Either

Two independently wound chokes placed directly adjacent on the same cable can often be approximated as series common-mode impedances. Separation creates a different circuit. The coax exterior between stages has inductance, capacitance, coupling to nearby conductors and electrical length; it can transform the second impedance or form a resonance.

Separate stages offer real advantages:

• independent core material, size, turn count and spacing;
• independent temperature sensing and replacement;
• the option to place each stage at a different current source or system boundary;
• easier fault isolation and service.

A compact mixed stack offers different advantages: one common-mode boundary, no intentional interstage cable segment, one winding and smaller packaging. The choice should follow a complete circuit and mechanical design, not a slogan that “separate is cleaner.”

Adhesive Is a Qualified Component

Adhesive suitability is product- and process-specific. Henkel’s current LOCTITE 401 product information, for example, lists ceramics among bondable substrates and specifies close-fitting parts, gap fill and an operating-temperature range. Those data still do not qualify it for a ferrite choke; the actual substrate, stress, RF and environmental requirements must be verified.

For a ferrite assembly, the selected product and process need documented answers for:

• adhesion to the actual ferrite coating or bare ceramic and any enclosure/support material;
• surface preparation, primer, permitted bond-line thickness, cure chemistry, time and temperature;
• elastic modulus, toughness, cure shrinkage and thermal-expansion behavior;
• glass-transition and service temperature—not merely decomposition temperature;
• thermal conductivity and whether the bond impedes or assists heat sharing;
• moisture, UV, salt, solvent and thermal-cycle ageing for the installation;
• dielectric strength, dielectric loss and RF voltage where adhesive approaches the winding;
• inspection, repair, disassembly and end-of-life requirements.

The importance of those properties is visible in a manufacturer data sheet. The 3M DP100 epoxy sheet, for example, reports glass-transition temperature, coefficients of thermal expansion in different temperature regions, thermal conductivity and electrical properties separately. Its much higher thermogravimetric weight-loss temperature is not the temperature at which stiffness, expansion or bond performance remains unchanged.

Core Stress Is Real—but the Outcome Is Process-Specific

Ferrite is a brittle ceramic. TDK warns that shock, rapid temperature change and tensile or cyclic load can crack a core, and that stress can lower initial permeability. Its processing guidance limits glue placement in certain assemblies specifically to reduce stress from differential thermal expansion.

Ferroxcube’s ferrite-core gluing study is equally useful because it avoids a universal verdict. Adhesive modulus, glass-transition temperature, bond thickness, cure process, moisture and thermal ageing changed the result; some tested bonds became stronger than the ferrite itself, so fracture moved into the ceramic. A stronger adhesive is not automatically a safer assembly if thermal stress or serviceability controls the design.

For stacked toroids:

• support the coax and enclosure separately so cable tension, vibration or drop load does not pass through the bond;
• avoid point clamps and uncontrolled cure pressure on the brittle rings;
• keep core markings and surfaces inspectable where possible;
• inspect after cure and after representative thermal cycles; repeat the impedance sweep because stress can alter magnetic behavior;
• reject chipped or cracked cores according to the component maker’s acceptance guidance.

For split or mated cores, also keep the mating faces clean and within the specified gap/clamping process. Adhesive wicked into that seam can directly change magnetic reluctance. That is a different failure mechanism from bonding two complete toroids face to face.

Mixed Materials Can Heat Unequally

In an ideal coaxial common-mode choke, equal-and-opposite differential currents largely cancel core magnetization. Net exterior/common-mode ampere-turns drive the ferrite. In a linear small-signal model, each material’s dissipation contribution is approximately:

The core with the largest resistive contribution at a given frequency can run hotter even though all cores carry the same linked common-mode current. As temperature rises, permeability and loss can change, moving the impedance and redistributing dissipation. The adhesive bond conducts some heat between rings while also adding a thermal interface; the direction and magnitude depend on bond thickness and thermal conductivity.

Actual heating also includes coax conductor/dielectric loss, connector loss, bending and enclosure temperature. High differential transmitter power is not by itself the ferrite drive, but it raises the consequence of winding asymmetry, leakage, mismatch, insulation stress and connector loss. A choke can also support RF voltage approximately proportional to I_CM|Z_CM|, so insulation and spacing remain part of the design.

“QRO” is therefore not a material or glue rating. Establish the operating envelope with the exact cable, connectors, cores, winding, enclosure, mismatch, frequency, waveform, duty cycle, ambient and mounting. Use low-power current mapping first; then increase power in controlled steps with remote observation of each core, the bond, coax and connectors. Stop before any component reaches its validated limit or shows thermal drift, odor, softening, cracking or impedance instability.

Serviceability Belongs in the Design

A permanent mixed stack is compact but locks material choice, turns and bond process together. Removing one core may chip both, damage the cable or hide the initiating failure. A removable enclosure with compliant supports can make inspection and replacement easier, but it too needs qualified temperature, flame, UV, moisture and mechanical performance.

Choose the retention method from the service plan:

Tape wrapped directly around bare cores is not automatically gentler: tension, creep, adhesive ageing, heat trapping and sharp-edge abrasion need evaluation. The absence of glue does not eliminate mechanical design.

A Measurement and Qualification Workflow

1. Define the outcome. Name the bands, required exterior-current or EMC reduction, maximum temperature, environment, mechanical load and service life.
2. Identify every part. Record manufacturer, material, part number, dimensions, coating, tolerances and lot where available; do not rely on color or seller labels.
3. Measure each candidate. With one repeatable common-mode fixture, capture R, X and |Z| for each core/material and same-material stack.
4. Build the shared winding mechanically. Respect cable bend radius and connector strain. Record turns, spacing, core order, cable and fixture reference planes.
5. Measure before bonding. Sweep the assembled mixed stack and compare complex values with the vector sum. Investigate resonance and fixture limits rather than forcing agreement.
6. Qualify the adhesive process. Follow its technical and safety sheets for surface preparation, gap, mix/dispense, cure and handling. Use representative coupons or sacrificial cores for mechanical/ageing tests.
7. Measure after cure and cycling. Inspect for chips/cracks and repeat R+jX after cure, temperature/humidity cycling and mechanical handling appropriate to the installation.
8. Map installed current. Verify that the choke reduces the intended common-mode path without moving unacceptable current to another cable or boundary.
9. Run stepwise thermal tests. Use representative frequency, mismatch, modulation and duty cycle. Observe individual cores, bond, coax and connectors remotely until stable.
10. Document service limits. Record maximum validated temperature/current/power conditions, inspection interval, reject criteria and whether the unit is repairable.

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