Magnetizing Reactance Is Not Common-Mode Isolation

When a voltage is applied across a winding, sufficient magnetizing inductance keeps the magnetizing-current burden under control. The required value is finite and application-specific: increasing it beyond the useful operating window does not by itself improve power capability, common-mode rejection or safety isolation.

Four Parameters, Four Different Jobs

The simple transformer equivalent circuit is only a starting point. Winding resistance, core-loss resistance, leakage inductance and multiple distributed capacitances create a frequency-dependent network. Above its first resonances, one measured “inductance” value cannot represent the component.

What Magnetizing Reactance Actually Tells You

For a linearized inductance at one frequency:

A larger X_L at the lowest operating frequency reduces the reactive current drawn by that branch. In an audio or broadband impedance transformer, this can reduce low-frequency loading and phase error. In many hard-switched power transformers, it reduces no-load or circulating magnetizing current. That is useful, but “large enough” must be derived from the circuit.

Suppose a resistive load reflected to the primary is R′ and the ideal magnetizing branch has X_L = 10R′. At the same winding voltage:

That idealized 10:1 ratio means 10% quadrature excitation current, not “99% power transfer.” Source impedance, winding resistance, core loss, leakage, waveform, mismatch and converter topology determine the actual loss and voltage transfer. Ten-to-one is an illustration, not a universal design rule.

Flux and Loss Set Harder Limits Than One Inductance Reading

Faraday's law connects winding voltage to flux. For an ideal sinusoid:

For a nearly constant voltage applied during a switching interval:

Here N is turns and A_e is effective core area. Flux imbalance, duty-cycle error and insufficient reset can walk a core toward saturation even if a small-signal LCR measurement looked excellent. Core loss also depends strongly on frequency, waveform, flux density and temperature. TDK's official Magnetic Design Tool exposes amplitude and complex permeability, power loss versus frequency, flux density and temperature, and topology-dependent transferable-power calculations—an appropriate reminder that permeability and saturation/loss are different material properties.

More turns can raise L_m approximately with N² and reduce flux per volt, yet also increase copper length, DC and AC resistance, winding volume and distributed capacitance. A larger core may obtain the required volt-seconds, copper area and inductance with fewer turns, but it is not automatically lower loss or safer; it still needs a complete design.

Some resonant converters deliberately use magnetizing inductance as part of their operating mechanism. The official TI UCC25800-Q1 data sheet, for example, calculates an L_m target from zero-voltage-switching requirements: too large gives too little magnetizing current for full ZVS, while too small adds conduction loss. That product-specific example disproves “maximize L_m” as a universal instruction.

Open-Secondary Inductance Is a Measurement Condition

Magnetizing inductance is commonly estimated at one winding with the other windings open. The result depends on test frequency, voltage or current amplitude, DC bias, temperature, core assembly and the instrument's series/parallel equivalent-circuit choice. At sufficiently high frequency, capacitance contaminates the result; near a resonance, the displayed inductance can become meaningless or change sign.

The Keysight Impedance Measurement Handbook distinguishes the required setups: windings open for primary/secondary inductance, a low-impedance secondary short for leakage inductance, and a separate connection for interwinding capacitance. It explicitly notes that a direct inductance reading includes capacitance effects unless an equivalent-circuit analysis separates them.

Record the fixture, terminal connections, frequency, amplitude, bias, temperature and winding state with every number. An unqualified “L = 1 mH” is not a broadband model.

Common-Mode Excitation Is a Different Port

Under ideal common-mode excitation, both terminals of one winding move together relative to another winding, chassis or earth. Their differential voltage is zero, so the magnetizing branch is not driven in that idealized mode. Current can still flow through electric-field paths from every turn to every nearby conductor.

Real circuits are not perfectly common-mode. Unequal capacitances, winding gradients, lead inductance, rectifiers, shields, source impedance and load imbalance convert some common-mode energy to differential mode and vice versa. A common-mode choke is also a different magnetic structure: its intended common-mode currents produce adding core flux, whereas differential load currents ideally cancel. Do not transfer a common-mode-choke impedance argument to an ordinary isolation transformer without modeling the actual ports.

Capacitance Creates Displacement Current

For one lumped capacitance with voltage v(t):

The scale can be surprising. One picofarad subjected to 1 V/ns produces an instantaneous displacement current of 1 mA while that slew rate persists. A lumped 100 pF path has |X_C| ≈ 159 Ω at 10 MHz. These are arithmetic examples, not predictions of cable current: the actual voltage distribution and all parallel/series return impedances set the result.

Texas Instruments' 2025 measured study How Isolated Bias Transformer Parasitic Capacitance Impacts EMI Performance compared several 15 V, 1.5 W isolated bias topologies. Its tested flyback transformer combined the lowest leakage inductance with the highest parasitic capacitance, while an LLC implementation accepted more leakage and achieved much lower capacitance. The result supports a geometry/topology trade, not a universal 2 pF or 20 pF specification.

Leakage and Capacitance Form a Design Trade

Close winding spacing, broad overlap and interleaving generally improve coupling and reduce leakage inductance. They also generally increase interwinding capacitance. A split bobbin or deliberate separation can reduce capacitance, but usually increases leakage, worsens regulation or bandwidth, and can increase switching overshoot and snubber loss.

TI's primary-source design seminar Magnetics Design 3—Windings documents this exact conflict. It also shows that winding capacitance resonates with mutual and leakage inductances. A first-order resonance estimate is:

A real transformer has several resonances, distributed fields and frequency-dependent loss, so one self-resonant-frequency number is only a landmark. Sweep complex impedance and transfer response beyond the intended band, with a fixture that represents the installed return path.

A Faraday Shield Is Not “Ground This Foil”

An electrostatic shield can intercept displacement current between windings, but its construction and termination are topology-specific. The shield must not form a closed conductive turn around the core; an overlap must remain electrically insulated. Its termination needs low inductance at the frequencies of interest and must route noise to a deliberate quiet node without exporting it through another cable or protective conductor.

The TI winding guide recommends its SMPS shield connection to the quiet side of the primary and warns that grounding it can propagate common-mode EMI to the input. Other equipment architectures may require chassis or protective-earth strategies under their product standard. That is why “ground the shield” is not a general prescription.

A shield also does not replace basic, supplementary or reinforced insulation. Its foil, lead-out, overlap and termination become conductors within the insulation system and must preserve required distances, solid insulation and fault behavior.

Functional EMI Isolation Is Not Protective Isolation

Finite capacitive current through intact insulation is normal. It may be undesirable for EMI, analog accuracy or user-perceived “tingle,” but it is not evidence that the dielectric has failed. Conversely, a low capacitance or impressive high-frequency CMRR trace does not prove that the transformer safely withstands its working voltage, transients, pollution, altitude, temperature and faults.

Why there is no universal touch-current number here

For one ideal 1 nF capacitor on 230 V RMS, 50 Hz, the sinusoidal current magnitude is:

That calculation is useful for an initial estimate, but it is not an IEC touch-current result. Waveform, frequency weighting, accessible part, equipment class, connection, normal/fault state and other current paths matter. IEC 60990:2016 defines measurement methods and body-impedance networks for touch and protective-conductor current; it explicitly does not set the product limit. The applicable product standard does.

Standards belong to a defined product

• IEC 60664-1:2020 provides insulation-coordination principles, requirements and tests for low-voltage supply systems, including the basis for creepage, clearance and solid-insulation criteria.
• IEC 62368-1:2023 is a current hazard-based product safety standard for audio/video, information and communication technology equipment. It is relevant only when the product falls within that scope.
• IEC 61140:2016 states common principles for protection against electric shock; product committees translate those principles into applicable requirements.
• IEC 60384-14:2023 covers safety-tested capacitors intended for EMI suppression and connection to the supply mains. A deliberate capacitor across an isolation boundary must use the correct safety class and application permitted by the relevant product standard.

Required distances and tests depend on factors such as insulation type, working and transient voltage, overvoltage category, pollution degree, material group, altitude, waveform/frequency and manufacturing controls. A generic transformer article cannot select them safely.

A Measurement Plan That Separates the Mechanisms

1. Define the application and reference planes. Record topology, turns ratio, voltage waveform, frequency range, source/load impedances, isolation function, cable paths and applicable safety standard.
2. Measure open-secondary impedance. Sweep complex impedance over frequency at documented amplitude, bias and temperature. Extract L_m and core-loss behavior only where the chosen equivalent circuit is valid.
3. Measure leakage correctly. Short the secondary with the lowest practical fixture impedance and measure the primary; state the fixture residual and frequency.
4. Measure winding resistance and AC loss. Include temperature, skin and proximity effect, not only room-temperature DC resistance.
5. Characterize electric-field paths. Measure interwinding and winding-to-core/shield capacitances with explicit terminal connections. Then measure common-mode transfer or admittance with the installed source, return and termination represented.
6. Find resonances and mode conversion. Sweep impedance or mixed-mode transfer beyond the operating band and change cable/load conditions to expose fragile nulls or resonances.
7. Verify operating stress. Measure winding voltage, magnetizing/current waveforms, reset balance, switching overshoot and hot-spot temperature at worst input, load, duty cycle and ambient.
8. Run system EMI tests. Use the applicable LISN, current probe, antennas, cable layout and load configuration. A transformer-only capacitance result does not replace conducted or radiated emissions testing.
9. Keep safety verification separate. Confirm certified construction, creepage, clearance, insulation, component approvals, dielectric tests and touch current under the applicable standard and fault conditions.

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