Inductor coupling is the magnetic linkage between two or more coils, quantified by the coupling coefficient (k). In power electronics and RF design, it is either your most critical design parameter (intentional coupling in SEPIC, Cuk, or flyback topologies where k > 0.95) or a silent layout killer (parasitic crosstalk between adjacent PCB components causing sub-harmonic oscillation). Understanding the physical reality of magnetic flux lines dictates whether your switching regulator runs cleanly or burns up a MOSFET on the bench.
The Workbench War Story: Parasitic Coupling in a 12V SEPIC
Every power electronics engineer has a war story about magnetic flux. Mine involves a 12V-to-5V SEPIC (Single-Ended Primary-Inductor Converter) that refused to regulate cleanly. A SEPIC topology requires two inductors. While application notes often show them as separate components, they strongly recommend using a single coupled core to reduce ripple and save board space.
The Setup: I needed two 10µH inductors rated for at least 4A. I selected two separate Würth 74477410 shielded drum-core SMD inductors. Because board space was tight, I placed them parallel to each other, just 8mm apart, with their winding axes aligned.
The Numbers: The simulated output ripple was 20mV. On the bench, my oscilloscope showed a massive 250mV ripple with erratic jitter. Worse, the main switching MOSFET was running 25°C hotter than the thermal model predicted.
The Outcome & What Went Wrong: I had created an accidental transformer. Even 'shielded' inductors leak magnetic flux. By placing the two inductors parallel and close together, I established a parasitic mutual inductance with a coupling coefficient of roughly k ≈ 0.12. Every time the primary inductor switched, it injected a coupled voltage spike directly into the secondary inductor. This noise coupled into the feedback node, tricking the PWM controller into skewing the duty cycle, resulting in sub-harmonic oscillation and massive switching losses in the MOSFET.
The Fix: I didn't need a new capacitor or a different controller. I simply desoldered the second inductor and rotated it 90 degrees. Orthogonal placement forces the magnetic flux lines to intersect at right angles, dropping the parasitic coupling coefficient to near zero. The ripple dropped to 28mV, and the MOSFET temperatures normalized.
Intentional vs. Parasitic: Which Inductor Type for Which Job?
Choosing the right physical construction determines how much flux escapes the core and how susceptible the component is to external fields. Here is how to select the right magnetics based on your coupling needs.
| Type | Construction & Flux Path | Typical Tolerance | Tempco (ppm/°C) | Typical Use Case | Coupling Profile |
|---|---|---|---|---|---|
| Unshielded Drum | Open bobbin core; flux radiates into surrounding air. | ±20% | ~1000 | Low-cost, low-density buck converters. | Extreme Risk: High parasitic crosstalk. |
| Shielded (Molded) | Powdered iron encapsulated in epoxy; mostly closed flux path. | ±20% | ~300 | High-density PMIC layouts, multi-phase buck. | Low Risk: Minimal leakage, but not zero. |
| Toroidal | Donut-shaped core; flux tightly contained within the ring. | ±10% | Highly stable | Audio crossovers, high-current EMI filtering. | Very Low Risk: Excellent self-shielding. |
| Intentional Coupled | Dual windings on a single shared core (e.g., Coilcraft MSD series). | ±10% (Matched) | Matched | SEPIC, Cuk, flyback, and Zeta converters. | Intentional: Designed for k > 0.95. |
For standard DC-DC buck or boost converters, always default to shielded molded inductors if your layout places components within 10mm of each other. If you are designing a SEPIC or Cuk converter, do not use two separate inductors unless you have the board space to isolate them; use a dedicated coupled inductor to leverage the mutual inductance for ripple cancellation.
Decoding the Markings and the Dot Convention
When you pull an SMD inductor from a bin, the top marking tells you the inductance value, but it does not tell you the polarity. For standard single inductors, polarity doesn't matter. For coupled inductors, it is the difference between a working circuit and a destroyed catch diode.
Reading SMD Inductance Codes
Most SMD power inductors use a 3-digit EIA-style code, measured in microhenries (µH):
- 100 = 10 × 100 = 10µH
- 471 = 47 × 101 = 470µH
- 2R2 = The 'R' acts as a decimal point = 2.2µH
The Dot Convention for Coupled Magnetics
According to standard mutual inductance theory, the physical 'dot' printed on the silk screen or the component body indicates phase relationship. Current entering the dotted terminal of the primary coil induces a positive voltage at the dotted terminal of the secondary coil.
If you wire a SEPIC converter with the coupled inductor dots out of phase (e.g., dot on the switch node for L1, but dot on the ground side for L2), the coupled voltage will oppose the switching action. The controller will max out the PWM duty cycle trying to compensate, inevitably causing the catch diode to avalanche and short circuit. Always verify the dot orientation against the silicon vendor's reference schematic before applying power.
Substituting Coupled Inductors Safely on the Bench
Supply chain shortages happen. If your BOM calls for a 1:1 coupled inductor (e.g., 10µH + 10µH on a single core) and you only have single SMD inductors in your lab stock, you can substitute them, but you must follow strict physical and electrical rules to avoid the parasitic coupling trap mentioned earlier.
- Verify Saturation Current ($I_{SAT}$): Ensure both substitute inductors have an $I_{SAT}$ rating higher than the peak switch current of your design. In a coupled core, the flux cancels out, allowing for a smaller physical size. Two singles will not have this ripple-cancellation benefit, so they must be rated for the full peak current without saturating.
- Match the DCR: Measure the DC Resistance (DCR) of both inductors with a multimeter. True coupled inductors have perfectly matched windings. If your two substitute singles have a DCR mismatch greater than 15%, the lower-DCR inductor will hog the DC bias current and overheat prematurely.
- Enforce Orthogonal Placement: Never place the two substitute inductors parallel to each other. Place them at a strict 90-degree orthogonal angle to minimize mutual inductance.
- Maintain Clearance: Keep a minimum of 10mm clearance between the two inductors, and route high-impedance feedback traces away from the air gap between them.
Failure Modes and Visual Symptoms
Inductors rarely fail open like a fuse; they usually fail due to thermal or magnetic abuse. Recognizing the physical symptoms on the bench saves hours of oscilloscope debugging.
Core Saturation
The Physics: When the current exceeds the core's saturation limit ($I_{SAT}$), the magnetic domains in the ferrite or powdered iron align completely. The core permeability drops to near that of air, and the inductance collapses. The inductor effectively becomes a low-value resistor, causing massive current spikes.
Visual Symptoms: Initially, there is zero physical damage. The component looks pristine. However, a thermal camera will show a massive hotspot (often 40°C to 60°C above ambient) on the inductor body. Over repeated thermal cycles, the potting compound may exhibit micro-cracks or a yellowed, baked appearance due to the extreme localized heat from the $I^2R$ losses during saturation.
Inter-Winding Short
The Physics: The thin enamel insulation on the copper windings breaks down due to voltage spikes, mechanical stress during manufacturing, or prolonged overheating. Adjacent turns short together, reducing the total number of active turns and drastically lowering the inductance.
Visual Symptoms: The top epoxy surface may look slightly melted, bubbled, or discolored brown. If you bring the component close to your nose, it will emit a distinct, sharp smell of burnt phenolic resin and caramelized copper enamel. An LCR meter (measured out-of-circuit) will show an inductance value significantly lower than the marking, and a multimeter will read a DCR much lower than the datasheet specifies.
