Inductors and Beads
Inductors and Ferrite Beads for Power and EMI
An inductor stores energy in a magnetic field and a ferrite bead turns high-frequency noise into heat. The two sit close on a schematic and look alike on a board, then split on what they do and the numbers they are read on. A power inductor is chosen on its saturation current, its DC resistance, and its inductance value, since it has to carry the converter’s current without saturating and without wasting it as heat. A ferrite bead is chosen on its impedance at a stated frequency and its rated current, since it has to block the noise without dropping the rail it sits on. Treating one as the other is the error behind a long list of failed power rails and failed EMI fixes, so the choice starts with knowing which part the job in front of you calls for, a power part or a noise part. Both are passives wound from copper on a magnetic core, and that shared look is why the two get confused on a parts list, where the inductor is built to store and release energy cleanly and the bead is built to lose it on purpose in a chosen band. A board carries both, a few power inductors on its converter rails and a scatter of beads on its supply and signal lines, and reading each on its own numbers is what keeps a rail regulated and a spectrum clean.
Power inductors and saturation
A power inductor sits in a switching converter and carries a current that ramps up and down every cycle, so the first number that matters is the saturation current, the point where the core stops storing more field and the inductance collapses. Below saturation the part behaves as its rated value, and above it the inductance falls away and the current spikes, which a converter reads as a fault or lets run until something fails. The inductance value itself sets how much the current ripples each cycle, since the volt-seconds the converter applies drive a current ramp inversely proportional to the inductance, so a larger value gives less ripple and a smaller one more, and the value is chosen with the switching frequency to land the ripple where the design wants it. Too small a value ripples hard and stresses the part and the capacitors around it, and too large a value costs size and DC resistance, so the value is a balance struck before the current ratings even come into the choice.
The trouble from getting this wrong is specific. Knowing why undersizing saturation current causes trouble starts with the peak current the inductor sees, which is the average load current plus half the ripple, and that peak has to sit below the part’s saturation rating with margin. A part chosen on the average current alone saturates on the ripple peaks, its inductance drops, the ripple grows, and the converter heats and grows unstable, so the saturation current is read against the peak current the part sees, with the average left to the heating check. The datasheet curve of inductance against current shows where it rolls off, and a soft-saturating core that rolls off gently is read differently from a hard one that gives way abruptly, so the part is chosen with its entire curve in view.
The other steady cost is resistance. Understanding how a small DCR still costs efficiency is the current squared times the DC resistance of the winding, a loss that runs all the time the converter delivers current, so even a few tens of milliohms turns into a real fraction of a watt on a high-current rail. A lower DCR means a thicker winding and a larger part, so the choice trades size against efficiency, and a design reads the DCR against the current it carries and the heat it can lose. The DCR also sets the temperature rise, since the winding heats from its own loss, so a part rated for a current on saturation may still run hot on resistance, and the two ratings are read together. The winding loss is not the whole of it, since the core itself gives up energy each cycle to hysteresis and eddy currents, a loss that climbs with the switching frequency and the ripple, so a part for a fast converter is read on its core loss alongside its DCR. The total of the copper loss and the core loss is what heats the part and what the efficiency pays, and a datasheet that lists only the DCR tells half the story on a high-frequency rail.
There are two current ratings on every power inductor, and they answer different questions that a design has to keep apart. The saturation current is the magnetic limit, the point where the core can hold no more field and the inductance starts to collapse, and the RMS or heating current is the thermal limit, the point where the winding’s own loss raises the part to its temperature ceiling. A part carries both, and a design checks its peak current against the saturation rating and its average current against the heating rating, since the two can sit far apart on the same part. A converter with a high ripple and a light average load is held by the saturation number, because the ripple peaks reach the core limit long before the average heats the winding, and a converter with a steady heavy current and little ripple is held by the heating number, because the winding overheats before the core saturates. The binding limit shifts with the application, so the part is read on whichever one it hits first and sized with margin against that one. The core material sets how the saturation behaves besides, since a ferrite core saturates hard, dropping its inductance sharply once the field is full, and a powdered-iron or a composite core saturates softly, rolling its inductance down over a range of current. A hard-saturating part is sized with generous margin so the design never drives it into saturation, and a soft-saturating part can run closer to its rated saturation since the rolloff is gradual and forgiving, which is part of why a converter that sees occasional current spikes often prefers the soft core. The datasheet shows this as a curve of inductance against current, and a design reads where the curve has fallen to perhaps eighty or seventy percent of the nominal value and treats that as the usable limit, ahead of the single saturation number the part is labeled with, since the inductance the converter sees at its peak current is what sets the ripple and the stability, and a part run past its real rolloff feeds a ripple larger than the design budgeted and a loop less stable than it modeled. The margin left on saturation is the cheapest protection on the rail, since a part one size up in current costs a little board area and a converter that saturates in the field costs a recall, so the margin is set generously and checked at the highest load and the lowest inductance the part is specified to hold.
Shielding is the third axis. Weighing the EMI gain and the cost of shielded inductors is that a shielded part keeps its magnetic field to itself, so it radiates less into nearby traces and sensitive circuits, at a higher price and sometimes a higher DCR than an unshielded one. A dense board near an antenna or an analog front end reaches for the shielded part, and a roomy, noise-tolerant layout saves the money with an unshielded one, so the shielding is matched to the EMI the design has to meet.
An inductor also stops behaving as an inductor above its self-resonant frequency, where the winding’s own capacitance takes over and the impedance falls away, so a part used near that frequency behaves as a capacitor and stops filtering. A power inductor on a switching rail runs well below its resonance and ignores it, and a part used to filter at high frequency is read on its resonance the way a bead is read on its curve. This is part of where the inductor ends and the bead begins, since above a few tens of megahertz a wound inductor has lost its impedance to its own capacitance, and a ferrite bead, built for exactly that band, takes over the job of stopping the noise.
Power inductors come in a few constructions, and the type suits different rails. A wirewound part on a drum or a toroid core gives a low DCR and a high current for its size and is the default on a power rail, a shielded version adding a sleeve or a magnetic coating that holds its field in. A molded part buries the winding in a composite core, which gives a soft saturation and a rugged, low-radiation body that suits a dense board near sensitive circuits. A multilayer or a thin-film part packs a small value into a tiny case for a light load or a high-frequency point-of-load, trading current for size. The construction sets the saturation behavior and the DCR, along with the height and the radiation, so a design reads the type alongside the value and the current, since a flat composite part and a tall wirewound part at the same inductance bring their own DCR and radiated field to the board. The core material runs underneath all of it, ferrite for the high-frequency low-loss rails and powdered iron or a composite for the high-current soft-saturating ones, so the type and the material are read together with the electrical numbers.
The inductor vendors
The premium end of the power-inductor market runs on a few makers known for tight specs and reliable data. Reaching for Coilcraft LPS and XAL in premium power designs buys a part whose saturation and DCR are characterized carefully and whose models match the bench, which a tight, high-performance converter pays for. The availability of low-profile high-current Vishay IHLP covers the parts a dense, high-current board reaches for, a composite-core family that takes a hard current in a flat package, broad enough in stock to design around. The two anchor the high end each on its own footing, the one on characterized data and tight tolerance and the other on raw current density in a flat case, and a design picks by whether it needs the modeling fidelity or the current-in-height, since a point-of-load beside a processor values the low profile and a precision supply values the data. Both publish the saturation and the heating curves a careful design leans on, which is part of what the premium buys over a part that gives a single headline number and leaves the rest to be measured.
The breadth of a line decides as many designs as its peak specs. Knowing when to choose Würth WE-LHMI and WE-HCI covers a wide catalog with free samples and solid data, which speeds a design that wants one supplier across many rails, and Sumida CDRH applications cover a long-established shielded drum family that fits the ordinary buck-converter rail at a sensible price. The mid-market makers compete on breadth and availability ahead of the last point of performance, so a design that needs a workable part across a dozen rails without qualifying each one reaches for a wide catalog with good stock and clear data. The free samples and the online design tools these makers offer speed the early work, since a part in hand and a model that matches the bench lets a converter rail settle quickly, and the design pays the premium only on the rails that need the fully characterized part.
One job needs the field kept in. The Bourns SRR shielded inductors and EMI cover the semi-shielded and shielded drum parts a design drops onto a rail that has to pass radiated-emissions limits, and the current uses of J.W. Miller axial inductors cover the leaded parts that still fit a through-hole filter, a high-current choke, or a repair where a surface-mount part will not do.
The lower-cost end is real and repays a careful read. Weighing Chinese inductors against Coilcraft is how far a cost-driven part has closed the gap on saturation, DCR, and data quality, which for many ordinary rails is far enough, so a design reads the curves and the lot consistency before it commits to the cheaper part on a rail that matters. The gap is narrowest on the ordinary buck and boost rails, where a cost-driven part holds its saturation and DCR well enough and the lot-to-lot spread is tight enough to design around, and it stays widest on the demanding rails where the characterized data and the tight tolerance earn their price. A design splits its bill of materials along that line, putting the cost part on the forgiving rails and the premium part where the margin is thin, which captures the saving without betting the hard rails on it.
Ferrite beads
A ferrite bead is a frequency-dependent resistor, low impedance at DC and a peak of resistance at high frequency where it turns noise into heat, so it passes the supply current and blocks the noise riding on it. The bead is rated by its impedance at a stated frequency, usually a hundred megahertz, and by the DC current it carries before it saturates and loses that impedance. The way it works is the key to using it: at low frequency the bead is a small inductance and passes the signal, and in its working band the ferrite material turns lossy and the bead becomes a resistor that absorbs the noise and dissipates it as a little heat. A bead is sorted by the band it targets, since a part tuned to suppress noise around a hundred megahertz does little at a gigahertz and the reverse holds too, so the bead is matched to the frequency of the noise the design has to kill. The materials differ by maker and grade, and a high-current bead trades some impedance for the ability to carry more bias, so the choice reads the impedance, the current, and the target band together.
The market leaders set the reference. The noise suppression of Murata BLM ferrite beads covers the broad family a design reaches for by default, sorted by impedance, current, and the frequency band each part targets, and the interchangeability of the TDK MPZ and Murata BLM covers how closely the two cross-reference, which a design leans on for a second source as long as it checks the impedance curve and the current rating, beyond the value alone. The two makers between them set the catalog the rest of the market cross-references, so a bead specified as a BLM or an MPZ part is one a buyer can source widely, and the cross-reference holds as long as the impedance band and the bias rating match, not the nominal impedance alone. A design that pins a bead to one maker without checking the curve of the proposed equivalent can find the substitute peaks in a different band and lets the noise it was meant to stop pass straight through.
The trap is reading the part on one number. Understanding the limits of selecting a bead on its impedance curve starts with the fact that the rated impedance is measured at a stated current and falls as the DC bias climbs, so a bead carrying real current delivers far less impedance than its headline figure. The impedance is also a mix of resistance and reactance, and only the resistive part damps noise, so a bead used below its resistive band stores the energy and fails to dissipate it, and can ring with a nearby capacitor, raising a peak where the design wanted a notch. A bead chosen on its hundred-megahertz number alone, without reading the bias derating and the resistive band, often makes the noise worse, so the part is read on its full curve at its real operating current.
The everyday bead job has its own logic. Knowing why a bead sits next to a digital power rail is about keeping the switching noise a digital chip makes from traveling back down the supply into the analog or the RF section that shares it, so the bead isolates the noisy local rail from the clean one. The bead is paired with the decoupling capacitors on each side, and the design watches the LC resonance that the bead and those capacitors form, damping it where it would peak in band.
A bead is the wrong part on a rail that draws heavy current, since it loses its impedance under bias and adds resistance the rail never wanted, the error behind a great many EMI fixes that cost a rail its regulation.
The bead also forms a filter with the capacitors on either side, and that filter has a resonance the design has to place with care. A bead and a clean ceramic capacitor make a high-Q LC circuit, and if its resonance lands on a frequency the noise carries, the filter peaks there and amplifies the same noise it was placed to stop, so a design either picks a lossy bead whose resistance damps the resonance or adds a small damping resistor to flatten it. This is why a bead dropped onto a rail on instinct sometimes raises an emissions peak that was not there before, and why the bead, the capacitors, and the damping are designed together as one filter, with no part dropped onto the rail on its own. The DC resistance of the bead, small as it is, drops a voltage at the load current and dissipates a power, so a bead on a rail pulling an amp can drop tens of millivolts and warm itself, which a sensitive low-voltage rail cannot spare. The right reach on a heavy rail is a proper inductor or a filter built for the current, and the bead is kept for the signal lines and the light local supplies where its loss of impedance under bias does not bite.
Common-mode and differential chokes
Some noise needs a choke rather than a bead, and the kind of choke turns on the kind of noise. Choosing between common-mode and differential-mode chokes starts with where the noise flows: a common-mode choke wraps both lines on one core so it ignores the signal current that flows out one line and back the other, and presents a high impedance only to the common-mode noise that flows the same way on both, which is the noise a cable radiates. A differential-mode choke is a plain inductor in one line that blocks the noise riding along the signal itself.
The choke is read on the same numbers as any inductor, with its job added, so a common-mode choke is chosen for its common-mode impedance across the noisy band and its current rating, and it is placed where the cable meets the board to catch the noise before it leaves. A data line uses a choke wound to pass its differential signal cleanly and block the common-mode noise, so the part is matched to the signal it has to leave alone as carefully as the noise it has to stop. A common-mode choke on a USB or an Ethernet line is specified to pass the data rate without distorting it, which sets how tightly the two windings are coupled and how much leakage inductance the part can have, since the leakage acts on the signal as a differential inductance and rounds the edges. The current rating matters too, since a choke on a power input carries the full supply current through both windings and saturates if it is undersized, so a power-line common-mode choke is read on its current the way a power inductor is. A differential choke or a plain inductor handles the noise that rides along the signal itself, and a design that knows which mode dominates picks the choke that fits and skips stacking both and hoping. A board that fails its radiated-emissions scan often traces the trouble to a missing or undersized common-mode choke at the cable entry, since a cable is the antenna that carries common-mode current out as radiation, so the choke goes in where the cable meets the board, sized for the band the scan flags and the current the line carries.
How inductors and beads get chosen and sourced
An inductor is read on its saturation current and its DCR, with the inductance value behind them, and a bead on its impedance curve and its rated current, so a substitution that matches the value can still fail if its saturation, its DCR, or its bias derating fall short of what the rail was built around. A broad-line distributor that carries the power inductors and the ferrite beads across the values, the current ratings, and the case sizes lets a design source the premium part where the converter needs it and the cost part where it does not, and find a checked equivalent on the full curve when a part runs short. The saturation current and the DCR are the constants a power rail answers to, and the impedance curve and the bias rating are the constants a bead answers to, so the part is checked on the curve and the rating, past the headline value, and a second source is qualified the same way. The supply picture splits by part: the beads and the jellybean inductors come from several makers and cross-reference cleanly, and the premium power inductors and the custom magnetics carry longer lead times and fewer drop-in equivalents, so a design that commits to a characterized part builds that lead time into its schedule. A part qualified to the automotive grade carries the screening a car program needs and narrows the field of makers, and a design reads the grade and the lead time against the market it ships into the way it reads the electrical numbers.
Read the curve, not the headline
Name the job first: a power inductor stores energy and a bead blocks noise, and the two are never swapped. Size the inductor’s saturation current against the peak current with margin, read its DCR against the efficiency and the heat, and shield it where the EMI calls for it. Read a bead on its impedance at its real bias current and its resistive band, not its headline number, and keep it off a heavy-current rail. Match a choke to the mode of the noise it has to stop, and design the bead, its capacitors, and any damping as one filter. The one habit that heads off the worst trouble is reading the curve, beyond the headline, since the saturation current, the impedance, and the bias derating all live on curves that the one number on the front of the datasheet hides. The fifteen pages below take the inductors, the beads, and the chokes one at a time, each by the number that decides whether it holds up a rail or clears a spectrum, and each read on the curve the datasheet plots beneath that headline figure.