Key Takeaways
- Three tank values, the series inductance, the series capacitance and the magnetizing inductance, settle efficiency, gain range and soft switching margin before any part is ordered.
- Soft switching margin at light load and minimum input tells you more about a design than a headline efficiency figure measured at nominal conditions.
- Editable component models let you sweep tank parameters in minutes and read the switching detail sitting behind every efficiency number.
An LLC resonant converter earns its efficiency inside a narrow band of operating conditions, and the design work that matters is deciding where that band sits.
Three passive values do most of the deciding. The series inductance, the series capacitance and the transformer magnetizing inductance set the resonant frequency, the gain curve and the range over which soft switching holds. Get them right and the topology hands you low switching loss across most of the load range.
Data centres used 415 TWh of electricity in 2024, about 1.5% of global electricity consumption, and the isolated DC-DC stage inside those server supplies is almost always an LLC. An efficiency figure won’t tell you why one tank design beats another. You need the switching detail behind that figure, so the model you design in has to be one you can open and change.
How an LLC resonant converter moves power efficiently
An LLC resonant converter feeds a square wave into a tank made of a series inductor, a series capacitor and the transformer magnetizing inductance. The tank filters that square wave into a near sinusoidal current, so the primary switches change state while current and voltage sit far apart in phase. Low turn-on loss comes out of that phase relationship.
Picture the 400 V bus to 48 V stage in a server supply. A hard switched bridge pays a turn-on penalty on every edge, and that penalty grows with frequency. The LLC stage moves the same power at 200 kHz with its switches turning on near zero volts.
The tank supplies that current phase across a limited frequency and load range, so a design that looks excellent at nominal input can fall out of soft switching when the bus sags or the load drops to 10%. You can’t have every condition at once, so choosing the tank means choosing which ones win.
Sizing the resonant tank around your operating point
Tank sizing comes down to three numbers. The resonant frequency, set by the series inductance Lr and the resonant capacitor Cr, fixes where the stage runs most efficiently. The inductance ratio Ln, which is Lm divided by Lr, sets your gain range. The quality factor Q ties tank impedance to the reflected load.
Take a 1 kW stage running from a 390 V bus with a 48 V output. Setting the resonant frequency at 100 kHz with Ln near 5 and Q around 0.4 holds soft switching from full load down to about a quarter load. Push Q higher to shrink the magnetics and peak gain drops, so the stage can’t hold the output through a bus dip. Those numbers pull against each other, which is why tank sizing rewards sweeping.
| Tank parameter | What it controls, and what mis-sizing costs you |
| Series resonant inductance Lr | Fixes the resonant frequency with Cr, and a small value leaves no gain range for a sagging input |
| Resonant capacitor Cr | Carries the full tank current, so an undersized part heats and drifts the resonant frequency |
| Inductance ratio Ln | Sets the gain reached before the peak, so a high ratio leaves no reserve at low input |
| Quality factor Q at full load | Links reflected load to tank impedance, so a high Q pushes the worst case toward the capacitive region |
| Magnetizing inductance Lm | Supplies the current that discharges switch capacitance in the dead time, so a large value costs light load soft switching |
Reading the gain curve before committing to hardware
The gain curve plots voltage gain against normalized switching frequency for a family of load conditions. It tells you where the stage sits at nominal input, how far the frequency must move to hold regulation at the extremes, and how close the worst case point comes to the peak where soft switching stops.
Battery chargers make the point sharply. An onboard charger holds regulation from a depleted pack near 250 V to a full pack above 400 V, a span a fixed output supply never sees. That workload is no longer a niche, since electric cars were 25% of new car sales globally in 2025, with sales above 20 million.
Watch the region left of the peak. Soft switching holds while the operating point sits above the peak gain frequency. Cross it and the tank turns capacitive, the body diodes take a hard reverse recovery hit, and switches fail within seconds. Margin there is worth more than a better nominal efficiency.
Where the soft switching window opens and closes
Soft switching in an LLC stage runs on magnetizing current. During the dead time between primary switch transitions, that current has to discharge the output capacitance of the switch about to turn on. The window stays open while there’s enough magnetizing current and enough dead time to finish the job.
Light load closes it first. At 5% load the magnetizing current is unchanged, since it doesn’t scale with output power, but the switching frequency has climbed well above resonance and the dead time shrinks with the period. Plenty of designs that hold zero voltage turn-on at full load lose it between 10% and 20%, and the symptom is a hot high side switch with a step in the drain waveform. The secondary has its own version, since rectifier turn-off stays lossless only while the stage runs at or below resonance.
“The window stays open while there’s enough magnetizing current and enough dead time to finish the job.”
Transformer and magnetizing inductance set the practical limits
The transformer does three jobs at once here. It provides isolation and the turns ratio that fixes nominal gain, its magnetizing inductance forms one leg of the tank, and its leakage inductance often supplies part or all of Lr. Those jobs conflict, and the magnetics design settles the conflict.
Designers lean on that overlap. Separated windings and a deliberate air gap push leakage up to the value Lr needs, which removes a discrete inductor from the bill of materials. The saving is genuine, though leakage now depends on build tolerance, so a 20% spread across a production run drags the gain curve with it.
Magnetizing inductance carries the sharpest tradeoff. Cut it in half and you double the circulating current, adding conduction loss at every operating point, including full load where you least want it. Double it and you halve the current available to discharge switch capacitance, so the light load window closes earlier. The right value falls out of the dead time you can afford and the switch capacitance you’re using.
Simulating an LLC stage with editable component models
Simulating an LLC stage means assembling the half bridge, the tank, the transformer and the rectifier from component blocks, then running switching level waveforms at a time step small enough to resolve the dead time. A model built that way shows what the switches, the capacitor and the windings each experience rather than only the output.
An LLC stage assembled from SPS component blocks inside Simulink gives you that view directly. The resonant elements and the transformer come from standard libraries, the solver and time step are yours to set, and every value sits in a dialog a sweep can reach. Nothing hides in a compiled block, so a sweep across Ln, Q and dead time finishes while you make a coffee.
That access turns an efficiency number into something you can act on. A bench measurement tells you the stage lost 18 W and nothing about where. Waveform level simulation lets you attribute the loss, and a few quantities carry most of the answer.
- Primary switch voltage through the dead time, which shows if zero voltage turn-on completes
- Tank current at the turn-off instant, which sets the loss you pay on every edge
- Resonant capacitor voltage stress across the full input and load range
- Magnetizing current amplitude, which is the budget your soft switching runs on
- Secondary current shape, which shows how cleanly the rectifier stops conducting
“That’s the difference between a converter you tuned and a converter you understand.”
What disciplined LLC design looks like in practice
Good LLC design is mostly bookkeeping. You choose an operating point, write down what the tank does at every corner of the input and load range, and check the soft switching margin before anyone orders parts. The topology rewards that patience and punishes bench tuning.
Engineers who get consistent results treat the model as the place where design decisions actually happen. A sweep across Ln, Q and dead time that once meant a week of prototypes now finishes in minutes, and it returns a map of soft switching margin against operating point rather than one efficiency figure. Because the SPS component models are open and editable, you can read tank current at turn-off in the same run that produced that figure.
None of this removes the bench. It changes what you bring to it. A prototype built after a swept design confirms numbers you already predicted, and the surprises left are small. That’s the difference between a converter you tuned and a converter you understand.
