Every MOSFET datasheet leads with RDS(on), and it is tempting to sort a supplier table by that number and take the smallest value. But on-resistance is bought with silicon area, and silicon area is bought with gate charge, switching loss, and cost. This article explains the economics behind low RDS(on), works through the conduction-versus-switching crossover, and gives a practical method for deciding when a lower-RDS(on) part pays for itself and when it is a wasted premium.
Where Conduction Loss Actually Lives
The conduction loss of a MOSFET is I² × RDS(on) × duty, evaluated at the operating junction temperature. Three details separate a real calculation from a headline comparison:
- The temperature multiplier. RDS(on) rises with junction temperature—typically 40–80% from 25 °C to 100–125 °C for low-voltage silicon parts. Comparing parts at 25 °C and operating them at 110 °C changes the ranking.
- The duty cycle. In a buck low-side switch the current flows for most of the cycle; in a flyback primary it flows for a fraction. The same RDS(on) difference means very different watts depending on duty.
- The RMS value, not the average. Switching waveforms carry ripple; the loss depends on the RMS current, which is always higher than the average for pulsed conduction.
A worked example sets the scale: a 20 A RMS current with a 5 mΩ part at 110 °C (say 8 mΩ hot) and a 50% duty produces about 1.6 W of conduction loss. Dropping to 3 mΩ hot saves about 1 W—real, but small next to a switching loss that can change by 3–5 W when gate charge and capacitances grow with die size.
The Die-Size Economics
RDS(on) scales inversely with the active die area: to halve the on-resistance, the die roughly doubles, or the technology must improve. The die-size trade appears in the datasheet as a family curve—the same voltage class, with RDS(on) stepping down and gate charge stepping up as the parts get bigger.
The consequence is a figure of merit: RDS(on) × Qg, which stays roughly constant within a technology generation. A “lower RDS(on)” part from the same family is simply a larger die with more gate charge and more capacitance. The trade is not free; it is moved from conduction loss to drive and switching loss.
The table below shows the typical shape of the trade within one voltage class and technology:
| RDS(on) at 25 °C | Relative die size | Qg (relative) | Conduction loss at 20 A RMS, 50% duty, 110 °C | Switching loss trend |
|---|---|---|---|---|
| 8 mΩ | 1× | 1× | ~2.4 W | Baseline |
| 5 mΩ | ~1.6× | ~1.6× | ~1.5 W | Higher (larger Qg, Coss) |
| 3 mΩ | ~2.6× | ~2.6× | ~0.9 W | Significantly higher |
The numbers are example values for method—exact values come from the datasheet curves—but the shape is universal: each halving of RDS(on) costs a near-proportional increase in gate charge and capacitance, and the switching loss rises with frequency.
The Conduction-Switching Crossover
The decision rule is a crossover calculation. At low frequency and high duty, conduction loss dominates and the bigger die wins. At high frequency, the switching loss grows linearly with frequency while conduction loss stays fixed, so the crossover shifts toward smaller dies.
A practical shortcut: compute the total loss at your frequency for two or three parts from the same family—
- Conduction: I_rms² × RDS(on)(TJ) × duty;
- Switching: (Eon + Eoff) × fsw, scaled from the datasheet test conditions to your voltage and current;
- Gate drive: Qg × Vgs × fsw.
Plot the total versus frequency. The part with the lowest total at your operating point is the economic winner, even if it does not have the lowest RDS(on) on the page. In a 100 kHz buck, the mid-RDS(on) part often wins; in a 20 kHz motor bridge running at high duty, the low-RDS(on) part usually pays.
When Low RDS(on) Clearly Pays
Four situations justify the die-area premium without much analysis:
- Continuous high current. Battery discharge paths, motor bridges, and output rectification stages carry tens of amps for long intervals; every milliohm becomes watts.
- Thermal-constrained enclosures. When the heatsink, airflow, or board area is fixed, lowering conduction loss lowers the junction temperature directly.
- Synchronous rectification. The SR MOSFET’s conduction interval is long and the current high; the loss comparison is covered in the companion article on synchronous rectifier MOSFETs in this series.
- Low switching frequency with high duty. Below a few tens of kilohertz, switching loss is small and conduction dominates.
The opposite cases—high frequency, light load, cost-sensitive production, or a small gate-drive budget—favor a smaller die. The SMPS application section of the Good-Ark site shows how MOSFET families are grouped for supply designs; to review RDS(on)-versus-temperature and gate-charge curves, use the Documents section.
The gate-drive term deserves explicit mention: the driver must supply Qg × fsw of average power, and a larger die demands more. In a design with a small gate-driver budget—a logic-level drive or a small controller pin—the practical switching frequency falls as Qg rises, and the low-RDS(on) part may simply be undrivable at the target frequency. This is why the crossover calculation includes the gate-drive term and why a mid-size die often wins in frequency-sensitive designs even when the conduction loss alone favors the bigger part.
A Die-Size Economics Worksheet
- Define the operating point: I_rms, duty, junction temperature, and switching frequency.
- Pull RDS(on) at that temperature from the datasheet curve, not the 25 °C table.
- Estimate switching energy at your voltage and current from the Eon/Eoff curves.
- Compute total loss for the smallest, mid, and largest die in the family.
- Add the gate-drive power and the thermal result (junction rise = total loss × Rth).
- Compare cost only after the loss ranking—a cheaper die that needs a bigger heatsink is not cheaper.
Reading the RDS(on) Specification
The headline RDS(on) number is meaningless without its test conditions. The datasheet specifies the value at a defined gate voltage (10 V, 4.5 V, or a logic-level voltage), a defined drain current, a defined junction temperature (usually 25 °C, with a normalized curve for higher temperatures), and a defined measurement method (pulsed versus DC). A part that guarantees 5 mΩ at 10 V may be 6.5 mΩ at 4.5 V and 8 mΩ at 100 °C—all legitimate, all different. The comparison table must normalize these conditions before the ranking means anything.
The datasheet also distinguishes typical from maximum values; the maximum is the guarantee the design should use, and the tolerance can be wide enough to change a loss ranking between two parts. Confirm that the supplier’s specified test conditions match the circuit’s gate voltage and current before treating the number as the design value.
The cost case completes the picture. A low-RDS(on) part costs more, but it can shrink the heatsink, the copper area, the fan, or the enclosure. A worked comparison: if a 0.5 mΩ reduction saves 1.5 W of conduction loss at the operating point, and the thermal system costs roughly a dollar per watt removed, the device premium is justified up to that value. This is the same crossover calculation as the loss ranking, done in currency rather than watts.
The worksheet method above is best implemented as a shared selection spreadsheet: one sheet per candidate, with the operating point, the RDS(on)-temperature curve, the switching-energy curves, and the thermal and cost inputs, so the ranking updates when any input changes. The spreadsheet is the deliverable the design team keeps; a FAE-provided template with the Good-Ark families pre-loaded is the practical starting point.
Paralleling, briefly. When one die cannot carry the current, paralleled parts share the on-state current well because RDS(on) rises with temperature, but the switching transient is shared by the layout, not by the resistance. Symmetric gate and source paths and individual gate resistors are the rules that make paralleling work; the full treatment belongs in a dedicated article rather than this one.
Frequently Asked Questions
Why does RDS(on) increase with temperature? The channel resistance of a silicon MOSFET rises as lattice scattering increases with temperature. The datasheet’s normalized RDS(on) curve shows the multiplier at your operating junction temperature.
Is the lowest RDS(on) MOSFET always the best for high current? Not necessarily. The larger die also carries more gate charge and output capacitance, raising switching loss. The best part is the one with the lowest total loss at your frequency, duty, and temperature.
Does low RDS(on) help at light load? Little. At light load the conduction loss is tiny, while the gate-drive and switching losses remain; a smaller die is often more efficient there. Some designs switch between dies or modes to cover both.
How do I measure RDS(on) accurately? Measure the drain-source voltage and the current simultaneously with a Kelvin-connected sense (separate force and sense leads on the source), because the shared return path adds voltage error that looks like resistance. Use a pulsed measurement to avoid self-heating shifting the value, and record the junction temperature.
The Crossover, Not the Headline
Low RDS(on) is a good thing when conduction loss dominates the budget, and a costly distraction when it does not. Run the total-loss calculation at your operating point—hot RDS(on), duty, frequency, and gate-drive power—and let the crossover decide. The part that wins the loss ranking at your frequency is the part that pays for itself; the lowest number on the page is only the starting point.
FAE note before publication: add the RDS(on)-versus-temperature and switching-energy curve excerpts for two candidate parts of the same voltage class and package, with the test conditions and the technical reviewer’s name.