A MOSFET datasheet quotes an RDS(on) at 25 °C, and the part on your board runs at 100 °C — where its on-resistance has climbed to nearly double and the conduction loss has grown by the square of any current increase. The temperature that raises its own resistance is the central reality of MOSFET thermal design: the hotter the junction, the higher the RDS(on), the more heat the conduction loss produces, and the loop runs until the junction reaches a balance or the part cooks. This article walks that loop, the loss math, the thermal-resistance chain, and the SOA de-rating for sustained load, ending with a worked 24 V switch worksheet.
The Temperature That Raises Its Own Resistance: RDS(on) vs Junction
The RDS(on) printed on a MOSFET datasheet is measured at 25 °C, and it is not the value the part carries in service. On-resistance climbs with junction temperature because the channel mobility falls as the silicon heats, and the datasheet shows this as either a multiplier curve or a normalized RDS(on)-versus-temperature plot. The practical multiplier for a silicon MOSFET is typically 1.4 to 2.0 at the 100–125 °C junction temperatures common in real designs — a 20 mΩ part at 25 °C can read 30–40 mΩ hot.
The consequence is that every loss calculation done with the 25 °C RDS(on) understates the real loss by the same factor, and the understatement compounds because the loss feeds the temperature. A design that sizes the heatsink with the cold resistance can come up short by a large margin. The low RDS(on) analysis develops the trade of chasing a low cold resistance; this article’s point is that the hot value is the one that decides the design.
The first habit is therefore to read the RDS(on) at the worst junction temperature, not at the 25 °C headline — every later step in this article depends on that single correction.
Loss Formula: Why Squaring the Current Hurts Twice
The conduction loss of a MOSFET is the on-resistance times the current squared: P = RDS(on) × I². The square is the first reason high current hurts — doubling the current quadruples the conduction loss at a fixed resistance. The thermal loop is the second: the loss heats the junction, the heat raises RDS(on), and the higher resistance raises the loss again at the same current. The two effects compound, which is why a MOSFET can exhibit a thermal runaway-like climb under a heavy sustained load.
The compounding is the “loss quadruples” story in the title: the current-squared term alone quadruples the loss when current doubles, and the temperature-driven resistance rise adds a further multiplier on top. A 20 A design at 8 mΩ dissipates 3.2 W; the same design at a hot 16 mΩ dissipates 5.1 W — a 60% jump purely from the temperature effect, before any current change.
The design response is to close the loop at the hot operating point: compute the loss at the hot RDS(on), feed it through the thermal chain, and confirm the resulting junction temperature matches the hot value assumed. That confirmation is the convergence check that tells a designer the loop is closed rather than merely estimated, and it is the step that separates a thermal budget from a thermal guess. The loss worksheet method is the rectifier analog of the same two-way math, and it is the pattern this article applies to the MOSFET.
Building the Thermal Loop: Rth Junction-to-Ambient
The thermal chain that closes the loop is the same resistance network used for any power device: the junction temperature equals the ambient plus the dissipation times the total thermal resistance from junction to ambient. The total Rth is the sum of the junction-to-case, case-to-heatsink, and heatsink-to-ambient terms, and each term is set by a different physical element.
The junction-to-case term is the package’s intrinsic ability to move heat from the die to the case; the case-to-heatsink term is set by the mounting, the thermal interface material, and the pressure; the heatsink-to-ambient term is set by the heatsink’s size and airflow. The weakest link in the chain dominates, and a MOSFET with a superb junction-to-case but a poor interface can run nearly as hot as a part with a worse package but a clean mount. The rectifier thermal design guide and the D2PAK thermal article develop the same chain for packages this MOSFET analysis shares.
The loop is iterative because the dissipation depends on the temperature: assume a junction temperature, compute the hot RDS(on), get the loss, run it through the chain, and check whether the resulting junction matches the assumption. A few iterations converge to the real operating point.
The thermal chain belongs in a table because each term maps to a distinct design action:
| Thermal term | Set by | Design action |
|---|---|---|
| Junction-to-case Rth(j-c) | Package and die | Choose the right package for the heat |
| Case-to-heatsink Rth(c-s) | Interface material, pressure | Clean surface, correct pad, proper torque |
| Heatsink-to-ambient Rth(s-a) | Heatsink size, airflow | Size the sink, add airflow |
| Ambient temperature | Enclosure and climate | Set the operating envelope |
The table makes the weakest-link principle visible: the total is the sum, and the largest term dominates. A design that pours effort into a low junction-to-case package while skimping on the interface or the heatsink gets a hot part anyway, because the big term in the chain decides the outcome. Reading the table before buying the heatsink is what turns the thermal design from a guess into a budget with a known dominant term.
The second iteration of the loop deserves its own note because it is where the real operating point appears. The first pass in the worksheet assumed a 125 °C junction and landed at 65 °C; the second pass, using 70 °C, converged near 64 °C. The convergence is fast because the RDS(on) multiplier changes slowly with temperature in the mid range, so two or three iterations suffice in practice, and the worksheet in the next-to-last section runs the full sequence. A design that skips the iteration and stops at the first cold pass has not found the operating point — it has found an optimistic bound that the hot-resistance reality will not honor.
Derating the SOA for Sustained Load
The SOA for a sustained load is the DC line, not the pulse lines — a continuous current the part carries indefinitely, bounded by the thermal box rather than the junction’s transient heat capacity. The de-rating for a sustained load is therefore the thermal de-rating: the DC current the part can carry falls as the ambient rises, because the junction has less temperature headroom.
The de-rating read combines the SOA DC line with the thermal chain. At a high ambient, the junction starts closer to its limit, so the acceptable DC current drops and the acceptable dissipation shrinks. A design that used the 25 °C DC current and the 25 °C RDS(on) is doubly over-optimistic: it assumes a cooler junction than the part will reach and a lower resistance than it will carry. The SOA reading article frames the same boundary; the sustained-load point here is that the DC line, corrected for the real ambient, is the number that matters for a continuously running switch.
A 24 V MOSFET Switch Thermal Worksheet
A worked worksheet ties the whole loop together. A 24 V load-switch carries 15 A continuously in a 55 °C ambient enclosure, using a MOSFET with RDS(on) 8 mΩ at 25 °C and a thermal multiplier of about 1.6 at 125 °C. The junction-to-case Rth is 1.5 °C/W, the interface and heatsink add another 2 °C/W, for a total of 3.5 °C/W to ambient.
Start by assuming a junction temperature of 125 °C: the hot RDS(on) is 8 mΩ × 1.6 = 12.8 mΩ, and the conduction loss is 12.8 mΩ × 15² = 2.9 W. Through the 3.5 °C/W chain, the junction rises 2.9 × 3.5 ≈ 10 °C above the 55 °C ambient, giving a junction of about 65 °C — far cooler than the 125 °C assumed. Re-run with 70 °C: the RDS(on) multiplier is about 1.25, the loss is 2.5 W, the rise is 8.8 °C, and the junction lands near 64 °C. The loop converges quickly and stably, and the design is comfortably within the limit — a 15 A switch on a modest heatsink works. Now raise the duty scenario to a 40 A inrush with a real duty, and the same worksheet shows where the margin tightens. The power MOSFET selection guide and the LV/MV MOSFET families supply the parts whose hot resistance and thermal numbers feed the worksheet, and the SiC MOSFET families cover the high-temperature extension where the same loop applies. The closing habit is to run the worksheet at the worst ambient and the hot resistance, not at 25 °C, because the whole thermal reality of a MOSFET is that the numbers move together — and a design that computes them together at the operating point is the one that stays cool for the life of the product.