NCE40H29D vs STB270N4F3 vs STB170NF04: Full Comparison

Read the two current rows together, because that is where the three parts separate. The NCE sheet's 290A falls to 205A at a 100°C case; both ST sheets print the same number at both temperatures and say why in a footnote.

From what we see across Shenzhen reorders (2025–2026), the D2PAK class gets substituted by package and by on-resistance far more often than by current label, and the parts that cause returns are the ones chosen on the label.

When to Use the STB270N4F3

Choose it when the board is already D2PAK and the design needs the lowest on-resistance in this footprint. At 2.0mΩ maximum tested at 80A, it is the only part here that is lower than the NCE's 2.4mΩ.

Its 160A rating does not derate with case temperature, because the sheet states the current is limited by the package rather than by the die. At a 100°C case the NCE part is rated 205A against this one's 160A.

The 47pF reverse transfer capacitance is the other headline. It is less than a twentieth of the NCE figure, which changes what a given gate drive has to fight during a switching edge.

When to Use the STB170NF04

Choose it when avalanche margin or a published board thermal figure is what you are missing. It prints 1.5J of single-pulse avalanche energy at 40A and a 35°C/W junction-to-PCB path — neither of which appears on the NCE sheet.

Its 8 V/ns peak diode recovery rating is more than twice the other ST part's, which matters in bridge circuits where the body diode recovers into a live rail.

It is the lowest-current option here at 80A, so it makes sense as a second source for a mid-current design rather than as a direct swap at the top of the range.

When Not to Use This Class at All

Do not substitute any of these on current label alone. The three sheets range from 80A to 290A in the same footprint, and the current row is the one figure the industry guidance explicitly warns against using for cross-reference.

Two conditions disqualify the whole class. A gate rail that cannot reach 10V rules out all three, since none of them publishes a 4.5V resistance row. And a junction temperature budget below 175°C removes the margin all three assume.

Pinout & Layout Notes

All three parts share one footprint and one pin assignment: pin 1 gate, pin 2 drain through the tab, pin 3 source. A swap between them is a tape-and-reel change, not a layout change.

The tab is the electrical drain and the thermal path at the same time, so it is soldered to copper rather than bolted to a heatsink. Everything about cooling this class runs through the pad under that tab.

Pin numbering is identical on all three sheets, which is what makes this class a genuine drop-in comparison rather than a footprint study.

Why the 290A Label Shrinks and the 160A Label Doesn't

A continuous current rating is calculated, not measured, and manufacturers pick between two limits. The overall rating is the lower of a thermally limited figure and a package limited figure.

The thermal figure assumes an ideal heatsink holding the case at 25°C, and it scales with the die's on-resistance. The package figure comes from the bond wires and lead frame, and it barely moves with temperature at all.

These three sheets sit on opposite sides of that choice. The ST sheets state the basis outright — current limited by package — which is why 160A and 80A read the same at 25°C and at 100°C.

The NCE sheet states no basis and derates from 290A to 205A, which is the signature of a thermally limited number.

The gap between the 290A part and the 160A part is not 81% at a hot case. It is 28%. Design to the 100°C row and the ranking changes shape.

There is a second consequence. A package limited part can be pushed past its rating without immediate failure — the documented long-term mechanisms are wire fusion, degradation of the molding compound and electromigration, not a sudden breakdown.

A thermally limited rating fails differently, because the junction temperature is the thing being held.

The part with the lowest current label has the lowest on-resistance, and the part with the highest label does not. That is the clearest argument for cross-referencing on resistance rather than on current.

The test currents differ, which is the usual caveat: 20A, 40A and 80A. None of the three is tested at the current printed in its own headline rating.

Two of the three sheets tell you what the part does on a real board, under the same 1in² 2oz copper condition. The third gives you the junction-to-case figure and leaves the rest to you.

That single missing row is often the difference between a substitution that works on paper and one that works in a sealed enclosure.

Other Alternatives Worth Knowing

STB120N4LF6 is the answer when the gate rail cannot reach 10V. It is a 40V D2PAK part in the same series with logic-level drive, so it covers the case that disqualifies all three parts above.

NCE40H21 is the in-family move. It carries the identical 310W budget, the identical 0.48°C/W path and the identical 2500mJ avalanche rating in a TO-220 you can bolt down instead of solder.

Two more are worth a look for a specific gap. STB100NF04T4 is the same ST family rated 120A with the same D2PAK footprint. And where the design needs avalanche margin above everything else, the in-family NCE30H15K carries 1700mJ if a 30V rail is enough.

Frequently Asked Questions

Q1: Is the STB270N4F3 a drop-in replacement for the NCE40H29D?

A: Mechanically yes, electrically with two checks. Same D2PAK footprint, same pin assignment. The checks are the gate threshold, which is 2–4V against 1.3–2.5V here, and the gate charge, which is lower on the ST part but specified at a much higher test current.

Q2: Which has lower on-resistance?

A: The STB270N4F3, at 2.0mΩ maximum against 2.4mΩ. That is a 17% advantage, inside the ±20% band the industry treats as a close match. Note the test currents differ: 80A for the ST part, 20A for the NCE part.

Q3: Why does the 290A rating fall to 205A while the 160A rating does not move?

A: Because they are limited by different things. The ST sheet states its current is limited by the package, and the package limit barely changes with temperature. The NCE figure derates with the case, which is the signature of a rating computed from the silicon's thermal limit.

Q4: Can I compare the avalanche ratings directly?

A: No — the conditions differ in three ways. The NCE sheet gives energy with an inductance and a supply rail; both ST sheets give energy with a test current and a rail. 2500mJ at 0.5mH works out near 100A of peak avalanche current, against 1.5J and 1J specified at 40A and 80A.

Q5: What does "current limited by package" actually mean?

A: The limit is the internal wiring and lead frame, not the die. Exceeding it does not destroy the part immediately; the documented long-term failure modes are wire fusion, degradation of the molding compound and electromigration. It is a reliability boundary rather than an instantaneous one.

Q6: Why does the NCE sheet have no junction-to-PCB thermal figure?

A: It publishes only the junction-to-case value, 0.48°C/W. Both ST sheets publish 35°C/W junction-to-PCB on the same 1in² 2oz copper condition, which is the number that tells you what happens without a heatsink.

Q7: Which parameter should I cross-reference on?

A: On-resistance, not current. That is the industry guidance for a substitution where the application is unknown, and these three parts show why: the labels run 80A to 290A while the resistances run 2.0 to 5.0mΩ in almost the opposite order.

Q8: Is a 20% difference in on-resistance acceptable?

A: It is generally treated as a close match. The caveat is that on-resistance alone stops predicting total loss in a switching converter, where gate charge contributes. Folding the two together — R_DS(on) × Q_g — is the usual second check.

Q9: Do I need to change the gate drive?

A: Probably yes. The gate threshold windows do not overlap much: 1.3–2.5V against 2–4V on both ST parts. A driver sized for one threshold window is worth re-checking against the other, especially in a bridge.

Q10: What should I verify on the bench before committing?

A: Three things: the switching waveforms at both edges, watching for voltage spikes and abnormal ringing; the reverse transfer capacitance, which differs by more than twenty times between the NCE part and the STB270N4F3; and the case temperature under load, since two of the three sheets give you a board figure to check against.