Where the -50A Rating Comes From

The sticker current is measured at T_C = 25°C with the case held there — a heatsink of infinite size, in other words.

Board reality is different. The junction temperature is set by the power you burn times the thermal path from die to air, and the path is the package, the pad, the copper, and the airflow — in that order.

Now the fine print that few people read: the on-resistance rows are measured at currents far below the stickers.

The 50A part's 7mΩ is tested at -10A. The 30A DFN's 10mΩ is tested at -15A. The 30A TO-252's 18mΩ is tested at -20A. None is measured anywhere near its own sticker.

Why does that matter? At the sticker current the die is near its limit and the measurement is not repeatable at pulse widths. The rows describe die quality, not 50A conduction.

Three Packages, Three Heat Exits, Three Envelopes

The same -30V rail, three packages, three ways for heat to leave the die — and the datasheets disagree on which envelope each one gets.

Read the last three rows twice: the 50A part carries the tightest envelope of the three big dies, and the TO-252 part runs the hottest junction ceiling.

That is not a typo. It is package attribution — each datasheet is internally consistent, and each tells a different story about what the die is for.

What a Real Board Delivers: 20A and 30A Worked Examples

Worst-case conduction loss at your current is I² times the max R_DS(on) — use the max row, not the typ.

At 20A: the 50G burns 2.8W, the G 4.0W, the K 7.2W. At 30A: 6.3W, 9.0W, 16.2W.

Now take the junction math with the published envelope: at 30A the G's 9.0W against R_θJC 1.56°C/W means the junction runs 14°C above the case before a watt leaves the copper.

That sounds small — until you remember the case sits at whatever your board's copper, vias and airflow deliver, which no datasheet row tells you.

The Honest N-ch Crossover

P-ch high-side switching is a convenience: the gate drive references the rail, no level-shift IC, no bootstrap. The price is roughly double the R_DS(on) per die dollar of an N-ch.

From what we see across Shenzhen lots (2025–2026), the crossover shows up in reorders: boards that pass about 20-25A continuous stop reordering P-ch and start asking about N-ch plus a driver.

Our own line makes the point: the NCE30H10K — an N-ch, 30V, 100A, 5.5mΩ, TO-252 — carries the current class the P-ch parts struggle to name.

What does that driver actually cost? About a dollar of BOM — and it's the cheapest half-loss you'll ever buy at this current class.

So the rule of thumb: P-ch below ~20A continuous, surge-tolerant; N-ch plus a high-side driver above it. The driver costs a dollar of BOM and buys half the conduction loss.

How We Test, and What We See

We batch-test every 30P lot at the datasheet's own points — not at the sticker. A -30A part gets tested at -15A or -20A per its row, and the milliohm max is the pass line.

What that catches, in practice: relabeled smaller dies. A 30A die relabeled as the 50A part fails the 7mΩ row at -10A; a 25S relabeled as a 30P30G fails the 4.5V row. The test conditions are the counterfeit detector.

Across lots (2025–2026) we also see the honest questions repeat: "can it really do 30A?" is almost always followed by the real question — "what does my board actually deliver?"

The answer is the same every time: measure it. Thermocouple on the pad, a current probe on the load, and the datasheet math on paper — that combination settles more arguments than any spec sheet.

Frequently Asked Questions

Q1: Can a DFN5×6 P-ch really carry 30A continuous?

A: Yes, when the pad delivers — and the envelope says how much of the job is yours. The G's 80W at R_θJC 1.56°C/W is a case rating; the junction stays cool only if the pad, pours, vias and airflow move the watts. On a good 2oz board with airflow, 20-25A continuous is a defensible design; measure with a thermocouple before production.

Q2: Why does the 50A part have a tighter envelope than the 30A part?

A: Datasheet package attribution — and a design statement. The 50G lists R_θJC 3.6°C/W with 35W at T_J 150°C; the 30P30G lists 1.56°C/W with 80W. Both are internally consistent. The practical reading: the 50G is specced for surge duty — pulse energy and duty set its junction temperature, not continuous watts.

Q3: When do I stop buying P-ch?

A: Above roughly 20-25A continuous — that is the honest crossover. P-ch die economics give roughly double the R_DS(on) per dollar of an N-ch. Add a high-side driver to an N-ch and the loss halves for about a dollar of BOM. P-ch stays the right answer for simplicity and surge duty below that line.

Q4: Do I need to test at the sticker current to know my part is good?

A: No — test at the datasheet's own test current. The rows are guaranteed at -10A, -15A or -20A depending on the die, not at -30A or -50A. Testing at the datasheet point catches relabeled dies; testing at the sticker just measures a die past its pulse-test region.

Q5: The DFN pad or the TO-252 tab — which is better?

A: The DFN pad wins on paper; the TO-252 tab wins in the rework bay. The G's R_θJC 1.56°C/W beats the K's 2.5, but the tab is inspectable, hand-solderable, and shares a land pattern with the whole N-ch TO-252 line. Pick by your assembly reality, not by the smaller number alone.

Q6: Is the pulsed current rating (-200A) a design number?

A: Yes, for single events — with the pulse-width caveat in the note. I_DM is limited by junction temperature, so it is a stall-and-inrush budget, not a switching design target. Size your repeated pulsing against thermal impedance, not the sticker.