Three NCE Power MOSFETs, three different jobs: NCE9435 is the P-Channel high-side load switch, NCE9926 is the dual N-Channel for logic-driven switching and battery protection, NCE30H10K is the 100A power stage.
Pick by asking one question first: which side of the load does the switch sit on, and who drives the gate?
From our Shenzhen distribution work (2025–2026), most boards end up with two of the three — a high-side rail switch, plus a power stage or a logic switch. The three parts share the same trench process and brand, but they're not interchangeable; each socket has one right answer.
| Parameter | NCE9435 | NCE9926 | NCE30H10K |
|---|---|---|---|
| Polarity | P-Channel (single) | Dual N-Channel (independent) | N-Channel (single) |
| Package | SOP-8 | SOP-8 | TO-252 (DPAK) |
| Drain-Source Voltage | -30V | 20V | 30V |
| Continuous Current | -5.1A | 6A per channel | 100A |
| RDS(on) best spec | 53mΩ max @ -10V | 28mΩ max @ 4.5V | 5.5mΩ max @ 10V |
| RDS(on) @ logic level | 85mΩ @ -4.5V | 37mΩ @ 2.5V | not specified |
| Gate drive needed | -4.5V to -10V | 2.5V to 4.5V | 10V (VGS(th) 3V) |
| Gate charge (Qg) | 11nC | 4nC | 70nC |
| Power dissipation | 2.5W | 1.25W | 110W (case temp) |
| Gate-Source max | ±20V | ±10V | ±20V |
| Thermal path | drain pins only | drain pins only | exposed tab (Drain) |
| Pinout | 1–3=S, 4=G, 5–8=D | S1/G1/S2/G2 + D1/D2 | 1=G, 2=D, 3=S, tab=D |
| MCU direct drive? | 5V logic yes | 3.3V/1.8V logic yes | no — needs driver |
| Typical socket | high-side load switch, reverse-pole protection | battery protection, logic load switching | motor drive, inverter, BMS discharge |
The whole line in one table: current class goes 5A → 6A×2 → 100A; gate drive difficulty goes 4.5V → 2.5V → 10V-plus-driver.
The 9435 is the only one that does high-side without a driver; the 9926 is the only one rated at 2.5V drive; the 30H10K is the only one whose RDS(on) is a fraction of the others.
The 9435 is P-Channel: source to the rail, drain to the load, gate pulled low to switch on. It's the natural high-side part because the gate references the rail — no charge pump needed.
The 9926 and 30H10K are N-Channel: they switch the low side (or need a driver/level shifter for high side). N-Channel wins on RDS(on) per die area, which is why the power stage is N-Channel.
The 9926 is specified at 2.5V — 1.8V and 3.3V logic turn it fully on. The 9435 needs 4.5V minimum for its rated RDS(on); a 3.3V pin switches it, but at 2–3× the resistance.
The 30H10K is a driver's part: 70nC gate charge and 3V threshold mean a GPIO holds it in the linear region and roasts it. Budget a gate driver delivering 10V with 2–4A peak.
At 10A, conduction loss is 5.3W on the 9435, 2.8W on a 9926 channel, and 0.55W on the 30H10K — a 10× spread. The 30H10K's 100A rating assumes a cool case and a copper pour on the tab; the SOP-8 parts are package-limited regardless of die capability.
Both SOP-8 parts dissipate through their drain pins — a few watts at most. The TO-252 has an exposed tab that is Drain and doubles as the thermal path — that's how 110W becomes physically possible. Same trench process, different heat budgets.
Packages at a glance: two SOP-8s (9435, 9926) cool through drain pins; the TO-252 (30H10K) adds an exposed tab — Drain and thermal path in one.
Conduction loss at 10A — the current class decides everything:
Same 10A, ten times the heat. The 9435's 5.3W in an SOP-8 is already beyond its 2.5W budget — that's why current class must be chosen before package.
Three questions, three parts: high-side without a driver → 9435; logic-driven switching or battery protection → 9926; real current with a driver → 30H10K. Most boards answer two of the three.
A: P-Channel (9435) when you want a simple rail-referenced gate; N-Channel with a driver when you need lower loss. High-side N-Channel requires the gate 10V above the floating source — a charge pump or level shifter. For loads under ~4A, the 9435's simplicity wins; above that, an N-Channel power stage with a proper driver is the standard answer.
A: The NCE9926 — its two independent channels tied drain-to-drain give true bidirectional blocking. One channel opens on over-charge, the other on over-discharge, and each body diode blocks the direction it would otherwise leak. The 9435 and 30H10K are single-channel and can't do this alone.
A: The NCE9926, rated at 2.5V drive. The 9435 needs 4.5V for its specified RDS(on) — at 3.3V it switches but at 2–3× the resistance. The 30H10K is not a logic-level part at all: 70nC and 3V threshold demand a driver.
A: Only if you rework the drive and the layout. Same 30V rating, but the 30H10K is N-Channel low-side with a 10V gate requirement and a TO-252 tab that needs a copper pour. The 9435 is P-Channel high-side with a 4.5V gate. They solve different problems — the polarity alone changes the whole circuit.
A: Because most parts are only specified at 4.5V or 10V — at 3.3V their RDS(on) is uncharacterized. The 9926 guarantees 37mΩ at 2.5V, so a 3.3V logic rail fully enhances it. On the 9435, 3.3V drive is exactly where the "unspecified resistance" problem starts.
A: 5A is the uncomfortable middle — the 9435 is at its edge, a 9926 channel needs headroom, and the 30H10K is overkill. At 5A the 9435 dissipates 1.3W (2.5W budget) — workable with good copper. If the design is battery-powered with logic drive, a 9926 channel at 5A is 0.7W — fine. If the system has a driver already, the 30H10K at 0.14W is the coolest answer.
A: Technically yes, practically no. Three paralleled 9926 channels give 84mΩ total — still 15× the 30H10K's 5.5mΩ, with three times the package area and gate drive complexity. For real current, use the power-stage part. The 9926's parallelism benefit is only for redundancy or current sharing at its own class.





