The NCE0115K is a 100V, 15A N-channel power MOSFET in TO-252 (DPAK) with three pins and an exposed tab — pin 1 gate, pin 2 drain, pin 3 source, and the big tab is also the drain, electrically live.
That last sentence is the root of most field failures: the tab being the drain means the heatsink copper under the part is at switch-node voltage. Anyone who ties it to ground or forgets the thermal pad kills the board.
From what we see across Shenzhen lots (2025–2026), TO-252 layouts with undersized pads and missing thermal vias are the #1 repeat design error on 100V power boards.
Pin assignment: pin 1 is the gate (control), pin 2 the drain (switch-node side), pin 3 the source (ground side).
The exposed tab under the body connects to the drain — same node as pin 2, electrically live. The PCB copper beneath must never touch ground or any other net.
| Pin | Name | Function |
|---|---|---|
| 1 | Gate | Control input — drive to 4.5–10V to turn on; VGS(th) 1.0–2.5V |
| 2 | Drain | Switch output — connects to the load/inductor side; 100V max |
| 3 | Source | Ground reference — returns the load current; keep traces short |
| Tab | Drain | Exposed pad — same node as pin 2, electrically live, also the thermal path |
| Parameter | Value |
|---|---|
| VDS | 100V |
| ID continuous | 15A at TC 25°C (10.6A at 100°C) |
| RDS(on) @10V | 80mΩ typ / 100mΩ max |
| RDS(on) @4.5V | 85mΩ typ / 110mΩ max |
| VGS(th) | 1.0V min / 1.6V typ / 2.5V max |
| Qg @10V | 22.3nC |
| EAS | 200mJ |
| RθJC | 3°C/W |
| PD | 50W |
Low-side load switch on a 48V bus: the load connects from VBUS to the drain (pin 2); the source (pin 3) returns to ground; a GPIO through a gate resistor drives pin 1.
The 100V rating absorbs the inductive spike when the switch opens — a 60V part would avalanche on exactly this node.
But why does the gate resistor matter on a load switch?
1. Tying the tab to ground. The tab is the drain — electrically live at switch-node voltage. Layouts that pour ground under the tab short the switch; always route the tab pad to the drain net, with the same voltage and ripple as pin 2.
2. Undersized thermal pad. The 50W rating needs a large top-layer pad, duplicated on inner and bottom layers with thermal vias. Datasheet RθJA assumes specific copper — real boards with small pads measure double the listed value, and the part throttles or dies.
3. Solid copper under the tab causing tombstoning. A big solid pad heats slowly while the pins heat fast, and reflow surface tension shifts the part. Use a grid or cross-hatch solder mask on the tab pad to keep the part seated.
4. Driving the gate at marginal voltage. VGS(th) sits at 1.0–2.5V; a 3.3V GPIO through a weak driver can leave the part half-on in the linear region. Give it 4.5V or a proper driver, or it runs hot without ever fully conducting.
5. Skipping the gate resistor. The gate is a capacitive load — a long trace without a series resistor rings and oscillates at MHz. A 10–100Ω resistor in the gate line damps the ringing; a 100pF gate-source cap helps on noisy boards.
6. Designing at the 25°C current rating. 15A is only at 25°C case; at 100°C it's 10.6A, and a real board with a small pad derates further to 6–8A. If the load pulls 12A steady-state, step up to the NCE0125AK class instead.
ICMASS stocks the NCE0115K with lot testing for threshold voltage and RDS(on). The 100V MOSFET class is a re-marking favorite — a 60V die dressed as 100V fails exactly on the spikes this part is bought to survive.
Contact us for pricing and availability on any quantity; volume orders ship directly from the NCE Power production line.
What about 3.3V boards that can't spare a 5V rail?
A: Pin 1 is the gate, pin 2 the drain, pin 3 the source — the exposed tab is also the drain. When facing the marked side of the TO-252 body with the pins down, pin 1 is on the left. The tab duplicates pin 2's net, so it carries switch-node voltage, not ground.
A: 4.5V turns it fully on (85mΩ); 10V gives the best case (80mΩ). The threshold sits at 1.0–2.5V, but running right at threshold leaves the part half-on in the linear region. Drive at 4.5–10V through a gate resistor, and give a 3.3V system a proper driver.
A: Yes — with a bootstrap driver, but the gate drive must reference the source, not ground. Stack Exchange threads on 48V motor drivers use parts like the IR2110 with a bootstrap capacitor sized from the MOSFET's gate charge — one 48V example runs a 22µF cap at 30kHz PWM. The low-side switch must turn on periodically to recharge the cap, and high duty cycles need a charge-pump top-up.
Watch the driver's supply limit: TLP250-class opto drivers top out at 35V and can't sit directly on a 48V rail. Most 48V designs still use the NCE0115K as a low-side switch for exactly this reason.
A: Roughly a 1-inch square, duplicated with thermal vias into inner and bottom layers. Teardowns that killed parts fixed them by adding eight 0.3mm vias under the pad. Without the copper, the effective RθJA doubles and the 50W rating becomes fiction.
A: Thermal imbalance between the big tab pad and the small pins — surface tension pulls the part while one side is still molten. Solid ground pours under the tab are the usual trigger. Use a grid or cross-hatch solder mask on the tab pad, or split it into smaller pads.
A: Nothing electrically — they're the same drain node on the same die. Pin 2 carries the current path and the tab carries the heat; both are at switch-node voltage. Route the tab pad to the drain net and let the pin carry what the layout needs.
A: Add a series gate resistor (10–100Ω) between the driver and pin 1, close to the part. The gate's input capacitance (830pF) with trace inductance forms an LC tank that rings at MHz. The resistor damps it; a 100pF gate-source cap suppresses residual oscillation on noisy boards.





