NCE9926 vs NCE4606 vs NCE4953 vs NCE4963: Side-by-Side Comparison

Read the four resistance blocks together and the family separates cleanly: one part has no 4.5V number, one has no 10V number, and only two reach 2.5V. There is no single drive voltage at which all four are specified.

Now the practical version — resistance at 4.5V gate drive, the rail most 5V logic actually produces:

At the drive voltage a 5V logic rail actually delivers, the NCE4953 is 3.3× the resistance of the NCE4963 — and the NCE4606 has no number at all. The gray bar is the important one: there is nothing to extrapolate from, and guessing is how boards run hot.

But why does a resistance number matter more than a current rating? Because the current rating is a ten-second figure, while the resistance is what turns it into heat. Here is each part at its own headline current, against its own documented budget:

Three of the four push past 80% of their own package budget with a single channel conducting, at the currents printed on their own datasheets. Only the NCE4963 keeps real headroom — and it earns that with the biggest input capacitance in the group.

Key Differences Between the Four

The gate-drive floor is the real spec

The NCE4606 is a 10V part: its only R_DS(on) rows are at 10V for both channels. The NCE4953 adds a 4.5V row, at 64% higher resistance. The NCE9926 and NCE4963 both publish 2.5V rows.

What about the drive your gate circuit actually produces? That is the number you must design at — and for three of these four, lower drive means measurably worse resistance.

Two of them share a pin map, and two of them don't

The NCE4953 and NCE4963 are pin-identical: S1 on pin 1, G1 on pin 2, S2 on pin 3, G2 on pin 4, D1 on pins 7–8, D2 on pins 5–6.

The NCE4606 is the opposite: S2 on pin 1, G2 on pin 2, S1 on pin 3, G1 on pin 4. Same package, same pin count, channel numbers swapped.

Conduction performance is paid for with switching speed

The NCE4963 has the lowest resistance of the four (27mΩ at 4.5V) and the best thermal path (42°C/W) — but 1210pF of input capacitance and a 70ns turn-off delay.

The NCE4953 is 2× the resistance and roughly 5× faster. Low threshold and low gate swing buy conduction loss back at the cost of switching loss.

Every one of them is rated for ten seconds, not forever

All four datasheets qualify R_θJA on FR4 at t ≤ 10 sec, and none publishes a steady-state figure. The NCE4606 is the only one that splits its thermal rows per channel.

The other three give a single package-level number, which both channels share. Read every current rating on this page as a ten-second number.

The color is channel number, not device type. On the NCE4606 channel 2 sits on the outer pins; on the NCE4953 and NCE4963 channel 1 does. The pin-1 dot moves with it — top-left on the complementary part, bottom-left on the dual-P pair.

Substituting one of these for another without re-reading the pin map is the most expensive mistake in this family. The packages drop straight in; the connections do not.

When to Choose Each One

Choose NCE9926 (dual N, 20V) when both switches are low-side. Two N-channels, 6A each, characterized at 4.5V and 2.5V, with independent D1/D2 and 10nC of gate charge. Reach for it when the load returns to ground and you don't need a high-side device.

Choose NCE4606 (complementary, 30V) only when you genuinely need both polarities. It is the only part here that pairs an N-channel with a P-channel, which is what a level-shifted high-side switch wants.

Just accept that its resistance rows are 10V-only. If your gate drive can't reach 10V, this is the wrong part regardless of the channel arrangement.

Choose NCE4953 (dual P, 30V) when the rail exceeds 20V, or when you need speed. It has the widest gate window at ±20V and the fastest switching of the four — 14ns turn-off delay.

Its resistance is the highest in the group, and it is the only part here that gets worse at 4.5V drive. Give it 10V if you can.

Choose NCE4963 (dual P, 20V) when drive voltage is scarce and the switch is slow. It is the only part characterized at 2.5V with low resistance, and it has the best thermal path at 42°C/W.

In exchange it is the slowest of the four and carries the narrowest gate window at ±12V.

And when the answer is none of them: if your load is inductive and unclamped, every part on this page lacks an avalanche rating and a body-diode recovery figure. If you need real power, a single TO-252 with a heatsink beats four SOP-8 channels stacked on the same leadframe.

Three questions and the family resolves. The order matters: the polarity question comes second because it eliminates a part rather than ranking the rest, and the drive-voltage question never appears as a question at all — it's a constraint you check against whichever part the tree lands on.

Frequently Asked Questions About the NCE Dual MOSFET SOP-8 Family

Q1: Why is R_DS(on) higher at lower gate drive for three of these four?

A: Because the channel is less enhanced, and the datasheets are honest enough to publish both numbers. The NCE4953 goes from 55mΩ at −10V to 90mΩ at −4.5V — 64% worse. Industry parts show the same step; the NDS9948 roughly doubles from 250mΩ to 500mΩ across the same range. Always design at the drive voltage your circuit actually produces.

Q2: Can I use the NCE4606 pin map for the NCE4953?

A: No — the channel numbers are swapped. The NCE4606 puts S2 on pin 1 and S1 on pin 3; the NCE4953 and NCE4963 put S1 on pin 1 and S2 on pin 3. Same SOP-8, same pin count, opposite assignment, and the pin-1 dot marks a different corner. Re-read the pin map every time you substitute.

Q3: Which one has the lowest conduction loss?

A: The NCE4963, at 27mΩ max at −4.5V. It also has the best thermal path at 42°C/W and the largest package budget at 3.0W — so at −7A and 2.5V drive one channel sits at 64% of budget, the only part here with real margin. The price is 1210pF of input capacitance and a 70ns turn-off delay.

Q4: Which one switches fastest?

A: The NCE4953, at 14ns turn-off delay, with the NCE9926 close behind at 15ns. The NCE4606's N-channel is comparable at 14.5ns. The NCE4963 is the outlier at 70ns — roughly 5× slower. If you're chopping at tens of kHz, that difference decides the design long before the resistance numbers do.

Q5: Does any of them have an avalanche rating?

A: No — not one of the four publishes an E_AS figure, and none publishes a body-diode recovery time. That is a real gap for motor, solenoid and relay loads, where unclamped inductive edges are normal. If your load is inductive, plan a clamp or snubber regardless of which part you pick — the datasheets give you nothing to design avalanche against.

Q6: Are the thermal ratings per channel or for the whole package?

A: It depends on the part, and the difference matters. The NCE4606 splits its thermal rows per channel — 62.5°C/W and 2.0W each. The NCE4953 and NCE4963 give one package-level row each (50°C/W / 2.5W and 42°C/W / 3.0W), which both channels share. The NCE9926 lists 1.25W and 100°C/W as package figures. Read carefully before assuming your budget doubles.

Q7: What gate voltage should I design for?

A: Whatever your driver actually delivers, checked against that part's published rows. 10V is the safe answer for the NCE4606; 4.5V or 10V suits the NCE4953; the NCE9926 and NCE4963 both publish 2.5V rows. The one thing you cannot do is take a 10V number and use it at 4.5V — that is the mistake this whole page exists to prevent.

Q8: When should I skip this family entirely?

A: When you need steady current or real avalanche capability. Every rating on this page is qualified at t ≤ 10 sec on FR4, with no steady-state thermal resistance published, and three of the four reach 80% or more of their budget on a single channel at their own printed current. Above an amp or two continuous, a single TO-252 or TO-220 with a heatsink is the honest answer.