What Is a MOSFET: Three Losses and One Package
A MOSFET is a switch you turn on with voltage. Its limits are heat, and the heat leaves through the package.
The letters stand for metal-oxide-semiconductor field-effect transistor. The useful part is simpler than the name.
Put a voltage on the gate and the path between drain and source becomes a low resistance. Take the voltage away and the path opens.
Nothing wears out, because nothing moves. What wears out is the thermal budget. Every watt lost inside the part has to leave through plastic and metal.
The Short Answer
A MOSFET is a voltage-controlled switch with three working terminals. The gate decides, and the drain-to-source path carries the load current.
It differs from a bipolar transistor in one practical way. The gate draws almost no current, because a layer of silicon dioxide insulates it from the channel.
That insulation is also the catch. The gate behaves like a small capacitor. It must be charged and discharged every time the part switches, as the maker’s technical overview puts it.
So the part is chosen on three numbers that pull against each other: how much resistance it has when on, how much charge the gate needs, and how much heat the package can shed.
Cheap parts are cheap because one of those three was traded away. The data sheet will not say which.
A Switch With an Insulated Gate
The gate sits above a channel region, separated from it by a thin oxide layer. Voltage on the gate attracts charge carriers under the oxide.
Enough voltage and a conducting channel forms between drain and source. That voltage is the gate turn-on level, and it is the first number a designer looks for.
Below that level the part is meant to be off. In practice a small leakage current flows, and it grows with temperature.
There are two families of behaviour. An enhancement-mode part is off at zero gate voltage; a depletion-mode part is on, and needs voltage to turn off.
The market is almost entirely enhancement-mode, and the reason is safety rather than performance. If a control signal fails to zero during a fault, the part goes off, as one maker’s FAQ explains.
A device that fails on is a short circuit across the load. That single sentence explains why the other family is a niche part.
N or P, and Why N Wins
The two channel types differ in which carrier carries the current. An N-channel part moves electrons; a P-channel part moves holes.
Electrons move more easily than holes. For the same process and the same die area, the N-channel part therefore has the lower resistance.
That is why the efficient designs are built on N-channel parts. Synchronous rectifiers and DC-DC converters use them almost exclusively.
| Channel | Carrier | Same die area | Natural place |
|---|---|---|---|
| N-channel | Electrons | Lower on-resistance | Low side, near ground |
| P-channel | Holes | Higher on-resistance | High side, simpler drive |
The choice is not only about the silicon. An N-channel part wants its source near ground, which suits a switch on the low side of a load.
Put the switch on the high side and the source sits at the load voltage. The gate then needs to be driven above the supply, which takes a driver circuit.
A P-channel part avoids that driver, at the cost of a larger die for the same resistance. It is the usual answer in small, simple circuits.
There is a limit worth knowing. A power MOSFET can need a gate voltage higher than the voltage it is switching. That is why high-side drive is a circuit of its own.
The Three Losses
Every watt that heats a MOSFET comes from one of three places. A part chosen on one of them will usually be wrong on the other two.
Conduction loss. While the part is on it behaves as a resistor, so the loss is the current squared times that resistance, held for as long as the switch is closed.
This is the loss a buyer can see on a data sheet, so it gets the attention. It also grows with the square of the current.
Switching loss. During each transition the part passes through a region where both voltage and current are present at once. The energy in that overlap is lost, every cycle.
Switching loss scales with frequency. Double the switching rate and the same transition costs twice as much per second.
Gate drive loss. The gate is a capacitor, so charging it costs energy, and that energy is thrown away each time it is discharged.
It rises with gate charge, with drive voltage and with frequency. A part with low on-resistance and a heavy gate can lose more in drive than it saves.
Put the three together and the trade becomes visible. A big die has low resistance and a heavy gate; a small die is the other way round.
Which one wins depends on how fast the part switches. At low frequency, conduction dominates and the big die pays. At high frequency, the drive and the transitions take over.
That is why the same circuit can want two different parts at two different switching speeds, and why a part number copied from another design is a guess.
Why the Current Rating Is a Thermal Number
The current on the front of a data sheet is not an electrical limit. It is the current the part can carry while its junction stays inside a stated temperature.
The maker’s own guidance says it plainly: the current rating holds “under specific thermal conditions”. Change the heatsink and the number moves.
The path from the silicon to the air has stages, and each one has a thermal resistance. The die-to-case figure is on the data sheet; the case-to-air figure belongs to your design.
Multiply the power lost by the total thermal resistance and you get the temperature rise. Add the ambient and you have the junction temperature.
That arithmetic is the whole design. The data sheet’s current rating assumes a case temperature the part will never see inside a sealed enclosure.
Two rules follow. Derate the current before you trust the number, and measure the case temperature on the first prototype rather than the tenth.
Losses also feed themselves. A hotter part has higher resistance, which brings more loss, which brings more heat, until the loop settles or the part fails.
The Package Is Often the Limit
Buyers treat the package as a delivery format. It is part of the circuit, and it appears twice in the numbers that matter.
The first appearance is inside the on-resistance. The figure on the data sheet is a sum: bond wires, source metallisation, the channel, the region under the gate, and the drift layer.
The reference on power MOSFETs breaks that sum down, and the bond wire is not a rounding error in a small package.
The second appearance is the thermal path. The same die in a leaded package and in a leadless one has two different current ratings. The heat leaves by a different route.
That is the honest version of “the package is often the limit”. It sets the resistance in, and it sets the heat out.
Two packages of the same die are therefore not a straight swap in a design that was tuned. The board layout, the copper area and the mounting all assumed one of them.
Where a single part runs out, two can be put in parallel. The current shares, but only if both are driven properly and both sit on the same thermal footing.
The Diode You Did Not Order
Inside every power MOSFET there is a diode between source and drain. It is a by-product of the structure, and it conducts whether the design wants it or not.
In most circuits it never turns on. Switch a load to ground and the body diode stays reverse biased for the whole cycle.
In a bridge it matters. When one device turns off, the current in an inductive load has to go somewhere, and for a moment the diode of the other device carries it.
That moment is the dead time, and it is deliberate. Both devices must never conduct at once, so the design inserts a gap, and the diode fills it.
The cost is recovery. A diode that has been conducting needs time and charge to stop. That charge is lost as heat in the device that turns on next.
So the body diode is a specification, not a footnote. Two parts with the same voltage and resistance can differ in how much that recovery costs.
There is a second use for it. In a synchronous rectifier the diode is bypassed by the channel, and the part conducts in the direction that looks like reverse to a simple switch.
Five Lines That Decide the Part
A part number carries five decisions. Written out, they are the difference between a design that runs cool and one that fails in the field.
| Line | What to write | What goes wrong without it |
|---|---|---|
| Voltage class | The drain-source rating, with margin for the spikes the load makes | A 30 V part in a 24 V motor circuit dies on the first switch-off |
| On-resistance | The figure, plus the gate voltage and temperature it is quoted at | A logic-level number is read off a standard-gate part |
| Gate charge | Total charge, and the drive voltage it was measured at | The driver cannot switch it fast enough, and losses move to the transition |
| Package and thermal path | Package, junction-to-case resistance, and the copper area assumed | The current rating is met on paper and not on the board |
| Body diode | Recovery charge, or a statement that the circuit does not need it | Efficiency and noise both get worse in a bridge |
The second line is the one most often misread. A low on-resistance quoted at a gate voltage the driver cannot reach is a number for a different circuit.
The fourth line is the one that gets skipped, because the package looks like a delivery detail. The thermal resistance and the copper area are what turn a die rating into a board rating.
Two more lines belong on the same sheet when the part switches an inductive load: the safe operating area, and the energy the part can absorb in avalanche.
Temperature is measured, not assumed. A bead bonded to the case is the cheapest way to find out whether the arithmetic held.
Where the load current has to be known, a magnetic part can measure it without a shunt. A sensor that reads the field around the conductor does that job.
Six Lines for the Order
Put the five lines above into six sentences and the part number stops carrying the whole specification on its own.
One. Name the channel type, the mode and the voltage class.
Two. Give the on-resistance with the gate voltage and temperature it holds at.
Three. Give the gate charge and the drive voltage used to measure it.
Four. Fix the package and the thermal resistance, and state the copper area assumed.
Five. State the body diode requirement, including recovery charge if the circuit is a bridge.
Six. Agree the incoming test: which curve, which gate voltage, and which reading counts as a pass.
The test is worth the paperwork. A batch with the right marking and the wrong die passes a visual check and fails on the first hot day.
For a comparison across suppliers, the answers need to line up. A single document for all five lines does that better than four data sheets in four layouts.
Before the order grows, handle the parts. The stalls that sell these components are also where the marking on a reel gets checked against the die inside it.
None of that replaces a look at the line. A visit before the first batch should include the curve tracer, because a counterfeit part matches the label and not the curve.
Send me the MOSFET data sheet you were quoted and I will read the five lines. I will say whether the on-resistance is quoted at a gate voltage your driver can reach, whether the package can shed the heat, and which line is missing. No cost, no obligation.
Check My PartQuestions Buyers Ask
Is a MOSFET a switch or an amplifier?
Both, depending on the circuit. In power and digital designs it is a switch. It can also amplify in its linear region, which is why old audio amplifiers used it.
Why does the gate draw no current but still need power?
The gate is insulated by an oxide layer, so no steady current flows. It behaves as a capacitor that must be charged and discharged every time the part switches.
What is on-resistance?
It is the resistance between drain and source while the part is on. The loss it causes is the current squared times that resistance, held for as long as the switch is closed.
Why is my MOSFET hot when the current is well under the rating?
Usually switching loss or a slow gate drive, not conduction. The part spends longer than it should in the region where both voltage and current are present.
Can I replace a MOSFET with one that has a higher current rating?
Only if voltage, gate charge and package also fit. A bigger die usually brings a heavier gate, which can cost more in drive than it saves in conduction.
Do I need to worry about the body diode?
In a simple low-side switch, no. In a bridge or a synchronous rectifier it carries current during dead time, and its recovery charge becomes part of the loss budget.
Questions & Comments
We read every one and reply within 24 hoursIf a part runs hot well under its rated current, send me the package and the gate drive. The transition, not the on-state, is usually paying for it.