Tutorials
A Gate Voltage Is Not an On-State Guarantee: MOSFET Basics
Published · Updated
A controller output goes high and a MOSFET is expected to switch a load. What voltage actually reaches the gate relative to the source, and what evidence says the channel has the resistance used in the loss estimate? Those two questions are more useful than treating a threshold-voltage number as a switch command. This lesson follows a low-side conceptual example, a source-reference counterexample and a conditional calculation without claiming a completed driver design.
Name gate, source and drain before following the load

A MOSFET has gate, source and drain terminals, with additional connections or arrangements possible in particular packages and devices. The schematic symbol describes an electrical device, while the package drawing establishes the actual terminal mapping for a selected ordering code. Do not infer a pinout from a familiar three-lead appearance. A correct conceptual circuit can still be assembled incorrectly when a different package or manufacturer connection is assumed.
Consider a conceptual N-channel enhancement MOSFET placed in a low-side load path. The load is between a positive supply and drain, while source returns toward the supply reference. A separate gate-drive signal controls the channel. This drawing explains references and paths only. It is not a reference design, does not specify an actual driver and does not authorize connecting an inductive or high-energy load without its required protection.
The gate is insulated in the intended device structure, but that does not mean it never draws current. Changing gate voltage requires moving charge, and the actual device has capacitance and leakage behavior. During switching, the drive path and its impedance matter. Treating the gate as an ideal permanent open circuit hides the work that the driver must perform and can lead to an unsupported switching-time assumption.
The body-diode path is another reason a MOSFET is not always equivalent to a mechanical switch. Its orientation and relevance depend on device type and circuit arrangement. A single MOSFET may not block both directions in the way a beginner expects. Follow the actual symbol and datasheet rather than treating off as a universal disconnected state. Keep this question separate from whether the channel has adequate on-state conduction.
The source is the reference that changes the answer
Gate-to-source voltage is VG minus VS when both node voltages use the same external reference. A gate at 3.3 volts with source at 0 volts gives3.3 volts of VGS. If source instead rises to 2 volts while gate remains at 3.3 volts, VGS becomes1.3 volts. The gate has not changed relative to ground, but the voltage controlling the channel has changed substantially.
This simple subtraction explains why a gate-to-ground reading may be insufficient in a moving-source circuit. A scope connection, meter reference and driver topology must be chosen appropriately for the real circuit. This introductory illustration does not provide a probing procedure for a high-side or hazardous switching node. It simply establishes that the relevant voltage difference must be identified before interpreting a reading.
A low-side example with source near a reference is convenient for learning, but even that reference may move because of interconnect impedance. Gate-loop and source-return behavior become important during fast switching. The same printed drive-voltage label can therefore conceal different conditions at the device terminals. A source connection used for sensing or driving needs to be understood from the actual package and layout.
Try stating the design requirement without using the word high. Name the desired VGS interval, load current, temperature and operating state. Then ask what gate and source conditions produce it, including transitions. This converts a vague logic description into a question that device and driver evidence can answer. It also makes clear why different source states can require different drive arrangements.
Threshold answers a different question from low on-resistance
A threshold-voltage specification describes a defined small-current test condition. It does not guarantee the headline low on-resistance at that same gate voltage. The on-resistance table has its own gate-drive, drain-current and temperature conditions. The practical reading task is to find evidence at the intended drive and operating conditions, rather than selecting the smallest threshold number and assuming the rest follows.
Toshiba's gate-source-voltage guidance makes this distinction explicit. The relevant reference is the applicable on-resistance condition, while the gate voltage including transients must remain within the device's absolute limits. That general explanation does not establish a suitable gate voltage for every MOSFET technology or ordering code.
A typical graph can help visualize a trend, but it is not equivalent to a guaranteed maximum table entry. Read its temperature, pulse and other conditions. If the desired drive condition is absent, keep the on-resistance at that condition unresolved. Do not silently interpolate a guarantee from a picture. A candidate may require better documentation, another drive arrangement or a different component before it can enter a qualified shortlist.
The label logic level is also not a universal promise for every logic supply. Check the exact evidence for the actual output level. A device characterized at one drive voltage is not automatically qualified at a lower one, and the controller output may sag under a transient gate load. The question is not whether two catalog labels sound compatible; it is whether their specified conditions support the intended circuit.
Estimate conduction loss while leaving the other losses visible
For an assumed on-state resistance and an appropriately described current, conduction loss can be estimated with I squared times RDS(on). Suppose a hypothetical operating case uses 2 amperes and 50 milliohms. The result is0.2 watts. That number is conditional on the resistance applying at the actual gate drive, current and temperature. It is not a specification for any particular MOSFET and is not a measured thermal result.
Doubling current to 4 amperes while retaining the same assumed resistance raises the result to 0.8 watts. That quadratic relationship makes the current assumption important. For a varying waveform, use the relevant RMS or interval treatment rather than substituting peak current as though it were a constant average. Resistance itself can change with operating temperature, so the first calculation may need an iterative device-specific analysis.
Conduction loss is only one part of a switching application. Charging and discharging the gate, transitions with simultaneous voltage and current, body-diode behavior and other circuit elements can contribute losses. Gate charge and the driver path affect timing. The loss estimate here contains none of those mechanisms. A lower assumed on-resistance does not prove a better total design when the rest of the operating conditions differ.
Power is also not junction temperature. A thermal calculation needs the actual board, package, boundary conditions and appropriate thermal model. Nor does a modest conduction estimate establish safe operating area during startup or a fault. The Nexperia MOSFET/GaN application handbook (source address: https://assets.nexperia.com/documents/user-manual/Nexperia_document_book_MOSFETGaNFETApplicationHandbook_2020.pdf) is a broader reference for these separate design questions. Use the exact current device documentation before making a selection claim.
Explore the reference voltage and challenge the evidence flag
The interactive worksheet takes gate and source voltages relative to the same reference, assumed current and assumed on-resistance in milliohms. It reports VGS and a conduction-only loss. It also asks whether the on-resistance evidence has been verified at the stated drive and temperature. Answering yes records your statement; the tool does not inspect a datasheet or certify the component.
Static reference: VG=3.3 V,VS=0 V,I=2 A,assumed 50 mΩ →VGS=3.3 V,conditional loss0.2 W. Interactive fields appear when JavaScript is available. The explanation and example remain readable without it.
Start with a 3.3 volt gate,0 volt source,2 ampere current and 50 milliohm assumption. Confirm VGS of 3.3 volts and conditional loss of 0.2 watts. Now move only the source to 2 volts. The subtraction gives1.3 volts of VGS, but the arithmetic loss still uses the entered50 milliohms. The tool warns about missing or conditional evidence because it cannot infer the new channel resistance from the voltage subtraction.
This is a deliberate teaching feature. A calculator that silently retained50 milliohms and called the switch suitable would combine one correct equation with an unsupported device conclusion. Keep the resistance assumption and its evidence visible. Change it only when an applicable source supports the change, or use the tool as a labelled sensitivity study while the selection remains unresolved.
Try a negative VGS and observe that the subtraction remains meaningful while the conduction assumption is still conditional. This worksheet does not calculate the transfer characteristic, channel state or body-diode conduction. Zero or negative on-resistance is rejected, and missing numeric fields produce a readable error. The copyable result includes the unresolved switching, thermal, SOA and driver questions so that a number cannot travel into a project note without its limitations.
Move from a controller signal to a scoped design request
Gate charge provides another useful planning question without supplying a complete switching model. If a hypothetical charge of 10 nanocoulombs must be moved in 100 nanoseconds, charge divided by time gives an average current of 0.1 amperes over that chosen interval. This is an illustrative arithmetic requirement, not a claim about a particular MOSFET or GPIO. The actual gate-charge curve, voltage excursion and driver behavior remain necessary.
A controller pin's static output-current capability cannot be compared casually with that average transient number. Output impedance, rise and fall behavior, source and sink paths, gate resistance and the device's nonlinear charge distribution all matter. A driver may be needed to establish the intended terminal waveform. Conversely, an application with modest switching requirements should still be evaluated on its own conditions rather than inheriting an unnecessary driver from an unrelated high-speed example.
Startup adds another state that a high-versus-low sketch can hide. The controller may be unpowered, configuring or temporarily not driving the signal. Define the desired load state in those intervals and obtain evidence for the actual pull path, driver and MOSFET arrangement. A resistor that biases a gate can support a default state, but it does not by itself qualify every transient, fault or power-sequence case. Avoid treating a copied resistor value as a universal startup solution.
When comparing candidates, keep a short table with the actual gate-drive condition, relevant on-resistance evidence, charge data, voltage stress, current waveform and unresolved thermal or SOA question. Separate guaranteed entries from typical curves. If one candidate lacks a required condition, record a gap rather than inventing an equal score. A documented shortlist with an unresolved item is more honest and more actionable than an unsupported best-part claim.
That table also improves a quotation request. Purchasing can verify the exact code and supply conditions, while the engineering review can address the missing operating evidence. Neither job should silently replace the other. If a proposed substitute changes package, drive requirements or protection behavior, reopen the affected design checks before approval. The original arithmetic can remain valid as an assumption study even while the actual component choice changes.
An FPGA output that controls a load needs a complete interface review. Output-drive conditions, bank voltage, pin constraints, startup state, external driver and load behavior may matter. For a project using XC7A100T-1CSG324C, begin with that exact device's evidence and the applicable manufacturer documents. The link identifies controller context; it does not recommend this FPGA as a MOSFET or validate a direct gate connection.
Organize the supply requirements with the rail-planning worksheet and the board checks with the PCB bring-up checklist. The engineering workbench helps identify other project tasks, while the PCBA preparation guidance covers assembly handoff. The relevant next step is the one that resolves a remaining constraint, not simply another article containing the same keyword.
For parts, request a MOSFET or driver quotation with exact manufacturer ordering codes and quantities after the electrical review. If the design is not selected, state the intended gate drive, load, temperature range and switching task instead of inventing a qualified substitute. Pricing, stock and lead time remain subject to confirmation. A device name or low resistance number is not a supply guarantee.
For a complete board, submit a project quotation request with board quantity, schematic/BOM revision, sourcing and test scope, and the unresolved interface questions. Use the form for a non-confidential brief and arrange suitable transfer for private design files. The useful result of this lesson is an evidence checklist and a bounded loss estimate. It should help you ask a precise design question while keeping threshold, conduction, switching and thermal conclusions distinct.
Show the original mosfet concept animation

Concept animation, not a hardware recording or a validated simulation. Close this disclosure to hide it. Static steps and numerical assumptions are explained in the article.