Field-Effect Transistors

A field-effect transistor controls conduction by changing channel charge rather than injecting base current. That makes the gate nearly open at DC, but it does not make the input irrelevant: every transition must move gate charge, and the drain–gate capacitance feeds fast drain motion back into the gate drive.

Families and regions #

A JFET uses a reverse-biased p–n junction to deplete an existing channel. It is normally on at $V_{GS}=0$ and pinches off as reverse gate bias grows. A depletion MOSFET also has a channel at zero bias; an enhancement MOSFET creates an inversion channel only after gate voltage crosses threshold.

For an enhancement MOSFET:

• below threshold, current is small but not identically zero;
• in the ohmic/triode region, the channel behaves approximately resistively;
• in saturation, drain current is mainly controlled by gate overdrive, enabling gain;
• beyond voltage, current, thermal, or safe-operating-area limits, the model is no longer a valid operating mode.

The word saturation is a classic trap: MOSFET saturation is the analog constant-current region, while BJT saturation is the hard-on switching region.

Threshold is not a drive voltage #

$V_{GS(th)}$ is measured at a tiny specified drain current. It says where the channel begins to form, not where the device has low on-resistance. A switch is fully characterized only at gate voltages for which $R_{DS(on)}$ is specified, including hot temperature.

In linear operation, the small-signal transconductance is

$$g_m=\frac{\partial I_D}{\partial V_{GS}}\bigg|_Q,$$

and a common-source stage converts $g_mv_{gs}$ into output voltage through the drain load. A source follower trades voltage gain for buffering, while a common-gate stage avoids much of the Miller effect.

The gate is capacitive and nonlinear #

The driver must supply approximately $Q_g f_s$ average charge current and much higher peak current during edges. The Miller plateau occurs while drain voltage changes and gate–drain charge is transferred. Switching speed therefore depends on charge, driver impedance, common-source inductance, and the commutation loop — not gate capacitance alone.

flowchart TB
  DRV["gate driver"] -->|"moves Qg"| G["gate voltage"]
  G --> CH["channel conductance"]
  CH --> DV["drain voltage/current transition"]
  DV -->|"Cgd / Miller feedback"| G
  DV --> PAR["parasitic L and C<br/>overshoot + ringing"]

The intrinsic body diode and reverse channel conduction matter in bridges. Reverse-recovery behavior, dead time, and gate off-state impedance can decide loss and shoot-through. Logic-level compatibility, negative transients, maximum gate voltage, avalanche rating, and linear-mode safe operating area are separate specifications.

Bias, temperature, and sharing #

For small-signal stages, source degeneration gives local feedback and stabilizes current. For power switching, $R_{DS(on)}$ usually rises with temperature, which can encourage current sharing at steady state, but dynamic sharing still depends on threshold, charge, layout, and cooling. Linear-mode operation is much more dangerous than a simple $I^2R$ estimate because hot spots can form inside the die.