Miller Effect
The Miller effect is what happens to a small capacitor when it has the misfortune of sitting across an inverting, high-gain amplifier stage: from the input's point of view, that capacitor behaves as if it were enormously larger than its physical value, simply because the voltage swinging on the far side of it is working against the input instead of with it.
The arithmetic is simple once the sign convention is tracked correctly. A small capacitor C_f connects an amplifier's input node to its output node. If the stage has inverting gain −A, a small input voltage change δv produces an output change of −Aδv — moving the opposite direction. The voltage actually dropped across C_f is the difference between the two nodes, δv − (−Aδv) = (1+A)δv, which is far larger than δv alone. Since the charge the input must push into that capacitor scales with the voltage across it, the input node effectively sees a capacitance of (1+A)·C_f — the Miller capacitance — even though the physical component never changed value. This is Miller's theorem: a feedback impedance bridging an inverting gain stage can be replaced, for analysis, by two impedances to ground, one scaled by (1+A) at the input and one by roughly (1+1/A) at the output.
This single mechanism is usually the reason an amplifier's bandwidth is set by one specific capacitor rather than by the sum of everything in the circuit. In a common-source or common-emitter stage, the gate-drain or base-collector capacitance is physically tiny — often a femtofarad-scale parasitic — but multiplied by a voltage gain of even a few dozen, it dominates every other capacitance at the input node, including the device's own gate capacitance to the channel. The input resistance driving that node, combined with this multiplied capacitance, usually sets the single dominant pole of the whole amplifier — which is why a transistor-level simulation can show the Miller-multiplied gate-drain capacitance alone predicting bandwidth almost as accurately as the full circuit.
The same multiplication that steals bandwidth by accident is deliberately exploited to stabilize amplifiers on purpose. A two-stage operational amplifier needs its two internal poles pulled far apart in frequency to stay stable with feedback; doing that with a plain capacitor to ground at the first stage's output would require an impractically large value. Placing a small compensation capacitor across the second, inverting gain stage instead multiplies it by that stage's own gain, producing the equivalent of a much larger capacitance at the first stage's output using a component small enough to fit on-chip. This technique, standard in nearly every integrated op-amp design, turns the same physical effect that is normally a parasitic liability into the principal tool for frequency compensation — Miller compensation is not a different phenomenon from the Miller effect, it is the identical multiplication, applied on purpose instead of suffered by accident.
Cascoding defeats the parasitic version by denying the capacitor the gain it needs to multiply against, rather than by removing the capacitor itself. A gate-drain capacitance cannot be shrunk away — it is a property of the device's physical layout — but stacking a common-gate transistor on top of the gain stage holds that node's voltage nearly constant regardless of the output swing, collapsing the local voltage gain the Miller capacitance sees down to roughly one. With almost no local gain to multiply against, the same physical capacitance contributes almost nothing to the input-referred load, which is why cascode stages are the standard high-bandwidth counterpart to the simple common-source amplifier whenever Miller-limited bandwidth is the actual bottleneck.
| Scenario | Capacitor | Gain seen | Effect | Design response |
|---|---|---|---|---|
| Parasitic Miller cap | Gate-drain / base-collector | Full stage gain A | Unwanted dominant pole, limits bandwidth | Cascode to suppress local gain |
| Miller compensation | Deliberately added | Second-stage gain A | Intentional dominant pole for stability | Standard two-stage op-amp technique |
| No Miller multiplication | Any cap to ground only | None (not bridging gain) | Ordinary, unmultiplied loading | No special handling needed |
The Miller effect is a reminder that a capacitance is not a fixed number once it sits across a gain stage — it is multiplied by whatever gain the circuit happens to provide at that moment, for better or worse. The entire difference between a parasitic bandwidth killer and a textbook compensation technique is whether the designer chose the capacitor, or the device physics did.