2026-06-03
Inside nearly every general-purpose op-amp — the 741, the LM358, the TL072 — there's a single small capacitor (typically 10–30 pF) bridging the input and output of the second gain stage. This is the Miller compensation capacitor, and it's the reason your op-amp doesn't oscillate when you wrap feedback around it.
The problem. An uncompensated two-stage amplifier has two high-impedance nodes, each contributing a pole. Two poles inside the loop means up to 180° of phase shift before you even get to unity gain — guaranteed oscillation when you close the loop. You need to force the open-loop response to look like a single-pole rolloff well past the unity-gain crossover.
The Miller trick. Place capacitor Cc across the second stage (transconductance gm2, gain A2). Two beautiful things happen simultaneously:
Concrete example. The classic 741: input stage gm1 ≈ 190 µA/V, Cc = 30 pF. GBW = 190e-6 / (2π · 30e-12) ≈ 1.0 MHz — exactly the datasheet number. That tiny 30 pF cap single-handedly defines the part's speed.
The right-half-plane zero. Miller compensation has a dirty secret: Cc creates a feedforward path from input to output of stage 2, producing a zero at z = gm2/Cc in the right half-plane. RHP zeros add gain but subtract phase — exactly what you don't want near crossover. Modern op-amps kill this with a series "nulling resistor" Rz = 1/gm2, which moves the zero to infinity (or into the LHP, where it actually helps phase margin).
Rule of thumb for discrete designs: If you're building a discrete two-stage amplifier, pick Cc so the second pole sits at least 2–3× above the unity-gain frequency. Phase margin of 60° generally requires the non-dominant pole to be ~2.2× GBW.
Decompensated op-amps (like the LM6171, OPA637) use smaller Cc values for higher GBW — but they're only stable at gains ≥5 or ≥10. Use them in a voltage follower and they'll happily oscillate at tens of MHz.