2026-09-07
Every op-amp you've used so far is a voltage-feedback amplifier (VFA): both inputs are high-impedance, and gain-bandwidth product is roughly constant. Crank up the closed-loop gain, and bandwidth collapses proportionally. The current-feedback amplifier (CFA) breaks that rule — bandwidth stays nearly independent of gain, which is why every high-speed video op-amp (AD8009, LT1223, THS3491) is a CFA under the hood.
The architectural twist: the inverting input is low-impedance — it's the emitter of a unity-gain buffer driven by the non-inverting input. The error signal isn't a voltage difference; it's the current flowing into the inverting node. That current gets mirrored to a high-impedance gain node with transimpedance Z(s), then buffered to the output. So Vout = Z(s) · Ierr.
Why bandwidth doesn't shrink with gain: in a VFA, closed-loop bandwidth ≈ GBW / gain. In a CFA, bandwidth is set almost entirely by the feedback resistor Rf, not the gain resistor Rg. The manufacturer picks an optimum Rf (typically 500 Ω–1.5 kΩ) that gives flat response. You then set gain with Rg: Gain = 1 + Rf/Rg, same formula as non-inverting VFA. But changing Rg barely touches bandwidth.
Real-world example: the AD8009 is spec'd at 1 GHz small-signal bandwidth at G = +2 with Rf = 1 kΩ. Crank the gain to +10 and you still get ~700 MHz. Try that with a VFA of the same GBW and you'd drop to 200 MHz. This is why CFAs dominate composite video line drivers, HDMI cable equalizers, and pulse amplifiers for oscilloscope front ends.
The rules you must not break:
Rule of thumb: if the manufacturer specifies Rf = 750 Ω and you want G = +5, then Rg = Rf/(G−1) = 750/4 ≈ 187 Ω. Round to 180 Ω and you're done — bandwidth stays within 10% of the datasheet number.