differential amplifier
A differential amplifier takes two input voltages and amplifies only their difference, rejecting whatever both inputs share in common — a long-tailed pair of matched transistors sharing a single tail current splits that current between the two sides in proportion to the input difference, which is why this one topology sits at the core of the op-amp, the instrumentation amplifier, and nearly every analog sensor front end that has to pull a real signal out of a noisy environment.
The tail current source is what makes rejection possible, not the matched transistors alone. If both transistors shift up or down together by the same common-mode voltage, an ideal tail current source holds the total current constant regardless of that shared shift, so the split between the two legs doesn't change and no common-mode signal reaches the output; this is why a differential pair's common-mode rejection ratio is fundamentally limited by how close that tail source comes to an ideal infinite-impedance current source, not by how well the two transistors are matched, though matching still controls offset and gain accuracy.
Device mismatch, not circuit topology, sets the real-world CMRR ceiling. Two transistors on the same silicon die track each other within a fraction of a millivolt in base-emitter or gate-threshold voltage, which is why integrated differential pairs achieve CMRR figures of 80-120+ dB that discrete designs built from separately-binned parts rarely match; precision instrumentation amplifiers from companies like Analog Devices and Texas Instruments spend most of their design effort on exactly this matching problem — laser-trimmed resistor networks, chopper stabilization, or auto-zeroing circuits that null out whatever mismatch remains after fabrication.
It's the universal front end because real-world signals always ride on unwanted common-mode noise. A thermocouple, a strain gauge bridge, a long cable carrying a sensor signal, or two chips on a noisy PCB sharing a slightly different ground reference all produce a wanted differential signal sitting on top of an unwanted common-mode offset or noise floor, so putting a differential amplifier at the very first stage — before any single-ended gain — rejects that common-mode content before it can be amplified alongside the real signal, which is the entire justification for using a differential pair rather than two separate single-ended stages subtracted afterward.
Op-amps are built around this stage, which is why op-amp behavior inherits differential-pair limitations. The input stage of essentially every operational amplifier is a differential pair, and quirks that seem like op-amp quirks — input bias current, input offset voltage, finite CMRR, and the input common-mode voltage range the op-amp can tolerate without distortion — all trace directly back to the physical limits of that differential pair, so understanding the differential amplifier is the fastest way to understand why op-amp datasheets specify the parameters they do.
Read a differential amplifier through a *subtraction-before-amplification* lens rather than a *two-input-amplifier* lens: the circuit's job is to cancel whatever the two inputs share before any gain is applied, and every downstream analog building block that needs to pull a weak, noisy signal out of a shared-ground or shared-environment mess — op-amps, instrumentation amplifiers, ADC front ends — inherits that cancellation for free by starting from this stage.