Instrumentation Amplifier
An instrumentation amplifier takes the plain differential amplifier's common-mode rejection and fixes its two worst field problems — low input impedance that loads down a delicate sensor, and gain accuracy that depends on multiple resistor ratios matching precisely — by buffering each input through its own high-impedance op-amp stage before a final difference stage sets the gain with a single resistor, which is why it is the default front end for bridge sensors, thermocouples, and any signal source too weak or too fragile to drive a plain differential pair directly.
Input buffering solves the loading problem a plain differential pair can't. A basic differential amplifier's input impedance is set by its own biasing resistors, which can be low enough to pull real current out of a delicate source like a Wheatstone bridge strain gauge or a thermocouple junction, distorting the very signal it's trying to measure; the instrumentation amplifier's two input buffer op-amps present an impedance in the hundreds of megohms to gigohms range, so the sensor sees essentially no load and the measured signal matches what the sensor actually produced.
A single resistor setting the gain is the real innovation, not just extra buffering. In a classic three-op-amp instrumentation amplifier, the two input stages share one gain-set resistor between their inverting inputs, and the differential gain works out to 1 plus twice the feedback resistor divided by that single resistor — which means gain accuracy depends on trimming or selecting one component instead of matching several resistor ratios across two separate amplifier stages, a difference that matters enormously once you're specifying a part to 0.01% gain error, which integrated instrumentation amplifiers from Analog Devices (the AD620 family) and Texas Instruments routinely hit.
CMRR gets better, not just preserved, because the architecture adds a second line of defense. The buffered front end keeps the symmetric high-CMRR behavior of the differential pair, and the final difference amplifier stage rejects common-mode content a second time before producing a single-ended output, which is why integrated instrumentation amplifiers commonly spec CMRR above 100 dB and some precision parts exceed 130 dB — numbers that matter directly in applications like ECG and EEG measurement, where the signal of interest is microvolts riding on volts of 60 Hz mains common-mode interference.
It trades flexibility for precision, which is exactly the right trade for its job. Unlike a general-purpose op-amp circuit built from discrete resistors, an integrated instrumentation amplifier usually exposes just one external gain-set resistor and a reference pin, deliberately limiting what the designer can configure in exchange for guaranteed CMRR, offset drift, and gain linearity specs that would be difficult to hit reliably with a hand-built three-op-amp circuit — which is why instrumentation amplifiers dominate industrial weigh-scale, medical, and precision data-acquisition front ends rather than being assembled from discrete parts in production designs.
Read an instrumentation amplifier through a *sensor-protection-first* lens rather than a *bigger-differential-amplifier* lens: every design choice — the buffered high-impedance inputs, the single gain-trim resistor, the second common-mode rejection stage — exists to let the circuit measure a weak, fragile, noise-ridden signal without disturbing or distorting the source it's reading, which is the job a plain differential amplifier was never precise or gentle enough to do on its own.