Wilson Current Mirror

A Wilson current mirror reaches the same high output impedance a cascode current mirror achieves, but through active feedback correction instead of passive voltage shielding — a third transistor closes a local feedback loop around the basic two-transistor mirror that actively holds the input-side collector voltage constant, which not only pushes output impedance up by roughly the same gm-times-ro factor a cascode gets, but as a side effect also cancels the base-current mismatch error that limits a plain mirror's absolute current-matching accuracy, a benefit cascoding alone doesn't provide.

Active correction, not passive shielding, is the real mechanical difference from a cascode mirror, and it's easy to conflate the two because the output impedance numbers land in a similar range. A cascode current mirror improves output impedance by stacking a transistor that simply presents a low impedance at its own emitter, which happens to sit at the sensitive node and blocks output voltage changes from reaching it passively, the same mechanism covered in this site's cascode-amplifier article; a Wilson mirror instead creates a genuine local feedback loop — the third transistor senses a disturbance and actively drives a correcting signal back to the reference transistor's collector — so the improvement comes from negative feedback doing real correcting work, not from one node simply being harder to disturb in the first place.

Canceling base-current mismatch is the Wilson topology's distinguishing bonus, and it comes directly from how the feedback loop is wired. In a plain two-transistor mirror, each transistor's base current is stolen from the collector current it's supposed to be mirroring, which introduces a small but real systematic error between input and output current that scales with 1/beta; the Wilson mirror's feedback path routes through a bias arrangement that makes the base currents of the two main mirror transistors cancel to first order rather than simply accepting the error, which is a genuine accuracy improvement a cascode mirror's purely-impedance-focused topology doesn't provide on its own — this is specifically why precision analog designers reach for Wilson mirrors when absolute current-matching accuracy matters as much as output impedance.

The headroom cost is comparable to a cascode mirror's, so the choice between them comes down to the accuracy bonus, not a free upgrade. Like a cascode mirror, a Wilson mirror stacks an extra transistor's worth of voltage drop into the signal path, so it needs roughly the same additional supply headroom a cascode design does — there's no free lunch where Wilson gets cascode-level output impedance plus better accuracy for zero additional cost, which is why the real engineering decision between the two topologies usually comes down to whether the specific application's error budget is dominated by output-impedance-driven current droop or by base-current-driven mismatch, since each topology addresses a different one of those two error sources most directly.

It remains a standard building block in bipolar precision analog design specifically because bipolar transistors have a base-current problem that MOSFETs don't. A MOSFET's gate draws essentially zero DC current, so the base-current-cancellation benefit that makes the Wilson mirror distinctive in bipolar design simply doesn't apply to a MOSFET-only implementation — Wilson current mirrors appear overwhelmingly in bipolar and BiCMOS precision references, bandgap circuits, and instrumentation front ends, while pure CMOS designs reach for cascode current mirrors instead since there's no base-current error to cancel in the first place, which is a clean illustration of how a circuit topology's value depends on which parasitic effect the underlying device technology actually has.

Read a Wilson current mirror through a *feedback-does-the-work* lens rather than a *fancier-cascode* lens: it reaches similar output impedance to a cascode mirror by actually correcting a disturbance through a local feedback loop rather than merely shielding a sensitive node from it, and that feedback mechanism happens to cancel a second, completely separate error — base-current mismatch — purely as a byproduct of how the loop is wired, which is why the topology stays a bipolar-design staple even where pure impedance improvement alone wouldn't justify the extra transistor.