current-mirror
A current mirror is a two-transistor circuit that copies a current from one branch of a circuit into another, forcing the output transistor to carry the same current — or a deliberately scaled multiple of it — as whatever reference current flows into the input transistor, regardless of what voltage the output branch happens to be sitting at. Built from a matched pair of transistors sharing the same gate or base voltage, a current mirror turns a single known, often precisely set reference current into as many identical or scaled copies as a circuit needs, which is why it's one of the most repeated building blocks inside any analog or mixed-signal chip. How a current mirror actually forces that copied current to match looks like this:
A few things are worth knowing about why forcing a specific current, rather than a specific voltage, turns out to be such a useful trick:
A current mirror's defining property is that its output current stays fixed regardless of the voltage across the output branch, which is precisely the opposite behavior of a simple resistor, and that voltage-independence is the entire reason mirrors are so widely used. A resistor's current changes proportionally whenever the voltage across it changes, by Ohm's law — but a current mirror, built correctly, holds its output current essentially constant across a wide range of output voltages, because the matched input transistor has already set the exact gate voltage needed to produce the reference current, and applying that same gate voltage to a matched output transistor reproduces that current almost independent of what's happening on the output side. This makes a current mirror behave like a near-ideal current source, a building block that supplies a fixed current no matter what load it's driving, which is a property plain resistors simply cannot provide.
Scaling the mirrored current to any desired multiple of the reference is as simple as changing the relative size of the output transistor compared to the input transistor, which is exactly why a single reference current can cheaply fan out into many differently sized copies across a chip. A mirror built with matched, identically sized transistors copies the reference current one-to-one, but making the output transistor twice as wide as the input transistor, while keeping its length the same, roughly doubles the mirrored output current for the same gate voltage; making it half as wide roughly halves it. Because a single precision reference current — often generated once per chip using a bandgap reference and an external resistor — can be mirrored and rescaled into dozens of differently sized bias currents feeding different analog blocks across a die, a current mirror is the standard, area-efficient way to distribute consistent, well-controlled bias currents everywhere a chip needs them without generating each one from scratch.
A basic two-transistor current mirror isn't perfectly voltage-independent in real silicon, because a real transistor's current does shift slightly with its drain voltage even once biased on — an effect called channel-length modulation — and that imperfection is exactly what drives designers toward more elaborate mirror topologies. The simple mirror's output current creeps upward slightly as the output voltage rises, a deviation from the ideal constant-current behavior that matters in precision applications. A cascode current mirror, which stacks an additional transistor on top of the basic mirror, dramatically reduces this voltage sensitivity by shielding the current-setting transistor from output voltage swings, at the cost of needing more voltage headroom to operate correctly. This tradeoff — accuracy and output impedance against the minimum voltage the circuit needs to function — is a recurring theme across current mirror design, and choosing which topology to use is a direct response to how much precision a given application actually demands.
Beyond simple biasing, a current mirror forms the active load inside nearly every operational amplifier's input stage, converting a differential pair's current signal into a usable voltage with far higher gain than a plain resistor load could provide. An op-amp's input stage typically produces a difference in current between two branches in response to the voltage difference at its two inputs; replacing a simple resistor load on those branches with a current mirror lets that current difference develop into a voltage swing with much greater gain, because the mirror presents a very high effective resistance to small signal changes while still passing the correct DC bias current. This is a large part of why real operational amplifiers achieve the enormous open-loop gains they're known for — the current mirror acting as an active load, not just the raw transistor gain, is doing much of that gain-boosting work.
Read a current mirror as a circuit that trades a known current for a known current, not a known voltage, which is exactly the opposite of what a simple resistor or voltage divider provides: that current-forcing behavior is what makes it a near-ideal current source for biasing, lets a single reference current cheaply fan out into many scaled copies across a chip, and — stacked into an active load — is a quiet, essential contributor to why real analog amplifiers achieve the gain figures their datasheets advertise.