A current mirror circuit is an active building block that copies a reference current from one transistor to another, forcing the output branch to source or sink an identical (or scaled) current regardless of the load attached to it. Instead of relying on bulky, temperature-drifting passive resistors to set bias points, a current mirror ckt actively regulates electron flow, drastically improving voltage gain, common-mode rejection ratio (CMRR), and thermal stability in both discrete designs and monolithic ICs. If you have ever tried to bias a differential amplifier pair with a simple tail resistor and watched the operating point drift as the room warmed up, you already understand the problem this circuit solves.
The Core Mechanism and a Worked Numeric Example
At its most basic, an NPN bipolar junction transistor (BJT) current mirror uses two matched transistors. Q1 is "diode-connected" (base tied to collector), forcing it to develop a base-emitter voltage (VBE) proportional to the reference current fed into it. Because Q2 shares the exact same base node and emitter ground, it experiences the identical VBE and consequently sinks the same collector current.
Think of it like a municipal water system where a master valve (Q1) sets a specific flow rate, and a slave valve (Q2) mechanically links to the master to guarantee the exact same flow rate down a parallel pipe, even if the downstream pressure (load) fluctuates.
Worked Example: Designing a 5mA Discrete Sink
Let’s design a basic current mirror to sink exactly 5mA from a load, using a 12V DC supply (VCC) and standard 2N3904 NPN transistors. We will assume a nominal VBE of 0.65V at this current level.
- Calculate the Reference Resistor (
RREF):
The current through Q1 is set by the resistor connectingVCCto Q1’s collector.
RREF = (VCC - VBE) / ITARGET
RREF = (12V - 0.65V) / 0.005A = 2,270 Ω - Select Standard Component:
We choose the nearest E24 standard value: 2.2 kΩ. - Verify Actual Reference Current:
IREF = (12V - 0.65V) / 2200 Ω = 5.16 mA - Account for the Early Effect (Output Impedance):
In the real world, Q2 is not a perfect current source. Due to base-width modulation (the Early effect), the output current will creep up slightly as the voltage across Q2’s collector-emitter (VCE) increases. For a 2N3904, the Early Voltage (VA) is roughly 100V.
ro = VA / IC = 100V / 0.00516A ≈ 19.4 kΩ
This means for every 1V increase in load voltage, your "constant" 5mA sink will increase by about1V / 19.4kΩ = 51 μA. In precision analog work, this 1% error is unacceptable, which brings us to advanced topologies.
Topology Comparison: Basic vs. Wilson vs. Cascode vs. Widlar
When a basic two-transistor mirror lacks the output impedance or headroom for your design, you upgrade the topology. Below is a data-dense comparison of the four most common BJT current mirror architectures you will encounter in analog IC design and discrete audio circuits.
| Topology | Output Impedance (Rout) |
Min Voltage Headroom | Component Count | Best Application |
|---|---|---|---|---|
| Basic 2-BJT | ro (~20kΩ) |
VCE(sat) (~0.3V) |
2 BJTs, 1 Resistor | General biasing, non-critical LED driving |
| Wilson | β · ro / 2 (~1MΩ) |
2VBE (~1.4V) |
3 BJTs, 1 Resistor | High-gain amplifier active loads |
| Cascode | gm · ro2 (~10MΩ) |
VCE(sat) + VBE (~1.0V) |
4 BJTs, 2 Resistors | Ultra-high CMRR differential pair tails |
| Widlar | High (boosted by emitter degeneration) | VCE(sat) (~0.3V) |
2 BJTs, 2 Resistors | Generating μA currents from mA references |
Where You Meet This in Practice (and Common Confusions)
You will rarely build a current mirror ckt to power a heavy load directly; instead, you use them as the invisible scaffolding inside analog circuitry. Here is where they show up on the bench:
- Op-Amp Internal Biasing: Open the datasheet schematic for an LM358 or NE555. The "tail" current source feeding the input differential pair is almost always a current mirror. This is what gives the op-amp its high CMRR.
- Audio Amplifier VAS Stages: In discrete Class-AB audio amplifiers, the Voltage Amplifier Stage (VAS) relies on a current mirror as an active collector load to maximize open-loop gain before global negative feedback is applied.
- LED Matrix Drivers: Integrated LED drivers use scaled current mirrors to ensure that every row or column of an LED matrix receives the exact same current, preventing uneven brightness across the display.
What People Commonly Confuse It With
The most frequent mistake hobbyists make is confusing a current mirror with a voltage follower (emitter follower). A voltage follower copies voltage (with a 0.65V drop) and provides current gain; its output impedance is very low. A current mirror copies current; its output impedance is intentionally very high.
Additionally, beginners often confuse a current source (which uses PNP transistors to push current from the positive rail down into a load) with a current sink (which uses NPN transistors to pull current from a load down to ground). The math is identical, but the polarity and transistor types are flipped.
Bench Realities: Thermal Runaway and the Early Effect
Simulators like LTspice will tell you your basic current mirror is perfect. Your breadboard will tell you otherwise. When building discrete current mirrors, you must manage two physical realities:
1. Thermal Coupling and Runaway
The VBE of a silicon BJT drops by about 2mV for every 1°C increase in temperature. If Q1 and Q2 are sitting on a breadboard an inch apart, and Q2 is dissipating more heat because it has a higher VCE voltage drop across it, Q2 will warm up. As Q2 warms, its VBE requirement drops, but the base node is still being forced to Q1’s higher VBE. The result? Q2 turns on harder, draws more current, gets hotter, and enters thermal runaway.
The Fix: Never use discrete, physically separated transistors for precision mirrors. Glue the flat sides of Q1 and Q2 together with thermal epoxy, or better yet, use a monolithic dual-transistor IC like the BCM847BS (Nexperia) or the classic MAT04 (Analog Devices). These packages guarantee the silicon dies are at the exact same temperature and have tightly matched β values.
2. Base Current Theft
In a basic 2-BJT mirror, the reference current (IREF) doesn’t just go into Q1’s collector; it also has to supply the base currents for both Q1 and Q2. If your transistors have a β of 100, you lose about 2% of your reference current just turning the bases on. This means IOUT will be roughly 2% lower than IREF. If you need exact 1:1 matching, you must use a Wilson topology or add a third transistor as a base-current buffer.
Frequently Asked Questions
Can I use MOSFETs instead of BJTs for a current mirror?
Yes, and in modern IC design, MOSFET mirrors are the standard. The primary advantage is that the gate draws zero steady-state DC current, completely eliminating the "base current theft" error. However, discrete MOSFETs (like the 2N7000) have much wider manufacturing tolerances in their threshold voltage (VGS(th)) than BJTs do in VBE. If building a discrete MOSFET mirror, you must hand-match the transistors with a curve tracer or buy them in matched arrays.
Why is my mirrored current drifting when I change the load resistor?
You are likely hitting the compliance voltage limit. A current mirror can only regulate current as long as the output transistor remains in the active region. If your load resistance is too high, the voltage at Q2’s collector drops below VCE(sat) (usually around 0.3V for small-signal BJTs). Once Q2 saturates, it acts like a closed switch, and the current is dictated entirely by Ohm’s law through the load, not by the mirror.
How do I scale the output current to be 3x the reference?
In a monolithic IC, you simply make the emitter area of Q2 three times larger than Q1. On the bench with discrete parts, you can place three identical transistors in parallel for Q2, sharing the same base and emitter connections. Alternatively, you can use emitter degeneration resistors (adding a small resistor to the emitter of both Q1 and Q2) and calculate the ratio based on the voltage drops, a technique thoroughly documented in advanced transistor tutorials.
