An NPN current mirror is a two-transistor circuit that copies (mirrors) the reference current flowing through the first transistor into the second transistor, providing a stable, predictable output current regardless of load voltage variations. In a real circuit or installation, this topology changes everything about how we bias amplifiers and drive loads. Instead of using bulky, heat-dissipating power resistors that waste voltage headroom and offer poor AC impedance, an NPN current mirror acts as an "active load." This vastly increases the voltage gain of differential amplifier stages and improves common-mode rejection, all while occupying a fraction of the silicon area on an integrated circuit or saving critical board space in discrete designs.
Think of it like a Y-splitter in a plumbing system where the main valve sets the flow, and the second pipe passively matches it based on identical pipe diameters. However, unlike plumbing, electrons are governed by semiconductor physics, meaning our "pipes" (transistors) must be perfectly matched to avoid significant errors.
The Core Mechanism and Component Selection
The basic NPN current mirror consists of two transistors, Q1 and Q2. Q1 is "diode-connected"—its collector is shorted to its base. This forces Q1 to operate in the active region with a fixed base-emitter voltage ($V_{BE}$) for a given reference current ($I_{REF}$). Because Q2 shares the exact same base and emitter nodes as Q1, it experiences the identical $V_{BE}$. Assuming the transistors are perfectly matched and at the same temperature, Q2 will draw the exact same collector current as Q1.
In discrete designs, finding perfectly matched transistors is the primary challenge. Even small variations in silicon doping or junction temperature will skew the mirrored current. Below is a data-dense breakdown of the parameters that dictate mirror accuracy when selecting discrete parts like the 2N3904 or BC547, or matched pairs like the MAT04.
| Parameter | Typical Discrete Range | Impact on Current Mirror Accuracy |
|---|---|---|
| Base-Emitter Voltage ($V_{BE}$) Mismatch | 1mV to 5mV | Causes direct proportional error in output current. A 2mV mismatch at room temp yields roughly an 8% current error. |
| Early Voltage ($V_A$) | 50V to 150V | Determines output impedance ($r_o = V_A / I_C$). Lower $V_A$ means output current sags or rises as the load voltage ($V_{CE}$ of Q2) changes. |
| Current Gain (Beta / $h_{FE}$) | 100 to 300 | Dictates base-current error. Low beta steals current from the reference branch, reducing the mirrored output by $2/\beta$. |
| Thermal Coefficient of $V_{BE}$ | -2.0 mV/°C | If Q2 dissipates more power and heats up more than Q1, its $V_{BE}$ drops, causing output current to rise uncontrollably (thermal runaway). |
For critical applications, engineers use monolithic matched pairs (like the Analog Devices MAT04) where all four transistors share the same silicon die, guaranteeing $V_{BE}$ matching within 50µV and perfect thermal tracking.
Worked Numeric Example: Building a 1mA Mirror
Let’s design a basic current mirror to sink exactly 1mA through a load using standard discrete 2N3904 NPN transistors. We have a 12V DC supply ($V_{CC}$).
Design Goal: $I_{OUT} = 1.0 \text{ mA}$
Assumptions: $V_{CC} = 12\text{V}$, $V_{BE} = 0.65\text{V}$ (typical for 2N3904 at 1mA), $\beta = 100$.
Step 1: Calculate the Reference Resistor ($R_{REF}$)
The reference current is set by the resistor connected from $V_{CC}$ to the diode-connected Q1. The voltage across this resistor is $V_{CC} - V_{BE}$.
$$R_{REF} = \frac{V_{CC} - V_{BE}}{I_{REF}} = \frac{12\text{V} - 0.65\text{V}}{1\text{mA}} = 11,350 \, \Omega$$
We select the nearest standard 1% E96 resistor value: 11.3 kΩ.
Recalculating the actual reference current with this resistor:
$$I_{REF} = \frac{11.35\text{V}}{11,300 \, \Omega} = 1.004 \text{ mA}$$
Step 2: Account for Base Current Error
In a basic two-transistor mirror, the reference current must supply the base currents for both Q1 and Q2. Therefore, the output current is not perfectly equal to $I_{REF}$. The exact relationship is:
$$I_{OUT} = I_{REF} \times \left( \frac{1}{1 + \frac{2}{\beta}} \right)$$
Plugging in our $\beta$ of 100:
$$I_{OUT} = 1.004 \text{ mA} \times \left( \frac{1}{1 + 0.02} \right) = 1.004 \times 0.9804 = 0.984 \text{ mA}$$
Our output is 0.984 mA, not 1.0 mA. This 1.6% error is entirely due to the base current stealing from the reference branch. If you need higher accuracy without switching to an IC, you must upgrade to a Wilson current mirror or add a third transistor (a beta-helper) to buffer the base currents.
Where You Meet This in Practice
While you might not build discrete current mirrors every day, you use them constantly inside integrated circuits and specialized modules.
- Differential Amplifier Tails: The classic LM741 op-amp uses an NPN current mirror to set the tail current of its input differential pair. This ensures the common-mode rejection ratio (CMRR) remains high, ignoring noise that appears equally on both inputs.
- Active Loads: In the gain stage of an op-amp, PNP current mirrors are used as the collector load instead of resistors. Because a transistor configured as a current source has a massive dynamic AC resistance (often >100 kΩ), it allows the amplifier to achieve massive voltage gain without requiring high DC supply voltages.
- LED and Laser Diode Driving: When driving high-power LEDs where brightness must remain strictly constant despite battery voltage sag, a current mirror (often implemented via dedicated ICs like the LM334 or AL8860) ensures the LED current doesn't drift, preventing color shift and thermal destruction.
- Biasing Networks: In RF amplifiers and audio power amps, current mirrors distribute a single master bias current to multiple gain stages, ensuring all stages track perfectly as ambient temperature changes.
Common Confusions and Failure Modes
When troubleshooting or designing with current mirrors, engineers frequently fall into a few specific traps. Understanding the difference between a mirror and a source, and respecting the Early effect, is critical for a working circuit.
Current Mirror vs. Current Source
People commonly confuse a current mirror with a standalone current source. A current source generates a fixed current from a voltage reference (like a Zener diode and a BJT, or a bandgap reference IC). A current mirror does not generate current; it merely copies an existing current. A mirror always requires an external reference current to function. In practice, you use a current source to feed the reference leg of a current mirror.
The Early Effect (Output Impedance Limitations)
A basic NPN mirror assumes that collector current depends only on $V_{BE}$. In reality, the effective width of the base region changes slightly with the collector-base reverse bias voltage. This is the Early effect. As the voltage across your load changes, the $V_{CE}$ of Q2 changes, causing $I_{OUT}$ to drift slightly. According to All About Circuits, the output impedance of the mirror is roughly $r_o = V_A / I_C$. If your load voltage swings by several volts, expect the mirrored current to vary by 1% to 5% unless you use a cascode mirror topology to shield Q2 from voltage swings.
Thermal Runaway in Discrete Designs
If Q2 is driving a heavy load, it will dissipate more power ($P = V_{CE} \times I_C$) than the diode-connected Q1. This heats up Q2's silicon junction. Because $V_{BE}$ has a negative temperature coefficient of roughly -2.0 mV/°C, a hotter Q2 will draw more current than Q1 for the same base voltage. This creates a positive feedback loop: more current causes more heat, which causes more current, ending in the thermal destruction of Q2. Fix: Always place a small emitter degeneration resistor (e.g., 10Ω to 47Ω) on both Q1 and Q2 to introduce negative feedback and stabilize the thermal operating point.
Frequently Asked Questions
Can I use MOSFETs instead of BJTs for a current mirror?
Yes, MOSFET current mirrors are actually more common in modern CMOS ICs because they have virtually zero gate current, eliminating the base-current error inherent in BJTs. However, discrete MOSFETs have much worse threshold voltage ($V_{GS(th)}$) matching than discrete BJTs have $V_{BE}$ matching, making discrete MOSFET mirrors less accurate without trimming.
What is a Wilson current mirror?
A Wilson current mirror adds a third transistor to the basic two-transistor topology. This configuration drastically reduces the base-current error (making the output current much closer to the reference current) and significantly increases the output impedance, minimizing Early effect errors. It is the standard upgrade when a basic mirror isn't accurate enough.
Why does my current mirror oscillate?
High-frequency oscillation in a current mirror usually stems from parasitic capacitance at the high-impedance base node of Q1 and Q2, combined with the Miller effect. If your mirror is driving a capacitive load or operating at RF frequencies, place a small compensation capacitor (10pF - 100pF) across the base-collector junction of Q1, or use a cascode configuration to isolate the base node from the output.
