ICMASS.COM - Global Electronic Component Solutions Provider

Language

Home > Blog > Blog

When to Use a Darlington Pair — and When a Single Transistor Is Better

2026/7/23 12:03:55

A Darlington pair gives you gain on top of gain - two transistors stacked so the first one drives the second, multiplying their betas. β = 100 × 100 = 10,000. Sounds like free lunch. It's not. You pay for that gain with doubled base-emitter voltage, higher saturation voltage, and painfully slow turn-off.

Most circuits that use a Darlington would work fine with a single transistor. Some would work better. Here's how to tell the difference - and the three questions that decide which one belongs in your design.

The Tradeoff at a Glance

Single 2N3904 vs Darlington MPSA13 vs Logic-Level MOSFET (2N7000) - driving a 100 mA load from a 3.3V GPIO

Current Gain (β)

2N3904: ~100

MPSA13: ≥5,000

2N7000: ∞ (V-driven)

Base/Gate Drive Needed

BJT: 1 mA base for 100 mA load

Darl: 20 µA base for 100 mA load

MOS: 0 A DC (gate capacitance)

Turn-On Voltage

BJT: 0.7V

Darl: 1.4V (doubled)

MOS: 2.1V (Vgs(th) typ)

Saturation Voltage (Vce sat)

BJT: ~0.2V

Darl: ~0.9V (can't saturate)

MOS: ~0.05V (Rds(on)×Id)

Turn-Off Speed

BJT: ~175 ns

Darl: ~1–10 µs (sticky!)

MOS: ~20 ns (with gate driver)

Power Dissipation at 100 mA

BJT: ~20 mW

Darl: ~90 mW (4.5× more)

MOS: ~5 mW (5Ω × 0.1²)

Single BJT (2N3904) Darlington (MPSA13) MOSFET (2N7000)

Why the Darlington Can't Saturate

This is the fundamental physics limitation that every Darlington design lives with.

In a single NPN transistor, saturation happens when V_CE drops to ~0.2V - the collector-base junction becomes forward-biased, and the device is as "on" as it gets. In a Darlington, the output transistor's collector-emitter voltage is: V_CE2 = V_CE1 + V_BE2. The first transistor can saturate (V_CE1 ≈ 0.2V), but the second transistor's base-emitter junction always drops ~0.7V. So V_CE2 can never go below ~0.9V.

What this means in practice: at 100 mA load current, a single 2N3904 dissipates ~20 mW (0.2V × 0.1A). An MPSA13 Darlington dissipates ~90 mW (0.9V × 0.1A) - more than quadruple. At 500 mA, the difference is 100 mW vs 450 mW. The Darlington needs a heatsink where the single transistor doesn't.

The Turn-Off Problem (And How to Fix It)

The Darlington's dirty secret: it turns on fast, but turns off slow. When you pull the base low, the first transistor cuts off immediately. But the second transistor's base is full of stored charge with nowhere to go - the first transistor is now an open circuit. The charge recombines slowly, and the output transistor keeps conducting for microseconds after you told it to stop.

_BE R_BE = The Fix Without R_BE: Q2 base charge has no path out → turn-off takes 1–10 µs With R_BE (100Ω–1kΩ): charge bleeds out fast → turn-off drops to ~200–500 ns Cost: slightly reduced input impedance (R_BE in parallel with Q1's load) The Turn-Off Resistor: Mandatory for Any Switching Darlington

The fix is simple and mandatory: add a resistor (R_BE, typically 100Ω to 1kΩ) between the output transistor's base and emitter. This gives stored charge a discharge path. Without it, a Darlington switching at even a few kHz will spend more time turning off than on - the output never fully cuts off between pulses, and the transistor cooks itself. The cost is slightly reduced input impedance, since R_BE sits in parallel with Q1's effective load. In switching applications, this is almost always an acceptable trade.

Integrated Darlington arrays like the ULN2003A include these base-emitter resistors on-die. If you're building a discrete Darlington from two separate transistors, you must add R_BE yourself.

The Three Questions That Decide

Where Darlingtons Actually Shine

Darlingtons aren't bad parts - they're specialized parts. Here's where they're genuinely the right answer:

Touch sensors and high-impedance detectors: when your signal source can deliver microamps, a Darlington's β ≥ 5,000 means 20 µA of base current switches 100 mA. A single BJT would need 1 mA - 50× more. Capacitive touch circuits, rain sensors, and photodiode amplifiers are classic Darlington territory.
Long cable runs: if your switch is at the end of 50 meters of wire picking up noise, the tiny base current of a Darlington means you can use aggressive filtering (large RC) on the input without voltage-divider losses eating your signal.
Darlington arrays (ULN2003A, ULN2803A): when you're driving seven relays from a microcontroller, a single DIP-16 package gives you seven Darlingtons with base resistors and flyback diodes already integrated. Done. The 1.4V V_CE(sat) doesn't matter at 12V relay voltages, and the single-package convenience outweighs the efficiency loss.
Audio emitter-followers: the high input impedance of a Darlington buffer stage can preserve the frequency response of a high-impedance source (guitar pickup, phono cartridge) better than a single BJT. But be careful - the extra phase shift can cause oscillation in feedback amplifier designs not originally compensated for Darlington output stages.

Where Darlingtons Don't Belong

Battery-powered low-voltage circuits (under 5V): lose 1.4V just turning on, lose another 0.9V across the saturated switch. On a 3.3V rail, that leaves 1.0V for your load. Use a logic-level MOSFET instead.
PWM above a few kHz: the slow turn-off means the transistor spends a significant fraction of each cycle in the linear region, dissipating heat. At 50 kHz PWM, a Darlington can spend 20% of each cycle partially on. Use a MOSFET with a proper gate driver.
Voltage amplification stages: the Darlington's lower transconductance means you get less voltage gain per milliampere of collector current than a single BJT. For a common-emitter gain stage, a single 2N3904 with proper biasing outperforms an MPSA13 for voltage swing.
Any circuit where you haven't checked the thermal math: P_dissipation = V_CE(sat) × I_C. At 0.9V and 500 mA, that's 450 mW. The TO-92 MPSA13 is rated for 625 mW at 25°C ambient - you're at 72% of maximum with zero margin. Add a heatsink or switch to a MOSFET.

Discrete vs Integrated: Build Your Own or Buy One?

You can make a Darlington from two 2N3904s in about 30 seconds. It works. But monolithic Darlingtons like the MPSA13 (onsemi) or BC517 are designed so the first transistor's collector current isn't starved at low total currents - a problem with discrete pairs where the first transistor operates in a low-β region and drags down the compound gain. For prototyping, discrete is fine. For production, use the integrated part unless you have a specific reason not to.

And if you're using an array like the ULN2003A, you're already using seven integrated Darlingtons with base resistors and flyback diodes. Don't overthink it - just use the array.