TL431 Pinout, Specifications & Common Mistakes — the 2.495V Shunt Regulator Explained

What Are the TL431 Pin Functions in Each Package?

The trap is the SOT-23: two different pin conventions exist in the wild. The U-1 style (Vishay, GS431) puts cathode on pin 1 and REF on pin 3; the U-2 style (Diodes Inc.) swaps them — REF on pin 1, cathode on pin 3.

Install the wrong convention and you get a short, not a regulator.

TO-92 (flat face forward): pin 1 REF, pin 2 ANODE, pin 3 CATHODE. SO-8: pin 1 cathode, pin 8 REF, four anodes tied together on the die flag for heat, pins 4/5 NC.

What Are the Key TL431 Specifications?

The tolerance ladder: the same die, three grades. For a 5V rail regulated to ±1%, the TL431B's 0.5% reference is the difference between hitting spec and chasing it — and it costs pennies more.

And the circuit that sets output voltage: VOUT = 2.495V × (1 + R1/R2). R1 runs from output to REF, R2 from REF to ground.

Keep the divider current at least 100× IREF — a few hundred microamps — so the divider dominates the 2µA reference current and its temperature drift.

How Does the TL431 Actually Work?

It's a comparator driving an NPN, nothing more. When REF sits below 2.495V the transistor stays off and the cathode floats; when REF crosses the reference, the transistor pulls the cathode toward the anode, sinking whatever the external circuit feeds it.

Regulation happens because the divider keeps trying to push REF across that threshold.

What Are the Most Common TL431 Mistakes?

❌ Mistake 1 — Wrong SOT-23 pinout. Two conventions exist (U-1: K/REF/A; U-2: REF/K/A). A board drawn against the wrong one installs a short. Check the manufacturer's footprint drawing, not a generic SOT-23 library part.

❌ Mistake 2 — Starving the cathode current. Below about 1mA (250µA on some variants) the TL431 leaves regulation. An opto LED that only wants 100µA still needs its bias resistor sized for the knee.

❌ Mistake 3 — Gigohm dividers in "ultra-low-power" designs. When R1/R2 get high enough that IREF's 2µA and its drift matter, output accuracy and noise both suffer. Rule of thumb: divider current ≥ 100× IREF.

❌ Mistake 4 — Treating SOT-23 as a 36V/100mA part. The spec table says 36V and 100mA — the package says otherwise: ~625°C/W means a few hundred milliwatts continuous. Move the heat to an external transistor or use TO-92/SOP.

❌ Mistake 5 — Ignoring the loop. In flyback feedback, the optocoupler adds a pole (PC817-class: a few kHz), CTR rolls off with frequency, and the output cap's ESR zero needs a Type II network.

Uncompensated loops ring, sawtooth, or oscillate — design the compensation, don't add it after the fact.

Frequently Asked Questions About the TL431

Q1: How do I set the TL431 output voltage?

A: VOUT = 2.495V × (1 + R1/R2), with R1 from output to REF and R2 from REF to ground. Keep the divider current at least 100× the 2µA reference current so IREF and its drift don't degrade the set point.

Q2: What is the minimum cathode current for regulation?

A: About 1mA typical; some variants regulate down to 250µA. Below the knee the TL431 leaves regulation entirely. In opto feedback circuits, size the LED bias resistor so the cathode always sees at least this current — even at no load.

Q3: Why does my TL431 feedback loop oscillate?

A: The optocoupler is usually the culprit — it adds a low-frequency pole (a few kHz) and its CTR rolls off with frequency. TI's application note computes an 8.12kHz opto pole in its example. Design a Type II compensation network around the TL431 and verify phase margin >45° with a Bode plot, not SPICE alone.

Q4: Why is the optocoupler slower than the datasheet says?

A: Datasheet speed numbers assume a 100Ω load; real circuits often use 33kΩ pull-ups, which discharge the coupler's output capacitance far more slowly. One Stack Exchange case rang at 1.3MHz despite a stable simulation — the fix was lowering the pull-up to get milliamps through the coupler.

Q5: Can I replace a zener with a TL431?

A: Yes — and the accuracy is the point: ±0.5–1.6% against a zener's 5% or worse. But it's a shunt — it needs an external resistor or current source to supply what it sinks, and it can't regulate on its own.

Q6: Which SOT-23 pinout is my TL431?

A: Check the manufacturer's drawing — two conventions exist. U-1 style (Vishay, GS431) is K=1, REF=3, A=2. U-2 style (Diodes Inc.) is REF=1, K=3, A=2. Installing against the wrong one is a short, not a tolerance problem.

Q7: Why is there sawtooth ripple at no load?

A: Many flyback loops are designed for loaded operation and their compensation runs out of margin at no load. TI's forum guidance: adjust the TL431 compensation network and verify with a Bode measurement; the no-load case is often the worst operating point for phase margin.

Q8: How much power can a SOT-23 TL431 dissipate?

A: Only a few hundred milliwatts continuous — RθJA is about 625°C/W. The 36V/100mA spec-row is theoretical (3.35W worst case); the package will not. Use TO-92/SOP or an external pass transistor for real power.

Q9: What tolerance grade do I need?

A: TL431B (±0.5%) if the rail spec is tight, TL431A (±1.0%) for most designs, plain TL431 (±1.6%) where the divider tolerance already dominates. The die is the same; the grade is a test bin, and the price difference is small.

Q10: Does the TL431 work as a comparator?

A: Yes — it IS a comparator (plus an NPN), and that's a classic trick. Feed a signal to REF and the cathode snaps between off and on around 2.495V. Just remember the anode-cathode path can only sink, and it needs a pull-up to make a logic output.