LNK306 vs TNY284 vs LNK625 — Power Integrations SO-8C Switcher Selection Guide

At a Glance: Three SO-8C Switchers, Three Design Philosophies

The Decision Tree: Which One Do You Need?

Start with one question. Then follow the branch.

Question 1: Does the output need to be galvanically isolated from the mains?

NO → LNK306. If your load can float at line potential — a non-isolated LED driver, TRIAC trigger, cap-drop replacement — the LNK306 buck is the answer. No transformer means smallest BOM, lowest cost, simplest layout.

The trade-off: the entire circuit sits at mains potential. Don't use it for anything with a user-accessible connector.

YES → Question 2. You need a flyback transformer. Now decide how you want to regulate the output.

Question 2: How tight does your output regulation need to be?

Tight (±1–2%) → TNY284. When the load swings 10%–100% and the rail must hold within 2%, secondary-side feedback with TL431 + optocoupler is the only way. TinySwitch-4 also delivers <30mW no-load standby. This is the reference design for USB chargers, appliance MCU rails, and anything that faces a compliance tester.

Good enough (±5%) → LNK625. If the load is predictable and ±5% works, primary-side CV eliminates the optocoupler and TL431. That's 4–6 fewer components, simpler safety cert (one less part crossing the isolation barrier), and ~$0.15–0.25 BOM savings.

The trade-off: looser regulation, higher no-load, and more care needed in PCB layout around the FB pin.

Deep Dive: The Three Decisions That Actually Matter

Decision 1: Isolation — Do You Need a Transformer?

The LNK306 buck ties the output to rectified mains. The output "ground" sits at ~160V DC above earth on 230VAC. That's fine for an LED string or TRIAC gate drive. It's not fine for a USB port.

The moment a human can touch the output, you need isolation — and that means a flyback transformer. The LNK306 saves ~$0.30–0.50 in magnetics cost. That saving stops mattering the first time someone touches the output.

Decision 2: Feedback — Optocoupler or Not?

The TNY284 uses a TL431 to compare output voltage against a 2.5V reference, and an optocoupler to send the error signal across the barrier. This is the gold standard for regulation — ±1–2% across line and load.

The LNK625 replaces all of that with a resistor divider on the bias winding, sampled 2.5µs after turn-off. It works. But it's inherently looser: bias winding voltage isn't a perfect proxy for output voltage. Leakage inductance, diode V_F variation, and PCB layout all introduce error.

Decision 3: Standby — How Low Do You Need to Go?

The TNY284 achieves <30mW no-load. The LNK625 manages <70mW. The LNK306 draws <150mW. For a wall adapter plugged in 24/7, 30mW vs 200mW is about 1.5kWh/year — negligible in dollars.

But it's the difference between passing and failing DOE Level VI or EU CoC Tier 2. If the spec says "meets all global standby regulations" with no qualifiers, that means TNY284.

Application Mapping: What Goes Where

BOM Cost Comparison (Rough Order of Magnitude)

Excluding PCB and connectors. Prices are production-volume estimates for the power stage only.

What drives the difference: The LNK306 needs no transformer and no feedback components — just an inductor, freewheeling diode, and output cap. The LNK625 adds a flyback transformer and bias winding but no optocoupler. The TNY284 adds everything: transformer, optocoupler, TL431, passives.

Each step up the BOM ladder buys you either isolation or tighter regulation. Pick the step that matches your spec. Don't pay for performance you're not measuring.

Frequently Asked Questions

Q1: Can I use one design and swap between these three by changing the IC?

A: No. Each requires a fundamentally different PCB. The LNK306 is a buck — no transformer, inductor in series with the output. The LNK625 and TNY284 are both flyback, but the feedback networks are different: LNK625 uses an FB pin with a resistor divider from the bias winding, TNY284 uses an EN/UV pin driven by an optocoupler. You're looking at three separate layouts.

Q2: If I need isolation but want the lowest cost, LNK625 or TNY284?

A: LNK625 saves $0.15–0.25 in production by eliminating the optocoupler and TL431. But those savings come with looser regulation (±5% vs ±2%) and higher no-load consumption (70mW vs 30mW). If the product spec allows ±5% and standby isn't the headline spec, LNK625 wins on cost. If the spec says "DOE Level VI" or "±2% regulation," the TNY284 optocoupler is not optional.

Q3: Can an LNK306 design be made "safe enough" without isolation?

A: For products with no user-accessible conductive parts, yes. This covers LED bulbs, sealed sensor modules, and enclosed industrial controls where the entire PCB is behind a UL-rated enclosure. Anything with an external connector, a metal chassis the user can touch, or a cable exiting the enclosure needs isolation. There is no safe way to float a non-isolated output at a user connector.

Q4: Which one is easiest to design with?

A: LNK306, by a wide margin. Only 15–20 components. Fixed 482mA current limit means you pick the inductor and output cap and it works. The LNK625 requires careful FB pin layout and transformer design. The TNY284 adds loop compensation around the TL431. Design effort scales with performance.

Q5: What about the LNK304 vs LNK306?

A: Same family, scaled-down current limit. The LNK304 (257mA) and LNK305 (375mA) are smaller-die versions of the LNK306 (482mA) in the same SO-8C package. If your LNK306 design has thermal headroom to spare, you can cost-down to LNK304/305. Same pinout, same frequency, same buck topology — just less current.

Q6: Is there a part that does isolation + no optocoupler + tight regulation?

A: Not from Power Integrations in SO-8C, not yet. Primary-side regulation is inherently less precise than secondary-side because it's inferring the output voltage through a transformer. The closest compromise is the LNK625 with careful FB layout and a well-characterized transformer, which can get you to ±3–4% in practice. For ±1–2%, the optocoupler is still the answer.