The Most Expensive Diode Mistake in Power Supply Design — and How a 2-Cent Part Causes It

The Physics Problem: Reverse Recovery Is a Heater You Can't See

When a PN diode switches from forward conduction to reverse blocking, stored charge in the junction must be swept out first. During that sweep-out time — t_rr — current flows backward while voltage rises. Current times voltage equals power. That power becomes heat.

A standard 1N4007 at 100kHz: t_rr is typically 2–30µs, the switching period is 10µs. The diode can conduct in reverse for more than an entire cycle. Every time. A UF4007 at 75ns max: reverse conduction lasts 0.75% of the period. Two orders of magnitude difference.

The Runaway: Why It Works on the Bench and Fails in the Field

NASA's investigation documented the thermal runaway sequence in three phases:

• Phase 1 (benign): At 53°C, reverse recovery energy was ~36µJ per cycle. The heatsink is warm but stable.
• Phase 2 (acceleration): At 59°C — just 6 degrees higher — recovery energy hit ~60µJ, nearly double. The extra heat raises junction temperature, which increases recovery energy. The loop begins.
• Phase 3 (runaway): At 135–140°C, the feedback loop became self-sustaining. Junction temperature exceeded 150°C. The diode failed short-circuit.

The bench test never hits Phase 2. The supply runs 30 seconds at room temperature, diode stays at 45°C.

But in the field — sealed enclosure, 40°C day, six months of dust — the diode idles at 55°C. One transient away from runaway. Once it enters, it doesn't come back.

The Decision Framework: Which Diode for Which Job

Here's the 30-second checklist. If any of these conditions are true, a standard rectifier (1N4007) does not belong in your circuit.

The Repair Shop Perspective: What Actually Comes Back

From repair diagnostics on SMPS units (consumer and industrial, 2025–2026), the pattern is consistent. When an output rectifier fails short, the damage cascade is predictable:

1. The diode fails short-circuit. This connects the transformer secondary directly to the output capacitor through a near-zero-impedance path during the flyback interval.
2. The switching MOSFET sees a reflected short. Primary current spikes past the MOSFET's pulse current rating. If the controller's cycle-by-cycle current limit catches it in time, the MOSFET survives. Often it doesn't.
3. The controller IC's gate drive pin gets exposed to drain voltage. Through the gate-drain capacitance of a failing MOSFET. The controller dies silently — no visible damage, but the gate drive output no longer switches.
4. If the output capacitor is rated for the rail voltage but not the flyback spike, it vents. Now you're replacing capacitors too.

The proper repair: replace the diode with a UF4007, replace the MOSFET (even if it tests good — it was stressed), verify the controller IC gate drive on a scope, and check the snubber resistor.

The repair that guarantees a second failure: swap the diode for another 1N4007 and call it done. But the diode wasn't defective. It did exactly what a standard recovery diode does at 65kHz: conduct in reverse, heat up, and self-destruct. The mistake was in the BOM, not in the silicon.

What This Means for Your BOM

If you're sourcing components for a new SMPS design or a production run, here's what matters:

• Never substitute "close enough" rectifier diodes without verifying t_rr. The 1N4007, UF4007, 1N5819, and 1N4007GP all say "1A diode" on paper. Their reverse recovery behavior is different by a factor of 400+. If your BOM calls for a UF4007 and your CM proposes a 1N4007 to save $0.02, the answer is no.
• At 100kHz+, even "fast" isn't always fast enough. A 75ns UF4007 at 100kHz spends 0.75% of the switching period in reverse recovery. That's acceptable for most designs. At 300kHz, it's 2.25% — starting to hurt. At 500kHz, 3.75% — now you need a 30ns part or a SiC Schottky.
• Thermal margin is not optional in enclosed supplies. The NASA data shows the runaway threshold is razor-thin. If your thermal simulation says the diode runs at 110°C in a 25°C ambient, it will hit 135°C at 50°C ambient. Design for worst-case ambient plus 15°C of dust/degradation margin.
• Schottky diodes solve the recovery problem completely, but only below 40–200V. For 12V and 24V output rails, a Schottky rectifier is almost always the right choice. For 48V+ outputs, you need a silicon ultrafast or a SiC Schottky (which can handle 650V+). The voltage rating of your output rail determines which technology you can use.

Frequently Asked Questions

Q1: How do I know if a failed diode was killed by reverse recovery or something else?

A: Look at the failure mode and the circuit. A diode that failed short-circuit in an SMPS output rectifier position, especially one where the PCB shows heat discoloration around the diode body, is almost certainly reverse recovery thermal runaway. A diode that failed open-circuit or with visible arc damage may have been killed by an overvoltage transient. Measure the snubber components and the switching frequency before replacing the diode.

Q2: Can I test a diode's reverse recovery time with a multimeter?

A: No. A multimeter diode test only measures forward voltage at ~1mA. It tells you the diode is a diode. It tells you nothing about t_rr, which requires a pulse generator and an oscilloscope to measure. If you don't know the diode's t_rr spec and the circuit switches above a few kHz, you're guessing.

Q3: What's the difference between UF4007 and 1N4937?

A: Speed and voltage. UF4007 = 1000V/1A/75ns. 1N4937 = 600V/1A/200ns. The UF4007 is 2.7x faster and handles higher voltage. The 1N4937 costs about half as much and was the "fast" diode of the 1980s. For any modern SMPS above 50kHz, use the UF4007. For 60Hz bridge rectifiers where a standard 1N4007 is too slow but you don't need 75ns, the 1N4937 is a viable middle ground.

Q4: Why don't diode datasheets make the thermal runaway risk obvious?

A: Because they characterize the diode, not your circuit. A diode datasheet gives you forward voltage vs. current, reverse recovery time, and thermal resistance. It doesn't tell you that at 59°C the reverse recovery energy doubles in a flyback converter. That's not the diode manufacturer's job — it's the power supply designer's job to understand the application. The datasheet gives you the numbers. You have to know which question to ask.

Q5: Is a glass-passivated diode like 1N4007GP any better for switching applications?

A: No — it's still a standard recovery diode. The GP version specifies t_rr at 2µs and handles 175°C junction temperature, both improvements over a generic 1N4007. But 2µs is still ~400x slower than a UF4007 at 75ns. The GP is an upgrade for line-frequency rectification where you want guaranteed specs and better reliability. It is not a substitute for an ultrafast diode in a switching circuit.

Related Products

• UF4007 — Ultrafast Recovery 1000V/1A Rectifier, DO-41 (75ns)
• 1N4007GP-E3/54 — Glass Passivated 1000V/1A Rectifier, DO-41 (Vishay)
• 1N5819 — Schottky Barrier 40V/1A Rectifier, DO-41
• 1N4937 — Fast Recovery 600V/1A Rectifier, DO-41 (200ns)
• 1N4007 vs 1N4007GP vs 1N5819 — Full Comparison Guide
• Standard vs Ultrafast vs Schottky — Rectifier Diode Type Comparison