Qualifying a Second-Source Power Device

Key points

• A matched-condition comparison kills half the candidate parts before you spend money on samples
• Board-level verification on 20–30 units catches what a datasheet cannot: switching behaviour in your layout
• Stress the failure mode your application has, not a generic qualification battery
• A PCN agreement is part of the technical qualification, not paperwork

Why second-sourcing goes wrong

The usual story: purchasing finds a part 15% cheaper, engineering compares a datasheet, the numbers look close, a sample works on the bench, and the change is approved. Six months later a small percentage of units fail at high ambient, or conducted emissions fail on a later production run, and nobody connects it to the change.

The reason is almost always the same. The comparison was made between two datasheets rather than between two devices in the actual circuit, under the actual stress.

Step 1 — Define the envelope before looking at candidates

Write down what the device actually sees. Not the maximum ratings, the operating values:

Parameter	Your value	Notes
Bus voltage, max and min	375 V / 300 V	At highest and lowest line
Peak drain voltage	640 V measured	Includes ringing at full load, cold start
RMS current	1.2 A	At 40 °C ambient, worst case
Switching frequency	65 kHz	Quasi-resonant, valley switching
Gate drive	12 V, 10 Ω turn-on, 4.7 Ω turn-off	From the actual driver
Ambient / case	60 °C / 92 °C measured	Inside the sealed enclosure
Duty profile	8 hours/day, 300 days/year	Thermal cycling implication

With this written down, half the candidate list eliminates itself: the wrong voltage class, the wrong gate drive, the wrong package thermal path.

Step 2 — Compare at matched conditions

Only now open datasheets. Compare each candidate against the incumbent using the same conditions:

• RDS(on) at your gate voltage and at 125 °C, maximum not typical.
• Qg at your drive voltage, because drive loss and switching speed both follow it.
• Coss / Eoss at your bus voltage, not the small-signal value.
• VGS(th) range, not typical — spread matters in synchronous designs.
• Trr and Qrr at your junction temperature, with test conditions stated.
• RθJC and the package thermal path, because a better die in a worse package is a worse part.
• Avalanche or SOA, if your topology has inductive events.

Anything that fails here does not get a sample. This step costs an afternoon and saves weeks.

Step 3 — Verify on a batch, on your board

Bench-test a single unit and you learn almost nothing. Run 20–30 units from two or more lots and measure:

1. Switching waveform at full load, high line, cold start — the worst-case overvoltage condition.
2. Case temperature at rated load after 30 minutes soak, in the enclosure if possible.
3. Efficiency at 25%, 50%, 100% load, because a part that wins at full load can lose at light load.
4. Conducted emissions, because faster edges can cost 3–6 dB of margin.
5. Dead-time behaviour in synchronous designs, because VGS(th) spread changes it.

The batch matters because production spread is the thing a datasheet cannot show. Two lots from the same supplier can differ by 0.4 V in threshold, which is enough to cause shoot-through in a marginal design.

Step 4 — Stress the failure mode you actually have

Generic qualification batteries waste money. Choose the stress that matches your application:

Your application	The stress that matters	The test that predicts it
Sealed adapter, indoor	Thermal cycling from daily on/off	TCT, 500+ cycles
Outdoor lighting, telecom	Humidity with bias	H3TRB, 85/85 with bias
Solar, storage	20-year life, wide temperature	H3TRB plus 175 °C storage
Welding, motor drive	High di/dt, repetitive surge	IFSM verification, soft recovery check
Automotive-adjacent	Qualification beyond our scope	AEC-Q101 — a different supplier, not a different test

If the candidate’s existing reports cover that stress for the exact ordering code, use them. If they cover a different package or a different die revision, they do not apply — and a supplier who waves a report for “the family” is telling you something about their document control.

Step 5 — Document the decision

A written record turns a one-off decision into something repeatable. Keep it short:

• The envelope from step 1.
• The candidate comparison from step 2, with the conditions stated.
• Batch results from step 3, with lot numbers.
• Stress evidence from step 4, with report references.
• The residual risks, stated plainly — for example, “lower avalanche margin; topology has no unclamped events, so accepted”.

The residual-risk section is what makes the document useful in two years, when someone asks why this part was chosen and whether the reasoning still holds.

Step 6 — Lock it in with a PCN agreement

Qualification is worthless if the part changes underneath you. Ask for:

• Written commitment to notify before die source, mould compound, lead frame or assembly site changes.
• The qualification evidence that would accompany such a change.
• Lot traceability from foundry wafer to shipping carton, so a field return can be traced.

This is a commercial document, but it belongs in the technical qualification file, because a silent die change is the failure mode that no amount of testing prevents.

What this process is not

It is not a substitute for your own type testing of the finished product. It is not a route to automotive qualification, which requires a supplier with IATF 16949 and AEC-Q101 evidence. And it does not make a marginal design safe — if the incumbent part was running at 145 °C junction, the second source will run at 145 °C too, and both will fail eventually.

If you are qualifying one of our parts, the low voltage and high voltage product pages list the ordering codes, and the traceability page describes what the lot record contains. Send the envelope from step 1 with your enquiry and the comparison starts from your conditions rather than ours.

This article describes general manufacturing practice for machined and fabricated components. It is not a substitute for a drawing-specific engineering review — tolerances, material selection and process capability depend on the individual part.