Alternate Electronic Component Qualification Checklist

Before you start: scope and tier the evaluation

Two decisions made in the first ten minutes determine how much of this checklist actually applies: how far the candidate is from the original, and how much the circuit matters.

Substitution distance is easiest to reason about as tiers, because each tier adds a category of work rather than just more of the same work:

• +Tier 1 — same manufacturer part number, differing only in packaging, marking, or date code. A handling and paperwork check.
• +Tier 2 — same manufacturer and family, different orderable part number (tolerance, grade, temperature range, or option). A parameter-by-parameter datasheet delta.
• +Tier 3 — different manufacturer, claimed pin- and function-compatible. Everything in this checklist applies, because nothing is guaranteed by a shared package name.
• +Tier 4 — different architecture requiring a circuit, layout, or firmware change. This is a design change with its own change-control path; the checklist is an input to it, not a replacement for it.

Electrical compatibility

Compare parameter tables, not summary paragraphs. Datasheet front pages are written to make parts look interchangeable; the tables are where they stop being interchangeable.

Supply and rails: absolute maximum and recommended operating voltage on every rail, minimum operating voltage against your worst-case brownout, separate analog and I/O rails, any sequencing requirement the candidate adds, and UVLO and power-on-reset thresholds with their hysteresis. A candidate whose UVLO sits 200 mV higher than the original will pass on the bench and fail on a sagging battery.

Current and power: quiescent, active, peak, inrush, and shutdown current at your operating point. Higher quiescent current matters on a battery product; higher inrush matters behind a hot-swap controller or a soft-start. Recompute worst-case dissipation yourself, and for regulators check dropout at maximum load across temperature rather than the typical figure.

Tolerance and accuracy: initial tolerance, temperature coefficient, drift over life, and line and load regulation. Passives hide the largest surprises here — a ceramic capacitor of the same value, case size, and voltage rating can lose a large fraction of its capacitance under DC bias, and an X5R substituted for an X7R changes behavior across temperature even though the part looks identical on the BOM line.

Timing and frequency: propagation delay, setup and hold, rise and fall times, channel-to-channel skew, bandwidth and gain-bandwidth product, switching frequency, and crystal load capacitance and drive level with startup margin. A faster edge rate on an otherwise-identical logic part is a real EMI change, and a switching regulator whose frequency has moved can put a harmonic into a band you are already certified against.

Drive and leakage: output sourcing and sinking current at the logic thresholds you rely on, VIH and VIL, push-pull versus open-drain outputs, internal pull-ups, input leakage into high-impedance nodes, ESD-diode leakage, capacitor insulation resistance, MOSFET off-state leakage, and diode reverse leakage. Leakage rises steeply with junction temperature, so a divider or sample-and-hold that is accurate at room temperature can drift out of spec at the top of the range. Finish with the protection ratings: ESD (HBM and CDM), latch-up immunity, reverse voltage, and any internal clamp the circuit quietly depends on.

Functional compatibility

Functional compatibility is about what the part does, and specifically about what your design uses it to do. Compare against your actual usage rather than the candidate's feature list — a candidate that is a functional superset is still a change, because a pin or a register that was inert on the original may now do something.

Check states and behavior: power-on default configuration, whether the part comes up enabled or disabled, fault handling (latching versus auto-retry), and the thresholds on protection features such as overcurrent, overvoltage, and thermal shutdown. A part that auto-retries where the original latched will behave differently in a fault the field will eventually produce.

Check interfaces in detail: protocol version, I2C address and whether it is fixed or strapped, SPI mode (CPOL and CPHA), maximum bus speed, and clock-stretching support. For data converters, compare reference tolerance, resolution, INL and DNL, and noise, since these determine whether calibration constants still hold.

The most dangerous category is dependence on unspecified behavior: an ESD diode used as a clamp, a specific power-up ramp, an observed-but-undocumented startup delay. Nothing obliges the candidate to reproduce behavior that was never in the specification, so identify these dependencies and test them explicitly.

Finally, confirm your existing test coverage still applies. In-circuit, flying-probe, boundary-scan, and functional tests were written against the original; verify they still pass and that they actually exercise the parameters that differ.

Package and footprint

A shared package name is not a shared package. Two QFN-24 4 x 4 mm parts can differ in exposed-pad size, terminal length, pitch tolerance, and standoff. Compare mechanical drawings dimension by dimension, then compare the candidate's recommended land pattern against the footprint you already have: pad length and width, pitch, exposed-pad geometry and via pattern, solder-mask-defined versus non-solder-mask-defined pads, and courtyard.

Physical fit extends past the board. Check maximum Z height under shields, enclosures, and in stacked assemblies, and check clearance to adjacent components and connector mating envelopes.

Terminations and process: lead finish and plating, any pure-tin restriction that applies to your program (tin whiskers remain a real constraint for long-life and high-reliability hardware), moisture sensitivity level and the floor-life and bake implications that follow, peak reflow temperature and profile compatibility, package warpage, and whether stencil aperture or paste volume needs to change. Suitability for wave or selective solder is a separate question from reflow.

Handling and packaging cause more line stoppages than physics does. Confirm carrier type (tape and reel, tube, tray), reel size and quantity, and pin-1 orientation within the carrier before the first build, not on the day of it. Where a bottom-terminated package is involved, confirm the inspection method you rely on — X-ray availability, rework capability — still works for the candidate.

Pinout and orientation

Pin-for-pin compatibility is claimed far more often than it is true, and the failures are quiet. Compare the full pin table rather than the signal names, and pay particular attention to pins marked NC, DNC, or reserved on either side. A no-connect on the original that is a test, program, or supply pin on the candidate is a latent failure that a functional test will not necessarily catch.

Confirm the mechanical conventions: pin-1 index feature and numbering direction, and the exposed pad's electrical connection — grounded, floating, or internally tied to a rail. An exposed pad that is internally connected to a rail on the candidate and to nothing on the original is a short waiting for a copper pour.

For polarized passives and connectors, verify marking conventions against the mechanical drawing rather than the schematic symbol. Mirrored and rotated variants exist within several MOSFET and connector families, and the symbol in your library reflects the original part, not the candidate.

Check strapping and configuration pins: default states, required pull-ups or pull-downs, and whether default addresses match. Then draw the conclusion honestly — if the pinout differs at all, this is a layout change. Route it through design change control with a prototype and a test cycle, and record it as a design revision rather than a part swap.

Firmware and software implications

This group applies to anything programmable or register-configured: microcontrollers, PMICs, sensors, transceivers, PHYs, memory, and FPGAs. Where it applies, it is usually the group that determines the schedule.

Register map: compare address by address, not name by name. Reset and default values, reserved bits (writing a value the candidate does not expect to a reserved bit is a classic silent difference), bit-field widths, and read-modify-write semantics all matter. Then check the identification path — device ID, part ID, and silicon-revision registers that drivers and production test read to decide what they are talking to.

Boot behavior: strapping pins, boot ROM version, boot time, reset and brown-out behavior, and watchdog defaults. Firmware that assumes a boot completes within a fixed window will fail intermittently against a slower candidate, and intermittent boot failures are expensive to diagnose in the field.

Driver support: whether a vendor driver exists, whether the OS or RTOS carries in-tree support, whether the driver already recognizes the candidate's ID, and any license implication of the driver you would adopt.

Memory deserves its own pass. For flash and EEPROM, compare page and sector geometry, erase time, endurance, and write-protect behavior against what the bootloader and the wear-leveling assume. For DRAM, compare SPD contents and timing parameters against the controller configuration.

Then make the build decision deliberately: can one firmware image support both parts through runtime detection, or does the BOM become coupled to a firmware version? Coupling is sometimes the right answer, but it creates a manufacturing and service obligation — the wrong image on the wrong board is now a defect — so it should be a recorded decision, not a discovery.

Thermal requirements

Do not assume thermal equivalence from a package name. RθJA is measured on a standardized JEDEC test board and almost never matches your product; use RθJC and ΨJT with your own copper area, layer count, and airflow to compute junction temperature.

Compute worst-case junction temperature at maximum ambient, maximum load, and end-of-life efficiency, then check it against the candidate's maximum rating — which may be 150 °C or 125 °C depending on the part — and against your internal derating policy. Derating rules frequently disqualify candidates that are technically inside the datasheet limit, and it is better to discover that before the qualification build than during design review.

Compare the thermal path itself: exposed-pad area, via count and pattern, copper pour, and layer stackup. A candidate with a smaller thermal pad on the same footprint changes RθJA materially even when everything electrical matches.

Passives have thermal limits too: capacitor ripple current and the self-heating it causes, inductor core loss and saturation at temperature, resistor power derating above roughly 70 °C, and fuse derating. Compare thermal-shutdown thresholds and hysteresis against the normal operating envelope so protection does not engage during legitimate operation.

Then verify rather than trust the model. A thermal-camera pass on a loaded board across the temperature range is inexpensive relative to a field return caused by a 15 °C modeling error.

Environmental requirements

Temperature grade is the single most common silent downgrade. Commercial (0 to +70 °C), industrial (-40 to +85 °C), extended and automotive (-40 to +105 or +125 °C), and military (-55 to +125 °C) parts can be otherwise identical, and a narrower-grade candidate will pass every bench test and then fail a cold start. Confirm the candidate covers the product's specified range at the specified performance — many parts operate outside the range they are guaranteed accurate in.

Humidity and contamination: moisture sensitivity level and floor life, conformal-coating compatibility, biased-humidity exposure, and corrosion of terminations. Sulfur-induced open circuits in thick-film chip resistors are a well-documented failure mode in some industrial environments, and anti-sulfur variants exist specifically for it — if the original was one, the candidate must be too.

Vibration, shock, and board flex: package mass and standoff, connector retention, board-level drop performance, and mechanical cracking in large ceramic capacitors near board edges or mounting holes. Soft-termination variants exist for that case and are a genuine functional difference, not a marketing option.

Altitude: creepage and clearance requirements derate with altitude for high-voltage parts, convective cooling weakens, and sealed components see a pressure differential. If the product is specified for altitude, this is a requirement, not a footnote.

Finally, check where the environmental specification came from. If it originated in a customer requirement document, the change may need customer notification regardless of your own technical judgment.

Reliability and qualification requirements

This group is where an otherwise-perfect candidate gets rejected, and where finding out late is most expensive. Establish the requirements before the lab work, not after.

Automotive: confirm the required qualification and grade — AEC-Q100 for integrated circuits, AEC-Q101 for discrete semiconductors, AEC-Q200 for passive components — and the temperature grade within it. Then confirm the supporting production requirements: whether a PPAP package exists at the level your customer requires, IMDS entry, and the change-notification obligations your customer contract imposes on you.

Functional safety: for ISO 26262 or IEC 61508 work, the part needs a safety manual, FMEDA data, and FIT rates with stated confidence. A candidate without them cannot be dropped into an existing safety case without analysis effort that often exceeds the effort of qualifying a different part.

Reliability data generally: qualification report, HTOL hours, FIT rate, and known wear-out mechanisms — electrolytic capacitor life, relay and connector mating cycles, flash endurance — checked against the assumptions already baked into your reliability prediction.

Regulatory approvals travel with specific part numbers, not with part types. Where the original carries a UL or IEC recognition (fuses, capacitors across the mains, optocouplers, transformers), an isolation rating with defined creepage and clearance, or a radio certification, confirm the candidate carries the equivalent and that the swap does not invalidate a product-level certification. Material compliance belongs here too: RoHS, REACH, halogen-free, conflict minerals, and country of origin where tariffs or export control matter.

Customer-specific requirements are the ones teams forget: an approved-vendor list, a contractual change-notification or approval clause, or a configuration freeze can make a technically excellent candidate unusable. Read the contract before booking lab time.

Then agree the requalification scope explicitly. A delta qualification targets the differences you identified in the groups above; a full requalification repeats the original programme. Which one applies is a quality decision, and it should be recorded before testing starts so the results are not later judged against a scope nobody agreed to.

Lifecycle health of the candidate

The most common own-goal in alternate selection is spending a qualification cycle on a part that is itself near end of life. Check the candidate's lifecycle position early — it is the cheapest disqualifier available.

Start with the manufacturer's own published status (active, NRND, last-time-buy, obsolete) rather than a third-party field, and record the source and the date you checked. Then look at the family rather than the single part: introduction date, whether the manufacturer is still adding parts to the family, whether a longevity or product-availability commitment is published for that product line, and whether the process node is one the manufacturer is consolidating away from.

Check for open change notices against the candidate as well. A part that has just been through an assembly-site transfer, a die shrink, or a material change may not behave exactly like the datasheet revision your evaluation is based on, and the notice will tell you which parameters the manufacturer thinks moved.

Expect sources to disagree. A distributor's status field, a data aggregator's status, and the manufacturer's page are three separate claims made at three different times, and they routinely conflict. Treat the manufacturer's published status as primary, note the conflict rather than silently picking one, and record which source you relied on. A lifecycle status without a source and a date is not evidence a year later, when someone asks why this part was chosen.

Supply availability

Availability today is not availability across the product's remaining life, and a candidate that solves this quarter's shortage while creating next year's is not a solution.

Look at the shape of supply, not just the headline number: stock across authorized distributors, factory lead time, minimum order quantity and order multiples, and whether stock is concentrated in one distributor, one date code, or one region. Broad stock in modest quantities across several distributors is generally healthier than a single large position.

Ask whether the candidate is itself multi-sourced. Replacing one sole-sourced part with another sole-sourced part is lateral movement, not risk reduction, and it should be described that way in the record.

Match supply to your demand profile: annual usage against typical stocking levels, whether your build quantity can be absorbed without a special order, and any non-cancellable, non-returnable terms attached to scheduled orders. Lead-time stability matters more than the absolute number — a consistent twelve weeks is easier to plan around than a figure that oscillates between four and forty.

Treat distributor stock, price, and lead-time figures as estimates that move. They are a snapshot for prioritizing candidates, not a commitment; re-check them immediately before the buy decision rather than relying on the numbers captured during evaluation. Discontinued parts are a special case — a factory lead time for a part that is no longer manufactured is not a meaningful number, whatever a field somewhere reports.

Authorized sourcing and provenance

Confirm the candidate is available through franchised (authorized) distribution for that manufacturer in your region, and confirm it from the manufacturer's own authorized-distributor list rather than a distributor's description of itself. Authorized purchase carries the manufacturer's warranty, a traceable provenance chain, and — importantly for this workflow — the manufacturer's change-notification chain, so future PCNs about the part actually reach you.

Independent (open-market) distributors are legitimate and are sometimes the only route to an obsolete part, but the evidence burden moves to you. Define in advance what you will require: certificate of conformance, traceability documentation back toward the manufacturer, and incoming inspection and test performed against a counterfeit-avoidance standard — SAE AS5553 and AS6081 describe the practices most programs adopt. Record what was inspected and what the result was; an undocumented inspection is indistinguishable from no inspection during an audit.

Treat independent-only availability as a technical signal, not just a purchasing detail. If the only stock of your candidate is in the open market, that tells you something about the candidate's supply health that belongs in the evaluation, not just in the purchase order.

Then record the sourcing policy for this specific part: authorized only, authorized-preferred with a documented exception path, or open market permitted with defined inspection. Making that policy explicit per part is what prevents a shortage-driven exception from quietly becoming the standing practice.

Cost and commercial constraints

Price the candidate at your actual order quantity, not at the 1k break, and compare landed cost: unit price, duty and tariff, freight, minimum-order overage, and any minimum order value. A cheaper unit price behind a 10,000-piece MOQ is not cheaper for a 1,200-unit build.

The cost of the change usually dwarfs the piece-price delta. Count engineering hours, layout revision, prototype builds, lab and EMC test, requalification, documentation updates, customer notification, firmware release, and write-off or rework of existing inventory. Once that total is visible, some price increases stop looking like a problem and some savings stop looking worthwhile.

Plan the transition: existing stock of the original, work in progress, and consigned material at the contract manufacturer all need a phase-in decision — run the original out and switch cleanly, release both parts as approved alternates, or switch immediately and dispose of the remainder. Each has a different cost and a different traceability consequence.

Check the timing against the reason for the change. An alternate whose lead time lands after the build it was meant to protect has not solved anything, and that is worth saying out loud early enough to pursue a second option in parallel.

Review commercial terms before committing: non-cancellable and non-returnable conditions, how long the quoted price is valid, minimum annual commitments, and whether a franchised distributor will actually support your volume. Where the alternate is more expensive, get the margin impact acknowledged by whoever owns it rather than letting it surface in a quarterly review.

Documentation and the decision record

The output of this checklist is a record, not a conversation. Capture the original part, the candidate, the reason the evaluation happened (which notice, which shortage, which design goal), a pass / fail / not-applicable result with supporting evidence for every group above, the residual risks being accepted, the approver, and the date.

Attach the artifacts rather than referring to them: both datasheets with revision numbers and dates, the parameter comparison table, test reports, the qualification plan and its results, the supplier's PPAP or certificate of conformance where applicable, and the original supplier notice if a notice triggered the work. Datasheet revisions matter — "we compared the datasheets" is not reproducible a year later unless the revisions are named.

Then update the downstream records, because an approval that lives only in the evaluation document does not reach production: BOM and alternate or AVL entries (including which assemblies the alternate is approved for), the ERP part master, test procedures, work instructions, service and spares documentation, and firmware release notes.

Identify notification obligations early. Many customer contracts require notice before a component change and some require approval; medical, aerospace, and automotive programmes may involve a notified body or a customer PPAP submission. Approval lead time is frequently the longest pole in the whole change, and it is not one you can compress with engineering effort.

Record the rejected candidates and the reasons too. The next shortage will surface the same three parts, and a written "rejected: temperature grade is industrial, product is specified to -40 °C at full accuracy" saves the entire analysis from being redone by someone who was not in the room.

This is the record PCNshark is built to hold: the lifecycle status with the evidence behind it (a notice in the system, an uploaded document, or a stated source), the manufacturer-recommended replacement and the specification-ranked functional candidates that were considered, distributor sourcing context, and part-level decision history showing what was decided and when. A manual decision is never silently overwritten — conflicting newer evidence is flagged for review, so the disagreement surfaces instead of disappearing.

Final engineering approval

Approval is a person, not a status field. Name who signs, at what authority level, and precisely what they are attesting to — that the candidate meets the requirements evaluated above, with the listed exceptions, for a defined scope.

State the interchangeability scope explicitly, because this is the detail most often lost. Is the alternate fully interchangeable in both directions, interchangeable one way only (the candidate may replace the original but not the reverse), or approved for a specific assembly, revision, or build? One-way interchangeability is extremely common when the candidate is a functional superset, and recording it as "equivalent" creates a defect the next time someone substitutes in the other direction.

Write conditions and expiry into the record. Approvals granted under shortage pressure often carry them — "build 12 only," "pending EMC retest," "until the original returns to stock" — and a condition without an owner and a date becomes permanent by default.

Confirm the approval actually reached the people who act on it: purchasing, the contract manufacturer, incoming inspection, and whoever maintains spares and service stock. An alternate approved in engineering and unknown to the CM is not an approved alternate.

Close the loop after the change. Define the post-change verification up front — first-article inspection, a defined number of monitored builds, and a field-return watch period — with a named owner and a date at which the alternate is either confirmed or revisited. Qualification ends when that check completes, not when the part ships.