Cross-references are leads, not verdicts
Every cross-reference database — including the ones run by manufacturers themselves — is a starting point, not a verdict. onsemi says it plainly on its competitor cross-reference page: the data is provided “for suggestion purposes only,” based on the company’s “best estimate” of other manufacturers’ published information, and buyers should “completely review our datasheet to confirm the device functionality” for their application. If the manufacturer publishing the cross will not stand behind it as a guarantee, no buyer should treat a third-party database match as an approval. Third-party databases add a second layer of risk: they aggregate other manufacturers’ published data, sometimes stale by years, and their “equivalent” flags are marketing as much as engineering. A database that lists fifty “equivalents” for one part has not verified fifty datasheets against your operating point — it has matched fifty headline tables. Use the list as a candidate pool, then run the checklist below on each candidate you actually intend to buy. The working rule for this article: a cross-reference tells you which datasheets to open — never which part to buy.
The four-step verification checklist
Run every candidate through the same four steps, in order, and record the result of each. Step 1 — form and fit as a gate, not a score: package, pinout, dimensions, land pattern, and assembly-process compatibility must match before anything else is compared; a candidate that fails here is rejected no matter how attractive its specs look. Step 2 — parametric comparison at the operating point: compare the parameter your circuit actually depends on — noise and bandwidth for a signal-path op-amp, not just headline gain — using the datasheet of the exact candidate ordering code, not the family summary. Step 3 — qualification and grade: temperature grade, AEC-Q or other reliability standards, and regulatory status must meet the original’s, because a commercial-grade substitute in an industrial-temperature design is a field failure waiting for summer. Step 4 — change control: record every material difference found in steps 1–3, keep the original MPN on the record, and require engineering sign-off before the alternate enters the BOM.
- Step 1 — form and fit: package, pinout, land pattern, assembly process. A gate, not a score.
- Step 2 — parametrics at the operating point: the parameter your circuit depends on, from the exact candidate datasheet.
- Step 3 — qualification and grade: temperature grade, AEC-Q, regulatory status must meet the original’s.
- Step 4 — change control: record every material difference, keep the original MPN, require sign-off.
Grade the risk before you approve
Not every approved alternate carries the same risk — grade it so the next engineer knows what was actually verified. Drop-in: package and pinout, headline parameters, qualification grade, and regulatory status all match — document the comparison and release. Conditional: one variable differs within a bounded, re-qualified range, such as a wider temperature grade or a tighter tolerance — approve only for the verified condition, and write the condition on the approval itself. Redesign-required: footprint, pinout, or qualification mismatch — treat it as a new design choice with layout review and re-qualification, never as a substitution. The grade is the point of the review: two approved alternates can look identical on a BOM line while carrying completely different risk, and the grade is what tells the next engineer which one they are looking at.
- Drop-in: everything matches — document the comparison and release.
- Conditional: one bounded difference, re-qualified — write the condition on the approval.
- Redesign-required: footprint, pinout, or qualification mismatch — new design choice, not a substitution.
Traps that pass the bench and fail in the field
The dangerous substitutes are the ones that work on the bench. A pin-compatible op-amp can pass every functional test and then fail on noise in the field — TL072CDR has several pin-compatible options with very different noise and bandwidth, and the replacement met every headline spec except the one nobody compared. A shared package name is not a shared footprint: the LM358 and NE555 families each ship in documented package variants — SOIC, VSSOP, TSSOP, PDIP — that are not interchangeable without checking the pinout. LM358DR is the SOIC-8 ordering code; the VSSOP, TSSOP, and PDIP codes are different orderable parts, and our catalog records the distinction because the footprint check is where substitutions most often die. NE555DR carries the same lesson on the timer side. And an “upgrade” can be a downgrade in disguise: a lower-ESR capacitor in an LDO circuit designed around a specific ESR window can turn a stable rail into an oscillator, and a faster comparator dropped into a tuned timing loop can shift the operating point the original designer settled around. “Better on paper” is not a qualification result — it is a reason to re-run step 2. In every case the fix is the same — compare at the operating point, not the headline, and distrust any approval granted from memory.
Who approves, and what the paper trail looks like
Engineering approves; procurement documents. The minimum paper trail for an approved alternate: the original MPN, the proposed alternate’s full ordering code, every material difference found in the four-step check, the risk grade, and a named sign-off with a date. Keep the original MPN in the RFQ alongside the proposed alternate so the distributor quotes exactly what engineering approved — our RFQ workspace captures part number, manufacturer, package, date code, target price, and quantity per line, plus a notes field where the substitution policy (“alternates acceptable where the four checks pass” or “exact part only”) belongs. An alternate approved on paper but ordered by part number alone is how the wrong reel arrives. Name the reviewer, not just the role: a sign-off from “engineering” with no name and no date is decoration. Record the date because alternates age — a conditional approval granted against a 2024 datasheet revision deserves a second look when the manufacturer issues a process-change notice in 2026.
- Original MPN plus the proposed alternate’s full ordering code — never a family name.
- Every material difference from the four-step check, and the risk grade.
- A named engineering sign-off with a date — verbal approvals do not survive audits.
When the alternate comes from a shortage
Shortage substitutions follow the same four steps — they just run under time pressure, which is when steps get skipped. If allocation forces the swap, the four-number cross-reference — capacitance, voltage, dielectric, and case size, re-checked against each maker’s curve — is the fastest legitimate gate before the full review, and our guide to finding alternative components safely covers the qualification paperwork that follows. The discipline that matters most in a shortage is step 4: record the approval while the original is still the reference, because six months later nobody remembers which differences were accepted and why.
Frequently asked questions
Quick answers to the questions buyers ask most about reviewing alternatives.
- Q: A cross-reference tool lists it as equivalent — is that enough? A: No. Even manufacturer-run tools disclaim it: onsemi’s cross-reference page states its data is “for suggestion purposes only” and recommends a full datasheet review. Treat databases as search leads; only a datasheet-to-datasheet comparison counts.
- Q: The substitute has the same package — is it a drop-in? A: Not necessarily. Check pinout, land pattern, and assembly compatibility — package variants are not interchangeable without checking the footprint.
- Q: Who approves the substitute? A: Engineering. Keep the original MPN in the RFQ, name the proposed alternate’s full ordering code, record every material difference and the risk grade, and require sign-off.
- Q: Only the brand differs — can I approve it? A: Brand is negotiable, but re-verify the non-negotiables at the operating point — two brands’ “equivalents” can differ in exactly the parameter your circuit depends on.