BOM & components
How to Handle Obsolete and EOL Components in an Electronics BOM
What NRND, EOL and obsolete really mean, the four options when a part goes away, and the question you have to answer first.
What do NRND, EOL and obsolete actually mean?
They are three points on the same path, and manufacturers use them fairly consistently even though no single standard forces them to.
| Status | What it means | What you should do |
|---|---|---|
| Active | in production, recommended for new designs | nothing |
| NRND — not recommended for new designs | still being made, still buyable, but the manufacturer is signalling the end | stop designing it in; existing products are fine for now |
| EOL — end of life | a discontinuation notice has been issued, usually with a last-order date | decide: last-time buy, alternative, or redesign |
| Obsolete | no longer manufactured | you are buying from stock, brokers, or redesigning |
NRND is the useful one. It is the only status that gives you time. A part that goes straight from active to EOL leaves you with a deadline; a part that sits at NRND for two years leaves you with a plan.
PCN and PDN: the notices that carry the news
A PCN (product change notification) tells you something about the part is changing — the die, the package, the fab, the marking. It is not necessarily a problem, but it can be: a die shrink can change timing or current consumption enough to matter in a marginal design.
A PDN (product discontinuation notice) tells you the part is going away. It normally carries two dates:
last time buy (LTB) — the last date the manufacturer accepts orders; last time delivery — when the final shipments arrive.
Both notices are usually published by the manufacturer and distributed by distributors. Whether you receive them depends on whether anyone subscribed, and for most small teams the honest answer is that nobody did.
XKEMA does not monitor supplier lifecycle. It does not watch distributor feeds and it will not tell you a part has gone EOL. Nothing in this article should be read as suggesting it does. What it holds is the information you need after you find out — which is a different and also useful thing.
The first question is not "what replaces it?"
It is "where is this part used?"
A replacement decision for one board is an afternoon. The same decision across eleven products, three of which are still being manufactured and one of which is in a certified subsystem, is a different problem — and you cannot start it until you know the scope.
This is where most teams lose time, and it is a filing problem rather than an engineering one. If every project entered the same chip as free text — STM32F407VGT6, STM32F407VG, STM32F407-VGT6 — then no search answers the question and somebody opens eleven spreadsheets.
Your four options
1. Last-time buy. Buy enough to cover remaining production. Works when you can forecast demand and have somewhere to put the stock.
The risks are real: you are converting cash into inventory, components age (electrolytics and some ICs have shelf-life considerations), and if your forecast is wrong you either scrap or redesign anyway. For a product with a clear end date, it is often the cheapest option. For one you plan to make indefinitely, it postpones the problem at a price.
2. Drop-in alternative. A part that is genuinely equivalent for this circuit. Cheapest when it exists — but see the qualification scale in approved alternatives, because a matching footprint is not the same as a working substitute.
3. Local redesign. Change the footprint and the surrounding circuit without touching the rest of the board. A new regulator in a different package, a different connector with a new land pattern. New PCB revision, same product.
4. Full redesign. When the obsolete part is a processor, an FPGA, or anything whose replacement changes firmware, layout and test. Expensive and occasionally the correct answer, especially if several parts are going EOL at once.
A decision rule that holds up
Is there a drop-in alternative?
├── yes → qualify it according to its position in the circuit → done
└── no
├── Is remaining production volume predictable and bounded?
│ ├── yes → last-time buy, and schedule the redesign anyway
│ └── no → continue
├── Can the change be contained to one area of the board?
│ ├── yes → local redesign, new PCB revision
│ └── no → full redesign
└── Is the part in a certified subsystem?
└── yes → add requalification cost and time to every branch aboveThe last line is the one that surprises people. A trivial component swap inside a certified assembly can cost more in paperwork and retesting than the redesign itself.
What to document when you replace a part
The replacement outlives everyone's memory of why it happened. Record:
the original MPN and its status notice — PDN reference and date if you have it; the replacement MPN, and what was compared; the qualification level: paper, bench, full; which products and revisions are affected, and from which revision the change applies; whether existing stock of the original is still to be consumed first; whether the change invalidates any test report or certification.
That last point gets forgotten most often, and it is the one that turns a component swap into a compliance problem eighteen months later.
A worked example
The fictional ORION CONTROL UNIT uses a connector that goes EOL.
| Part | J1, original 10-pin connector |
| Notice | PDN received via distributor, last-time buy in four months |
| Used in | Main Control Board rev B and rev C — two products |
| Remaining production | ~400 units/year, indefinite |
| Drop-in available? | No — closest equivalent has a different land pattern |
| Decision | Local redesign. LTB would cover two years at best, and the product has no end date |
| Result | 1-2112931-1, new land pattern, Main Control Board rev C |
| Knock-on | Gerbers regenerated, BOM rev 4, assembly drawing reissued |
| Not affected | Power Supply Board — does not use J1 |
Sample data. Note the last row: knowing what is not affected is as valuable as knowing what is, and it is the part you can only answer if the component is recorded once and referenced everywhere.
Common mistakes
Finding out from a failed order. The first notice many teams get is purchasing reporting that a part cannot be bought. By then the LTB window has usually closed.
Treating NRND as "fine for now" forever. NRND is a warning with no deadline attached, so it gets ignored until it becomes EOL with a deadline attached.
Replacing on one board and forgetting the others. Especially when the same part is entered differently in each BOM.
Last-time buying without a redesign plan. The stock runs out eventually, and usually at a worse moment than the original notice.
Not recording the reason. Two years later nobody knows whether the old part was obsolete or just more expensive, so the change gets questioned every time.
Finding out that a part is obsolete is a supply-chain problem. Answering which of our products use it, in which revisions, and what else has to change is a filing problem — and it is the one that decides whether the response takes an afternoon or a fortnight.
In XKEMA a component is a record in a shared library, and every BOM line points at it rather than repeating its name as text, so the question "where is this used?" has an answer.
Component library BOM management Approved alternatives in an electronics BOM Reopening an old PCB design