Obsolescence management: last time buy or redesign?

Obsolescence management is the discipline of keeping a product buildable when its components stop being made. On a recent project for a customer, the power management chip and the DDR memory both went obsolete: the customer made a last time buy, and we did a full re-layout that became the next version of the product. Both answers were needed.

Contrary to the usual framing, the choice is rarely last time buy or redesign. In our practice the useful question is different: how much stock buys enough time to redesign properly. At AESTECHNO, an electronic design firm in Montpellier, we handle this on boards we design and on boards we take over. This guide covers the notices, the four possible responses, how to size a last time buy, the decision matrix and the field report, and links to our related work on component shortages and the 2026 memory shortage.

What obsolescence management is

Obsolescence management is the set of policies, monitoring and decisions that keep an electronic product in production when a component it uses is discontinued. According to IEC 62402:2019, the international obsolescence management standard, obsolescence is an item that is still needed becoming unavailable from its original manufacturer, not an item falling out of use.

The standard, published on 29 May 2019 by IEC Technical Committee 56 (Dependability), is written as a set of requirements rather than a guide. Its IEC webstore summary lists what it expects an organization to have: an obsolescence policy, an infrastructure and organization, an obsolescence management plan, design strategies that limit obsolescence, a way to choose an approach and a resolution, and performance measurement. In US defense programs the same discipline goes by the name Diminishing Manufacturing Sources and Material Shortages (DMSMS).

Obsolescence is not a shortage

Unlike obsolescence, a shortage means a part is still made but hard to get or expensive; obsolescence means it will not be made again. The two often arrive together, as the 2026 memory market showed, but they call for different answers: a shortage is usually waited out or bought through, while obsolescence ends with a stock, a substitute or a new board. Our article on managing electronic component shortages covers the first case.

How a component goes end of life: NRND, PCN, EOL and LTB

A component goes end of life through a sequence of supplier notices. It is first marked Not Recommended for New Designs (NRND), then a Product Change Notification (PCN) or discontinuance notice announces the End of Life (EOL) and opens a Last Time Buy (LTB) window, after which the part is shipped one last time and stops.

NRND is a warning, not a deadline: the part is still produced and sold, but the manufacturer tells designers not to start a product on it. Contrary to a common reading, an NRND status on a part already in production is not a reason to panic; it is the cheapest moment to plan, because no clock is running yet. The general notion of an end-of-life product covers the same idea across industries: the vendor stops marketing, selling and eventually supporting the item.

What JEDEC requires in a discontinuance notice

For semiconductors, the reference is JEDEC JESD48, Product Discontinuance. According to JESD48C (December 2011), section 3.1, the supplier shall give notice "allowing a minimum of 6 months from the notice to place final orders, and 12 months from the notice for final shipments". Section 3.3 requires the notice to contain, as a minimum, the effective date for last purchase orders and the affected supplier part numbers.

JESD48C has since been superseded by J-STD-048, published jointly by JEDEC, IPC and ECIA (the Electronic Components Industry Association) in November 2014 and revised as J-STD-048A in late 2025. Individual suppliers publish their own terms on top. A Nexperia discontinuation notice we read for this article, CN 202410011DN, cites JESD48 and allows "a six (6) month last order period for multiple-source products and a nine (9) month last order period for sole-source discontinued products", with a twelve-month delivery period from the notice date.

What those months mean for a product team

Six months is short. It is the time to notice the email, find the affected products, forecast the remaining demand, get the purchase approved and place the order. A notice sent to a former purchasing contact or filtered as a newsletter can consume half the window before anyone in engineering hears of it. The same notice also says what it does not promise: Nexperia states that the last time buy date "automatically expires" when the remaining inventory is sold, and that some parts cannot be bought at all.

Why industrial and embedded products are exposed

Industrial and embedded products are exposed to obsolescence because they are sold and serviced for many years, while many of their components follow the shorter cycle of consumer electronics. A product certified once, qualified by its customers and integrated into their systems is expensive to change, so its life is set by the market, not by its parts.

Memory is the clearest example. Memory makers stop older generations when capacity is worth more elsewhere: in 2025 the large manufacturers announced the end of DDR4 production for mainstream markets, as we described in our article on the 2026 RAM shortage and embedded products. DDR4 is still a valid standard, defined by JEDEC JESD79-4, but a standard does not keep a production line open. There are exceptions: Micron started long-lifecycle DDR4 and LPDDR4 production at its Manassas, Virginia site for automotive, industrial and medical markets.

Which parts tend to go first

In our practice, the parts that go first are those tied to a consumer generation: memory, application processors, power management chips designed around a specific processor, displays and some radio modules. Passive components and common logic rarely force a redesign because drop-in equivalents exist. The risk concentrates on single-source parts that the rest of the board was designed around, which is exactly where a substitute is hardest to find. Our comparison of DDR4 and DDR5 explains why a memory change is rarely a drop-in.

The four responses: last time buy, second source, module or redesign

The four responses to an obsolete component are a last time buy, a second source, a move to a module that absorbs the part, and a redesign. They differ in engineering effort, cash tied up and how long they keep the product buildable, and the right choice depends on how long the product still has to be sold.

A last time buy needs no engineering but ties up cash and storage and only lasts as long as the stock. A second source, a qualified drop-in or near drop-in part, needs a test procedure on the affected functions but usually no layout change. A module, such as a system-on-module that carries the processor and memory, moves the obsolescence problem to the module vendor. A redesign resets the clock for the whole board but costs a design cycle, tests and often a new certification.

The same comparison, with what to watch

The table below adds the trap each response hides. None of them is free, and the cheapest-looking one is the one whose cost arrives later.

How to size a last time buy

Sizing a last time buy means estimating how many parts the product will consume until it is retired or redesigned, then adding what production and service will lose along the way. The quantity is a forecast, so the method matters more than the precision: a documented estimate can be revised, a guess cannot.

Our procedure has five terms. Remaining production demand is the units still to build multiplied by the parts per unit, over the period until end of sales or until the redesigned version ships. Service spares cover repairs and warranty over the support period. Production attrition covers parts lost to placement errors, rework and test failures, which an EMS (Electronic Manufacturing Services) partner can quantify from its own records. A safety margin covers the forecast error. Finally, stock already on hand and at the assembler is subtracted.

Storage, moisture and cash

A last time buy is also a storage commitment. Moisture-sensitive surface-mount parts must be stored and baked according to IPC/JEDEC J-STD-033, which sets handling rules by Moisture Sensitivity Level (MSL); parts stored for years in the wrong conditions can fail at reflow. As an example of the scale, an MSL 3 part has a floor life of 168 hours at 30 °C and 60% relative humidity once its dry bag is opened. Even though the purchase looks like a one-off, its cost includes dry storage, inventory tracking and the cash immobilized until the last unit ships. Despite these costs, a well-sized buy remains the cheapest way to keep shipping while an alternative is prepared.

Field report: an obsolete PMIC and DDR, a last time buy and a full re-layout

On a recent client project, two central components became obsolete: the Power Management Integrated Circuit (PMIC) and the DDR memory chip. The customer made a last time buy to keep production running, and we replaced both parts with a full re-layout of the board, which became the new version of the product.

The case shows why the "last time buy or redesign" framing is too narrow. Neither response alone would have worked well. The stock alone would have ended production when it ran out; a redesign alone, started on the notice date, would have left a gap between the last parts and the first boards of the new version. Together, the stock covered production while the new version was designed, tested and released. The product is under NDA, so we do not give part numbers, quantities or dates here.

Why a PMIC and a DDR chip tend to travel together

In general, and not only on this product, the PMIC and the memory are coupled on a processor board. The PMIC generates the rails the processor and the DDR need, with their voltages, current limits and power-up sequence. Memory generations differ here: DDR3 runs at 1.5 V, DDR3L at 1.35 V, DDR4 at 1.2 V, and LPDDR4 uses a 1.1 V I/O rail. A memory of another generation can therefore need different rails and sequencing, and a new PMIC changes the placement and the power plane layout.

The DDR routing itself is length-matched and impedance-controlled, as our guide to LPDDR4 memory design shows, so a new memory package rarely fits the old footprint, stackup, trace lengths, via placement and differential pair routing. When two parts of that cluster go at once, a partial patch costs most of a redesign and keeps the old constraints; a full re-layout lets the board be rebuilt around parts with a future.

What we recommend after this case

We recommend treating the processor, its memory and its PMIC as one obsolescence cluster in the bill of materials review, and checking their lifecycle status together. When one of them receives an NRND or EOL notice, size the last time buy on the redesign horizon, not on the end of sales, and start the redesign while the window is still open. In our lab, the test procedure on a re-layout is the one we apply to any new board: signal integrity checks on the DDR interface with our Tektronix oscilloscope and TekExpress suite, power sequencing on the bench, then the validation and certification steps described in our guide to electronic product testing and validation.

Decision matrix: last time buy or redesign

The choice between a last time buy and a redesign depends mainly on the remaining sales life of the product, on whether a qualified substitute exists, and on how many of the board's central parts are affected. A short remaining life favours stock; a long one with no substitute favours a redesign, usually bridged by a last time buy.

In our practice, the flow below is the one we use in obsolescence reviews. It starts with the substitute because a validated drop-in is almost always the cheapest answer, and it treats the bridging buy as a normal part of a redesign rather than as an alternative to it. Each answer ends with a date to review it.

The matrix in a table

Our method: an obsolescence review in five steps

An obsolescence review is a structured check of a product's bill of materials against the lifecycle status of each part, ending with a decision and a date for each risk. We run it when a notice arrives, when we take over an existing product, and before a planned new version, so that the next design does not inherit the next shortage.

Step one is the inventory: an exact Bill of Materials (BOM) with manufacturer part numbers, not distributor references. Step two is the lifecycle status of each line from the manufacturers and distributors, flagging NRND, EOL and single-source parts. Step three groups the parts into clusters that cannot change independently, such as processor, memory and PMIC. Step four estimates the cost and lead time of each response per cluster. Step five records a decision with a review date, because a status that is fine today can change next quarter. For a second source, the review also defines the test procedure that will validate it.

When we take over a board designed elsewhere

End-of-life notices also hit boards designed by someone else, sometimes years ago, with incomplete files. In that situation the review starts by rebuilding a trustworthy BOM and schematic from what exists. Our article on taking over a stalled electronics project describes that recovery work, and our electronic project risk management guide puts obsolescence among the other supply risks.

Designing for obsolescence from the first version

Designing for obsolescence means choosing parts and architectures on the first version so that the inevitable end-of-life notices cost a validation instead of a redesign. IEC 62402 lists design strategies among its requirements, and most of them cost little when applied at the start and a great deal when retrofitted.

The levers we use are well known: prefer parts with several sources or published long-term availability programs; check lifecycle status at schematic review, not after the prototype; group the processor, memory and power into a module when volumes and lifetime justify it, as discussed in our guide on choosing between SoC, SoM, SBC or custom; keep spare footprint options for passives and regulators; and keep the design files complete, so that a future re-layout starts from a clean base. Our design for manufacturing guide covers the production side of the same choices.

Certification after a component change

A component change can reopen certification. A new PMIC changes switching frequencies and the emission profile; a new memory changes fast edges on the board. Depending on the change and the directives that apply, a partial or full retest of electromagnetic compatibility may be needed, as our guides to EMC and CE RED certification explain. In practice, a full re-layout is the moment to fix known EMC margins at the same time, since the tests will run anyway.

Bottom line

Obsolescence management is less a choice between a last time buy and a redesign than a question of timing: how much stock buys the time a proper fix needs. The notices give a short, fixed window; the work that follows depends on which parts are hit and how long the product still has to sell.

• The window is short. JEDEC JESD48C requires at least 6 months to place final orders and 12 months for final shipments after a discontinuance notice.
• Look for a substitute first. A validated drop-in second source is almost always the cheapest answer.
• Size the buy on the right horizon. Until end of sales for a mature product, until the next version ships when a redesign is planned.
• Processor, memory and PMIC are one cluster. On a recent project both the PMIC and the DDR went obsolete; a bridge buy plus a full re-layout was the answer.
• Design for it from the first version. Lifecycle checks at schematic review cost little; retrofits cost a design cycle.

Related reading: managing electronic component shortages and obsolescence, the 2026 RAM shortage and embedded products, DDR4 versus DDR5, embedded power management, PCB stack-up and impedance design, COTS versus custom electronics, what drives electronic product development cost, from prototype to manufacturing, the PPE resin shortage and PCB laminates, the DDR4 DIMM and SO-DIMM pinout, our hardware design services and our embedded firmware expertise.

Written by Hugues Orgitello, an electronic-design engineer with 10+ years of experience who founded AESTECHNO in Montpellier in 2022.

FAQ: component obsolescence management

What is obsolescence management in electronics?
It is the set of policies, monitoring and decisions that keep an electronic product buildable when its components are discontinued. IEC 62402:2019 sets the requirements: an obsolescence policy, an organization, a management plan, design strategies that limit obsolescence, a method to choose a resolution and performance measurement. In practice it means tracking the lifecycle status of every part in the bill of materials and deciding, before the window closes, between a last time buy, a second source, a module or a redesign.

How much notice does a supplier give before end of life?
JEDEC JESD48C requires a minimum of 6 months from the discontinuance notice to place final orders and 12 months for final shipments. It was superseded by J-STD-048, published by JEDEC, IPC and ECIA in 2014 and revised in 2025. Suppliers add their own terms: one Nexperia notice allows 6 months for multiple-source parts and 9 months for sole-source parts. Contracts can set other terms, and stock can run out before the stated date.

What does NRND mean on a datasheet?
NRND means Not Recommended for New Designs. The part is still produced and sold, but the manufacturer signals that it should not be used in a new product, usually because an end of life is planned or a successor exists. On a product already in production, NRND is not an emergency: it is the best moment to check for a substitute, plan a last time buy and decide whether the next version should change the part.

How do you calculate a last time buy quantity?
Add the remaining production demand, the service and warranty spares, the production attrition from placement, rework and test, and a safety margin for forecast error, then subtract the stock already on hand and at the assembler. The horizon decides the result: until end of sales for a mature product, or until the redesigned version ships when a redesign is planned. Account for storage too, including the J-STD-033 rules for moisture-sensitive parts.

When is a redesign better than a last time buy?
When the product still has several years of sales ahead, no qualified substitute exists, or several central parts such as the processor, the memory and the PMIC are affected at once. In that case the two are combined: a last time buy sized on the redesign horizon keeps production running while the new version is designed, tested and certified. On a recent project with an obsolete PMIC and DDR, that is what we did, with a full re-layout.