Obsolete component replacement is the step where a part on your board that nobody makes anymore gets swapped for something you can actually buy — without changing how the circuit behaves. It happens after the board is mapped and the bill of materials is confirmed, and it involves three things: identifying exactly what the dead part did, finding a current part that does it inside the same electrical and thermal envelope, and proving that on hardware before you commit to a build.
If you’re staring at an EOL notice for a linear regulator, an opto-isolator, or a 20-year-old 8-bit MCU, this page describes what actually happens on the bench.
Why obsolescence shows up in the middle of a board copy, not at the start
Most customers don’t come to us for obsolescence. They come because a machine is down, or because the original layout files are gone, and they want the board rebuilt. Obsolescence is what we find along the way.
The pattern is consistent. On a 15-year-old industrial control board, the passives are all fine — 0805 resistors and X7R caps are eternal. The connectors are usually still available or have a drop-in equivalent. What bites you is the semiconductors: a Fairchild opto that ON Semi discontinued, a Maxim RS-485 transceiver that got NRND’d, an ST 8-bit MCU that hasn’t been in the catalog for a decade, a magnetics part from a vendor that no longer exists.
So the work order effectively splits. Part of the board is a straight duplication job. Part of it is a small redesign problem that has to be solved before anything gets fabricated.
Step one: knowing exactly what you’re replacing
You can’t substitute a part you haven’t positively identified. This sounds obvious, and it’s where most in-house attempts stall out.
Markings on old parts are frequently a problem. Laser marks fade. Date codes and house numbers get confused for part numbers. Boards that ran hot for years have components whose top surface is baked to the point where only a couple of characters are legible under a stereo microscope. Sanitized boards — where the OEM deliberately ground off or over-marked the part numbers — are their own category.
What we actually do, in order:
- Optical read under magnification and angled light. Raking light at a low angle brings back faded laser marks that look blank under direct illumination surprisingly often.
- Package and pinout constraint. An SO-8 with pin 1 tied to a bulk cap, pin 4 to ground, and pin 8 switching into an inductor is a buck controller, not a dual op-amp. Topology narrows the candidate list fast.
- In-circuit and out-of-circuit measurement. Curve-tracing a transistor or diode, measuring an inductor’s DCR and inductance, checking a crystal’s frequency. This is how we pin down values that were never printed on the part in the first place.
- Cross-reference against the surrounding circuit. A part that reads plausibly as three different devices gets resolved by asking which one makes the circuit work.
This whole stage overlaps heavily with the component identification work we do on every board. If the part turns out to be a programmed device — a locked MCU, a serial EEPROM holding calibration data, a CPLD — identification is only half the problem, and the firmware side gets handled separately.
Step two: choosing a substitute an engineer can defend
A cross-reference from a distributor site is a starting point, not an answer. The parametric tables catch the headline specs and miss the ones that break boards.
Here’s what we check before calling a substitute valid:
- Absolute maximums with real margin. Not “close enough.” If the original was rated 60 V and the rail sees 48 V transients, a 50 V replacement is a field failure waiting for a hot day.
- Thermal behavior in the actual copper. A modern part in a smaller package often dissipates the same watts with a third of the pad area. IPC-2152 is the reference for how much current a given trace and copper weight can carry, and the same logic governs how much heat a pad field can pull out of a part.
- Timing and drive. Faster logic isn’t a free upgrade. A modern gate with a 2 ns edge into a long unterminated trace will ring where the original CMOS part never did.
- Quiescent current and startup behavior. Regulators are notorious here. Newer LDOs often need different output capacitance and have different ESR stability windows than the part they replace.
- Footprint and polarity. Same package family doesn’t mean same pinout. Tantalum vs. MLCC vs. polymer in the same case size behave differently on ripple and surge.
- Analog subtleties. Op-amp input bias current, offset drift, common-mode range. In a precision measurement front end, a “better” op-amp can shift a calibration curve enough to fail acceptance test.
- Authenticity of supply. If a genuine obsolete part is available from a broker rather than a franchised distributor, it gets inspected before it goes near a board — marking consistency, lead condition, and X-ray where the internal structure needs confirming.
The four routes out of an obsolete part
| Route | When it fits | What it costs you |
|---|---|---|
| Genuine stock from a verified source | Low volume, short remaining product life, part still floating in the channel | Cheapest engineering, highest counterfeit and supply risk; inspection required |
| Drop-in equivalent, same footprint and pinout | Standard logic, common regulators, optos, transceivers, passives | Verification effort only; layout untouched |
| Functional substitute needing local layout changes | Different package, extra decoupling, changed compensation network | A small re-layout of that block plus a prototype round |
| Section redesign around a modern part | Discontinued MCU family, custom ASIC, vanished magnetics | Real engineering time; firmware port may be needed too |
Most boards land in the first two rows for the majority of the BOM and in the third row for one or two parts. The fourth row is the case worth being honest about, and we’ll get to it.
Step three: verifying the substitute against the original board
A substitution isn’t finished when the part number goes into the BOM. It’s finished when the rebuilt board behaves like the original.
The reference is always the working original, if you can send us one. We measure the same nodes on both boards under the same conditions: supply rails under load, switching node waveforms, output ripple, reference voltages, timing on any bus that crosses the substituted block, and current draw at idle and at full load. Where a part sits in a feedback loop, we look at step response and overshoot, not just the DC value.
Thermal gets its own pass. An infrared image of the assembled board under representative load, compared against the original, catches the substitutions that pass electrically and then cook. Fabrication itself is checked against IPC-A-600 and IPC-6012 for the class the board needs — a modified footprint means annular ring and spacing get re-examined, not assumed.
The changes then get recorded properly: an annotated BOM listing the original part, the replacement, the reason, and the verification result, plus updated Gerber and schematic files that match what was actually built. This forms part of the file set you receive at handover, so the next engineer to touch the board isn’t reverse engineering your substitutions all over again.
What to send us, and what comes back
Getting a real answer instead of a vague one depends almost entirely on the first message. Send:
- Clear photos of both sides of the board, whole board in frame, on a dark background, in focus and evenly lit — no flash glare across the silkscreen.
- Close-ups of the parts you already know are obsolete, tight enough to read the marking. Type the marking out in text too, line by line, including the second and third lines.
- Board dimensions and layer count if you know it.
- Quantity: are you fixing one machine, or building 500 units a year for the next decade?
- One line on deliverables — replacement BOM only, or updated files, or assembled and tested boards.
- Whether the board has a programmed device on it, and whether you still hold the firmware.
What comes back is a feasibility assessment per obsolete part (drop-in, substitute with layout change, or needs redesign), a lead time, and a price. Cost is driven by how many parts are affected, whether any of them force a layout change, whether firmware is involved, and how much verification you need — see what actually moves the price on a reverse engineering job for the full breakdown, and how schedules typically run for timing.
Projects run under NDA, and we work on the basis that you hold the rights to the design you’re asking us to maintain and are responsible for compliance with applicable IP law.
Where obsolete component replacement gets hard, and where it sometimes doesn’t work
Being straight about this saves everyone a week.
Discontinued microcontrollers are the biggest single obstacle. If the original MCU is gone and you don’t have the firmware, replacing the chip means porting code you don’t possess. Recovering firmware from a read-protected device depends on the specific family and protection state — STM32 parts at RDP Level 1 behave very differently from Level 2, MSP430 devices with a blown JTAG fuse and no BSL password are a different problem again, and lock bits on old 8051 and AVR parts vary by die revision. We assess each case on the actual chip and tell you what’s realistic. No one can promise a given locked part will be readable.
Custom and house-marked silicon may have no path. An ASIC built for one OEM, or a mask-ROM part with proprietary code, sometimes has no functional equivalent. The honest answer there is either source remaining genuine stock or redesign that section around a modern equivalent — which is engineering, not substitution.
Magnetics and electromechanical parts are underrated. A discontinued cust
Working on a board like this?
Send the chip marking or two photos. You get feasibility, lead time and price within 24 hours, and the check costs nothing.
Get a free quote