Swapping an out-of-stock part without missing a pin

A part on your bill of materials is out of stock, or its last-time-buy date is on the calendar. The distributor shows a similar one in the same package. You can place the order in five minutes. The risk is a board that powers up, then fails in a way nobody connects to the swap.

This comes up constantly. One recent forum thread described a replacement LDO in the same SOT-23 package. Pins 2 and 3 were swapped, and the quiescent current was very different. The board design tool’s design rule check flagged neither, because it checks your layout against your netlist, not one part against another. The replies were sensible: keep both datasheets open side by side, treat a new manufacturer part number as a new schematic symbol, and keep a written checklist. This note is that checklist, with the reasons behind each line.

Same package does not mean same part

A matching package name tells you the outline is similar. It does not tell you the pin assignments, the thermal pad connection, or the land pattern the manufacturer recommends. Read the pin table and the package drawing of the replacement, not the old part’s. Do not assume a wider footprint will take a narrower package, and do not assume two package families share a pinout because they share a name.

If the pinout differs, you have a real choice:

• Re-spin the board. Cleanest, slowest, and the only honest answer when several pins move.
• Use a small adapter. A carrier board that maps the new pinout onto the old footprint. Fine for a short run or a rework, awkward in production, and it changes height and thermal behavior.
• Choose a different candidate. Often a second search turns up a part that really is pin compatible.

Whichever you pick, make a new symbol and a new footprint from the new datasheet. Editing the old symbol’s part number is how the pin 2/pin 3 mistake gets in.

The electrical checks that rarely show up in a parametric search

Distributor filters sort on headline numbers: voltage, current, package. The failures come from the numbers below the headline.

Quiescent current. On a battery product, an LDO or supervisor with ten times the ground current can erase your sleep budget. If a board’s standby current matters, compare the figure at your actual load and temperature, not the typical value on page one. Our note on why sleep current beats the datasheet covers how to measure what you really get.

Output capacitor requirements. Linear regulators differ a great deal in what they need on the output. Older designs depend on the capacitor’s equivalent series resistance to stay stable. Texas Instruments explains how the ESR range affects stability in ESR, Stability, and the LDO Regulator, and shows in Ceramic Capacitors Replace Tantalum Capacitors in LDOs that a low-ESR ceramic can leave such a regulator unstable unless the ESR is restored. Newer regulators may tolerate ceramics, but that is a property of the specific part, so read its output capacitor section. Also check the effective capacitance at your operating voltage, because ceramic capacitance falls with DC bias. If the old part wanted a tantalum and the new one wants a ceramic, or the reverse, the capacitor on your board may be wrong for the replacement. Our note on substituting a ceramic cap covers the capacitor side.

Absolute maximum and recommended operating ratings. Input voltage margin on a rail that sees transients is the usual trap. A part rated a few volts lower may run fine on the bench and fail on a hot-plugged supply.

Enable, reset and power-good behavior. Polarity, thresholds and whether the pin has an internal pull-up can all differ. A floating enable pin that happened to default high on the old part can leave the new one off.

Startup and sequencing. Soft-start time, inrush and the order rails come up can change. A microcontroller that expects a rail within a time window will notice.

Switching regulators. Frequency, inductor value and saturation current, feedback divider and compensation all change with the part. Do not assume the old inductor and capacitors suit the new controller.

Footprint, thermal and assembly

Compare the land pattern the manufacturer recommends for the replacement against the one on your board. A recommended pattern is a manufacturer or library convention; your assembler may have its own requirements, so ask before assuming. If the new part has an exposed pad the old one lacked, or the reverse, copper and vias under it matter for heat and for soldering.

Check moisture sensitivity level and packaging. A replacement with a stricter moisture rating changes how parts are stored before reflow, as in our note on parts that sat open needing a bake. Check the reel or tray format with whoever places the parts.

Check orientation. Pin 1 marking can sit in a different corner on a different manufacturer’s part, and the assembly program has to follow it.

Is it really out of stock, or about to be?

Two different situations need different responses.

A shortage is temporary. A second source, a distributor with stock, or a wait may fix it, and you should keep the original part in the design.

An end-of-life notice is permanent. Manufacturers publish discontinuance notices with dates for last orders and last shipments. Microchip’s policy document describes how those notices work and refers to the JEDEC product change notification standard for what counts as a change in form, fit or function. The ECIA position paper on product change notifications recommends that same standard for component change notices. Read the notice itself for the dates. If the date is a last-time-buy, you can buy a lifetime supply if cash and shelf life allow, which buys time to redesign properly instead of swapping in a hurry. For a microcontroller the stakes are higher; see when your microcontroller goes end-of-life.

A short ECO checklist

Fill this in per candidate, in writing, with a datasheet page reference for each line.

1. Pin table compared pin by pin, including exposed pad and no-connect pins.
2. Package drawing and recommended land pattern compared.
3. Absolute maximum ratings against worst-case rail voltage and transients.
4. Quiescent or supply current at your load and temperature.
5. Output or compensation components required by the new part, against what is on the board.
6. Enable, reset and power-good polarity and thresholds.
7. Startup behavior and sequencing against anything downstream.
8. Moisture sensitivity, packaging and pin 1 orientation for assembly.
9. Firmware-visible differences: registers, ID codes, timing, default states.
10. Whether the part is itself single-source or near the end of its life.

Prove it on the bench

A paper check finds most problems. It does not find all of them. Build one board, then measure: the rail on a scope during power-up and under a load step, standby current across a full sleep cycle, and the function that depends on the part at temperature extremes if your product sees them. Keep the old part’s measurements for comparison. If the board is already built in volume, fit the replacement on a small lot first and run your normal functional test, so that a difference shows up on a fixture and not in the field.

Record the change. Update the schematic, the bill of materials, the approved-vendor list and the revision, so the next engineer sees why the part changed instead of reverting it.

How we approach it

We compare the candidate against the schematic, layout and firmware as a set, flag what has to change, and say plainly when a swap is not worth doing. Send us the board files or bill of materials, the part that is going away, and any notice you received.