← All articles
HardwareMCUDesign

Choosing an MCU that survives the next redesign

Second sources, pin-compatible families, and why the cheapest part is rarely the cheapest product.

The chip shortage that started in 2021 caught most embedded teams off guard. We had clients whose production lines stopped because a 40-cent microcontroller went from eight weeks lead time to 52 weeks, at 10× the price. The lesson wasn't "order more safety stock" — it was "design in parts that can be replaced."

The second-source question

Before you commit to an MCU, ask: if this part disappears tomorrow, what drops in?

For some families the answer is easy. The STM32F4 and STM32L4 families have binary-compatible siblings from multiple vendors. You can often swap in a GD32 or an APM32 with a recompile and a resistor change — no board re-spin required.

For others — particularly some Arm Cortex-M0+ and proprietary-ISA parts — there is no second source. If you pick one of these, you own that supply chain risk.

Pin-compatible families

The strongest protection is a pin-compatible upgrade path within a single vendor's lineup. STMicroelectronics does this well: the STM32G0, G4 and L4 families share enough pinout overlap that you can often step up to a part with more flash or a faster core with only a footprint change and no schematic revision.

On Aurora we originally used an STM32L433. When a 240 kB binary outgrew its 256 kB flash, we swapped to an STM32L476 — same LQFP48 package, same pinout, rebuilt the firmware the same afternoon.

The cost argument

The "cheapest part" calculation usually omits the cost of a board re-spin when the part is discontinued. A board re-spin for a four-layer PCB runs to about €8k in engineering time and €3k in prototype tooling. If a €0.30 saving per unit drives you into a dead-end part and you ship 10,000 units, you're trading €3k for €8k+ of risk.

Our default position: pick a part from a family with an active roadmap, at least one second source, and a vendor with a track record of long-term supply commitments (10+ years is a meaningful signal). Pay the extra €0.50 per unit. It is almost always the right call.

A practical checklist

Before committing to an MCU:

  • Is there a pin-compatible upgrade path in the same family with 2× the flash?
  • Does the vendor publish a product lifetime commitment (longevity program)?
  • Is there at least one second-source or pin-compatible alternative from another vendor?
  • Is the part available from two or more distributors today with > 10k units in stock?

If you can't answer yes to all four, the part needs a stronger justification.