Circuit Design Risks: The Ones Engineers Keep Hitting

Most circuit designs do not fail because a formula was wrong. They fail because a margin was assumed rather than calculated, because two devices were expected to agree on a level that neither specified clearly, or because a part that was available during development disappeared before production.

Those are the recurring circuit design risks, and they are recognisable because they appear in almost every post-mortem. The useful response is not to work harder on the schematic; it is to add specific reviews that target them at the point where they are still cheap to fix.

Power Supply Margin and Load Behaviour

Power design is where optimistic assumptions are most comfortable. A regulator is selected on its nominal output, and the calculation stops before the worst case load, the worst case input voltage and the worst case temperature are combined.

Power supply margin has three components that are frequently treated separately. The first is steady state: the load at maximum activity, with every rail at its tolerance limit and with the derating the application requires. The second is transient: the response to a load step, including the voltage excursion and the time to recover, compared with what the supplied devices tolerate. The third is thermal: the dissipation in the regulator and in the surrounding copper at the ambient temperature inside the enclosure, not on the bench.

The failure mode when any of these is neglected is a product that works in the lab and resets in the field, or one that behaves correctly at room temperature and fails after an hour of operation. Adding the three calculations to the design review costs an afternoon.

Interface Compatibility Between Devices

Interface compatibility problems occur where the specification is silent or where two suppliers interpret a standard differently. A device that guarantees a minimum high output level, connected to a device that requires a minimum high input level, leaves a margin that is often smaller than expected and sometimes negative.

The check should be done numerically, using the guaranteed levels at the relevant supply voltage and temperature, and it should include the termination and any series components. Level shifting, open drain buses, differential pairs and clock inputs each have their own conventions, and a design that mixes them needs the calculation written down.

Timing is part of the same review. Setup and hold times, the rise time of a bus, and the propagation delay through a level translator can combine into a timing failure that no amount of firmware adjustment will fix.

Component Lifecycle and Availability

Component lifecycle is not a purchasing concern that appears after the design is complete. A part selected because it fits the footprint and the budget may be obsolete within a year, and the cost of a redesign during production is far higher than the cost of choosing a different device at the start.

The review should classify every part by availability risk: the number of sources, the lifecycle status, the lead time and the consequence of a shortage. Single sourced parts that are essential to the function are the ones that need an approved alternate or an early purchase commitment.

This is also where component lifecycle checks pay for themselves, because a substitution proposed during development is a design decision, while the same substitution during production is an engineering change.

Verification Gaps and the Design Verification Plan

The most expensive risk is the one that is never tested. A design verification plan names what will be measured, under which conditions and against which limits, and it turns assumptions into evidence.

The plan should cover the conditions that the product will actually see: the supply extremes, the temperature range, the interfaces in their worst case configuration and the load conditions that stress the power design. Each test should have a criterion written before it is run, so that a marginal result is recognisable as marginal rather than acceptable.

It is also worth planning the negative tests. A device that must respond to a fault should be exercised with the fault present, and a protection circuit should be tested with the transient it is supposed to absorb. Those tests are the only way to know that a protection feature exists in practice rather than in intention.

Working with a partner that measures rather than observes helps here. A design verification plan executed during the prototype stage, with the results attached to the revision that was built, converts a design review from opinion into evidence.

Reviews That Catch These Risks

• A power review that combines worst case load, input and temperature, including transient response.
• An interface review with guaranteed levels and margins written down numerically.
• A lifecycle and availability review for every part, with alternates named.
• A verification plan with pass criteria defined before measurement.
• A mechanical review confirming the thermal path and the connector access.
• A manufacturing review that confirms the design can be built at the intended volume.

Each of these reviews belongs to a stage rather than to a person. They work when they are scheduled, and they are skipped when they are optional.

Building the Marchand Plan Into the Schedule

A review only changes an outcome if it happens before the layout is committed. The practical way to guarantee that is to attach each review to a milestone that already exists, rather than creating a separate approval step that will be skipped when the schedule tightens.

Put the power and interface review at the point where the stack-up is fixed. At that stage the current budget, the connector pinout and the level compatibility are all still editable in a document, and a change costs a comment instead of a board revision. Check the power supply margin against the worst case load and the worst case input, and confirm that every interface has a single owner who has stated the voltage thresholds in writing.

Attach the lifecycle check to the bill of materials release. That is the moment when purchasing can see the parts, when a second source can still be qualified without delaying the build, and when an obsolete component can be replaced while the schematic is still open. Waiting until production releases the order turns a short engineering task into a schedule problem, which is exactly how an avoidable component lifecycle issue becomes the largest line in the post-mortem.

The verification plan belongs with the test fixture, because the plan and the fixture have to agree. List the measurements, the limits and the conditions before the fixture is designed, so that the fixture measures what the plan requires rather than the reverse. When the design verification plan is written first, the test points appear on the layout as a requirement instead of being added as a favour at the end.

None of these reviews needs a new meeting series. Each one needs a named owner, a defined input and a recorded outcome, and each one has to be closed before the next milestone opens. A review that can be deferred is a review that will be deferred, and a deferred review is indistinguishable from a risk that was never examined.

FAQ

Which risk causes the most schedule loss? A component that becomes unavailable late, because the substitute has to be verified and the documentation changes.

Is a design review the same as a verification plan? No. The review examines the design; the plan produces measurements that confirm it.

How much margin is enough? Enough that the worst case combination still meets the requirement. The calculation, not the habit, determines the number.

What should be reviewed first on a new design? The power architecture and the interfaces, because they are the most difficult to change once the layout exists.

Summary

Circuit design risks concentrate in four places: power supply margin, interface compatibility, component lifecycle and the absence of a design verification plan. Review them deliberately at the stage where a change costs a comment rather than a board revision, and the design has a chance of working the first time.