Design review checklist
Describe the board. Get the pre-tapeout checklist we would actually run against it — filtered to what applies, ordered by what costs most to miss.
Your checklist
—items
—
How to run a review that finds things
Someone who did not draw the board
This is the entire method. The person who did the layout has already considered every item here and concluded each is fine — which is exactly why they will not see the one that is not. A reviewer with no prior commitment to the decisions finds things the author structurally cannot.
Tick every item explicitly
Scanning a list for problems finds problems you were already worried about. Recording a verdict against every line, including the boring ones, is what surfaces the thing nobody thought about. It is slower and it is the point.
Schematic first, and separately
Most expensive findings are schematic or architecture issues: a missing supply sequence, a regulator that cannot start into its load, a connector with no ESD protection, an update path that was never designed. Finding those after routing means re-routing. Review the schematic when it is complete, before layout starts, and treat that as a distinct gate.
Rank by cost of missing it, not by ease of fixing it
A silkscreen error costs nothing. A missing thermal path costs a respin. An unprotected input on a cable that leaves the enclosure costs a field failure and a recall. The list below is ordered on that axis, which is not the order the items occur in the design.
The findings that recur
- Return path discontinuities. A fast signal crossing a plane split, every time. Free to fix before fab, expensive after.
- Supply sequencing. Rails coming up in an order the silicon does not allow, latching up parts that then run hot.
- Thermal paths that were never designed. A pad-cooled part with no vias under it.
- ESD on anything that leaves the enclosure. Routinely forgotten on debug and service connectors, which are exactly what field engineers plug into.
- Test access. No way to measure the rail that is misbehaving, on a board already in an enclosure.
- Update path. No mechanism, or an unsigned one — and under the CRA this is now a regulatory problem, not just an engineering one.
Design guidance, not a manufacturing instruction. Verify every result against your fabricator's stack-up and the applicable standard before release. Closed-form models are approximations; the fab's field solver and process window are the authority.
Questions
What people ask about this
Why generate a checklist instead of publishing one long one?
Because a 300-item list that mostly does not apply gets skimmed, and skimming is how items get missed. A board with no radio should not be reading antenna keep-out items, and a board with a radio absolutely should. Filtering to what applies is what makes a checklist get used rather than filed.
When should this be run?
Twice. Once when the schematic is complete and before layout starts — most of the expensive findings are schematic and architecture issues, and fixing them after routing means re-routing. Then again before releasing fabrication files, when the layout-specific items become checkable. Running it only once, at the end, catches the cheap problems and misses the costly ones.
Is this what you actually use?
It is the core of it. Our own review adds project-specific items from the requirements, findings from previous spins of the same product, and anything the client has been bitten by before. The generic items here are the ones that apply to almost every board of a given type, which is exactly what is useful to someone else.
How do I use it in a team?
Print it, and have someone who did not do the layout walk it. The value of a checklist is mostly in the second pair of eyes — the person who drew the board has already decided each of these is fine, which is precisely why they will not see the one that is not. Mark every item explicitly rather than scanning for problems; the discipline of ticking is what surfaces the gap.
Want a second pair of eyes that has seen this fail before?
We review one schematic or layout free — written findings ranked by severity, each with a specific remedy. If it goes no further, keep the findings.