Bring-up method

How a board goes from a box of bare PCBs to a closed log. The checklist is the terse form stuck to the bench; this page is the reasoning behind it, which is the part that keeps you from skipping a step the one time it mattered. Current checklist revision is 3.

Two rules above the steps

• Never raise the limit to make it work.
• Write the number down before you change anything.

The first rule is the one people break under deadline: raising a current limit or a timeout until the board "works" hides the fault instead of finding it, and the board fails later where it is expensive. The second rule is why every log has numbers in it: a value written down before a change is evidence; a value remembered after is a story.

The shape of a session

Cold checks first, with the supply disconnected: inspection under magnification, then resistance from every rail to ground. Nothing is powered until resistance is sane, because powering a shorted board turns a cheap problem into a burnt one. Then first power, current-limited and minimally populated, watching the rail order and timing on the scope with all channels at once.

Only after the rails are trusted do the active checks start: clock at the pin that consumes it, reset behaviour, straps pin by pin against the schematic, then the loader and the console. A board that gets a clean console has earned the peripheral walk, one peripheral at a time, each with its current draw noted. Thermal pass in the enclosure to the claimed maximum, then an unattended soak of at least 24 hours in the enclosure, because free air is not the product.

The last step is the write-up, before the bench is cleared: tables filled, dashes where a value was not measured, outstanding items named. A board whose log is not written is not closed, however well it ran.