EC-24.8 Bring-Up: Finding the Fault, and Proving the Board

What this is and why it exists

The first article of a new design almost never works completely.

How a team handles that first hour decides how many revisions the product needs. Panic and random probing produce a damaged board and no information. A method produces a fault, a cause and a fix.

The method is unglamorous. Power it in a controlled way. Check the supplies before anything else. Work outward from what is known good, and write down every change as it is made.

The vocabulary

• Current limiting — supplying power through a supply that caps the current.
• Bring-up order — the sequence in which parts of a board are proved.
• Bisection — choosing a test that halves the remaining possibilities.
• Rework — physically modifying a manufactured board.
• Cut and jumper — severing a trace and adding a wire in its place.
• Red-lined schematic — a marked-up drawing recording every modification.
• Acceptance test — the measurement showing the board meets its requirements.
• Revision record — the list of faults, causes and fixes for the next version.

The mental model

First power is where boards are destroyed. Apply it through a supply with the current limit set to a little above the expected draw, and watch the current rather than the board. A short circuit then becomes a shrug instead of a burnt trace. Confirm every supply rail before asking any other question.

Bring-up order follows dependency. Supplies, then clocks, then reset, then communication, then function. Each stage depends on the one before, so testing in order means a failure is located rather than merely observed. Skipping ahead produces symptoms that cannot be interpreted, because you do not know what was already broken.

Bisection is the habit that separates the experienced. Rather than probing the most suspicious component first, choose a measurement that halves the remaining possibilities. That finds a fault in a predictable number of steps. Following a hunch feels faster and usually is not, especially on the third hunch.

Rework must be recorded. A modified board is no longer the design, and an undocumented cut means the next revision inherits the original fault. Mark each change on a schematic copy as you make it, not afterwards from memory, because memory reliably omits the third one.

Proving the board is a separate step from making it work. Working is not a specification. The board has to be measured against the requirements written before it existed, and that measurement is what justifies moving to production.

Finally, the revision record. A list of every fault, its cause and its fix turns bring-up into evidence. Without it, the second board fixes what people remember rather than what was actually wrong.

What you should now be able to explain or do

• Power a new board safely and say what to check first.
• Follow a bring-up order and explain why each stage precedes the next.
• Isolate a fault by bisection rather than by suspicion.
• Record a rework so the next revision inherits the fix.
• Distinguish making a board work from proving it meets its requirements.
• Write a revision record that decides what the next version changes.

Check yourself