A first-power-on sequence that finds faults in order
Most bring-up time is lost by advancing past an unverified stage. Here is the order that stops that.
A fabricated board is not a working board. Bring-up is where the design meets reality, and the difference between two days and two weeks is almost entirely procedural.
The principle is simple: never advance past a stage you have not verified. Every skipped stage means that when something fails later, several variables changed at once and you no longer know which one matters.
1. Visual and continuity — before power
Inspect before energising. Solder bridges on fine-pitch parts, tombstoned passives, a connector rotated 180°, a polarised part backwards. Then check continuity on every power rail to ground with the board unpowered.
A short found here costs nothing. The same short found by a bench supply costs a board.
2. Rails, one at a time, current-limited
Bench supply, current limit set just above the expected draw. Bring up the input rail first and measure it. Then enable each downstream regulator individually where the design allows, and measure each output against its design value — not “is there voltage” but “is it the right voltage”.
Note the current draw at each stage. A rail that is correct but drawing three times its expected current is telling you something important, and it is much easier to hear now than later.
3. Clocks and reset
Is the oscillator actually oscillating? Probe it — with a proper probe, because a 10:1 probe on a small crystal can load it enough to stop it. Confirm frequency and amplitude.
Then reset: is it asserted at power-up, and does it release? A processor held in reset behaves exactly like a dead processor.
4. Boot straps — the usual culprit
This is where most stuck boards actually are.
Measure the strap levels at the processor pin, during the window where they are sampled. Not at the resistor, not from the schematic. A stuffing error, a wrong-value pull, or a strap fighting a peripheral that drives the same net all look completely correct on paper.
A processor in the wrong boot mode draws normal current, produces no output, and gives you nothing to work with. It is the most common cause of “the board does nothing” and the easiest to miss.
5. Debug access
Attach JTAG or SWD and halt the core.
This is the pivot point of the entire process. Before it you are inferring from the outside; after it you can read registers, single-step and ask the processor direct questions. If debug will not attach, that is now the only problem worth working on — everything downstream is unreachable until it is fixed.
6. Boot media
Is the bootloader on the device the straps selected, at the offset that device expects? Can the processor read it at all?
Bring the serial console up early — a bootloader printing its version string is the first unambiguous evidence that hardware and software have met.
7. Interfaces, one at a time
Now, and only now, go through peripherals individually against the schematic. Each one gets proven before the next is enabled. Memory first, since everything else depends on it, then storage, then the buses, then the high-speed interfaces.
Why the order is the whole technique
Each stage depends only on the stages before it. When something fails, the fault is in the stage you are on — because everything earlier is verified. That is what turns debugging from a search into a lookup.
The temptation is always to skip ahead: power looks fine, just plug in the console and see what happens. Sometimes that works. When it does not, you are now debugging a system where power, clocks, straps and boot media are all unverified simultaneously, and every measurement is ambiguous.
Keep a log
One line per stage: what was measured, what value, pass or fail, what was changed. It takes a minute per stage and it is the difference between a bring-up report you can hand to a client and a memory of roughly what happened.
That log also becomes the next board’s checklist — and the next spin’s change list.
we bring up boards we designed and boards other people designed. A board that will not boot has a finite list of causes, and working that list in order finds it far faster than probing hopefully. Send us the symptoms.
Questions
Follow-ups
How long should bring-up take?
For a board with no serious faults, first power-on to a booting system is typically one to three days. What turns that into three weeks is almost never a hard fault — it is changing several variables at once and losing track of which change did what.
The board draws current but does nothing. Where do we look?
Boot straps first, then reset, then clocks. A processor sitting in the wrong boot mode draws close to normal current and produces no output, which is the single most common cause of this symptom. Check the strap levels at the pin, not at the resistor — a stuffing error and a wrong-value pull look identical on the schematic.
Do we need a current-limited supply?
Yes. Bringing rails up on a bench supply with the limit set just above expected draw turns a destroyed board into a supply that clicks into constant-current. It is the cheapest insurance in the process.