The rail nobody enabled
An EN pin held low by a missing strap or an unpopulated 0Ω link. You measure 0 V on the output and start replacing a regulator that was never switched on.
Check: enable-network state before the rail is trusted.
VIANORA · PCB BRING-UP ASSISTANT
Bring-Up Assistant reads your netlist and BOM and writes the power-on plan for that specific board: every rail in order, the probe point, the reading that means pass, the reading that means stop.
Founding pilot · 10 boards · run with you by an engineer · no card
WHY THE CHECKLIST YOU HAVE DOESN'T SURVIVE
Before the board arrives, the plan is one of four things: the memory of the last board, a wiki page written for a different design, a forum checklist written for a different project, or the most expensive resource in the building — the one engineer who has brought up forty boards and is now in every bring-up.
“Expected values are design-specific. The schematic, datasheets, tolerance analysis, and known-good evidence define them.”
That sentence is the whole product. The order of operations, the probe points and the expected numbers all fall out of the design files you already have — they just have to be read, rail by rail, before anyone solders a supply lead to the board.
WHERE IT GOES WRONG
None of these are design errors a DRC catches. They are order-of-operations errors — which is why they are found with a probe in your hand rather than in the CAD tool.
An EN pin held low by a missing strap or an unpopulated 0Ω link. You measure 0 V on the output and start replacing a regulator that was never switched on.
Check: enable-network state before the rail is trusted.
3V3 before 1V8, or I/O ahead of core. A device latches up or back-powers through its ESD diodes. Nothing in the schematic stops you — only the order you switched things on did.
Check: the required order, derived from the rails each device depends on.
A bridge that reads fine at 1 mA and welds a trace at 800 mA. Diode-mode checks miss it. The current limit on your supply is the only thing standing between a repair and a rebuild.
Check: the current limit, set from the board's expected startup load.
A reversed electrolytic, a flipped connector, a QFN rotated ninety degrees. The schematic is correct and the assembly is not, and power is the first thing that finds out.
Check: unpowered resistance from each rail to return, both probe polarities.
You expect 1.2 V, measure 1.18 V and lose a day chasing a regulator that is inside tolerance. Or you expect 3.3 V, see 3.1 V, and miss the fault that actually matters.
Check: tolerance, not a round number — taken from the part's own table.
The MCU boots, the LEDs light, the board "works" — until you find the sensor bus is dead because a pull-up is DNP. Bring-up quietly stops at powers-on.
Check: pull-ups, straps and bus idle levels, before the firmware is blamed.
HOW IT WORKS
Netlist and BOM from KiCad, Altium, OrCAD/Allegro or Eagle/Fusion. Two files. Read-only — we never modify a design, and the pilot asks for no Gerbers and no firmware.
Every rail, its source, its loads, its enable, its reference and its sequencing dependency, read out of the netlist. Not a template with your board's name at the top.
One step at a time: the node to probe, the expected reading and tolerance, the instrument, the pass condition and the stop condition. Nothing to interpret at the bench.
Log what you measured as you go. The next board, the next engineer and the client handover start from evidence instead of "it worked last time".
WHAT YOU GET
One row per step. The expectation is never a round number someone liked — it carries a tolerance and, where it matters, the reading that means stop now.
| # | Node | Probe at | Expect | If it is out of spec |
|---|---|---|---|---|
| 01 | VBUS, before protection | J1 pin 1 → TP1 | 5.00 V ±5%, absolute max 5.5 V | Stop. Reverse polarity or wrong supply — do not proceed to the rail checks. |
| 02 | After input protection | D1 cathode | 4.60–5.00 V, < 30 mA unloaded | Check F1 and D1. A drop here means a clamp is conducting before the board is even on. |
| 03 | 3V3, unloaded | C12 / TP2 | 3.30 V ±3%, startup < 45 mA | Above 120 mA at startup: stop, find the short before it becomes a burn. |
| 04 | 1V8 | C21 / TP3 | 1.80 V ±3%, after 3V3 is stable | 0 V with 3V3 present: check the EN divider R7/R8 — the rail is off, not broken. |
| 05 | MCU core current | JP2 (VBAT link) | 8–25 mA after release from reset | 0 mA: the MCU is not booting. 60 mA+: check for a latch-up path before touching firmware. |
| 06 | I²C pull-ups present | SDA to 3V3, unpowered | 2.2 kΩ–10 kΩ | Open circuit: a pull-up is DNP. The bus will fail later, at 400 kHz, on a customer's desk. |
| 07 | USB enumeration | Host, dmesg / Device Manager | Device appears within 2 s of attach | No device: crystal, strap or VBUS sense — in that order, not firmware first. |
Illustrative rows for an STM32-class four-layer board with a 3V3/1V8 rail pair and USB. Your plan is generated from your netlist; these are examples of the shape, not a template to follow.
WHY YOU CAN TRUST A LINE
A checklist that invents an expected value is worse than no checklist, because you will believe it. So each line is tagged with its evidence, and the tool is allowed to say it does not know.
Rails, enables, straps, pull-ups and test-point access are read from your design files. Deterministic extraction, no inference.
The regulator's or the device's own output table sets the nominal and the tolerance, with the page cited so you can check it in ten seconds.
A previous board, a measured baseline or a lab note overrides ours. Your board beats our table every time.
Where a value genuinely cannot be derived — a sensor's quiescent current, a thermal target, a vendor's undocumented pin — the step says so instead of guessing.
WHO THIS IS FOR
Not for enterprises running Cadence and Siemens review flows, not for people without electronics background, and not for medical, high-voltage or safety-critical products — the plan is a bench aid, not a safety case.
Controllers, sensor boards and power stages in the same small team, with a new revision every few weeks.
Trigger: the board arrives on a Friday and nobody has written the plan.
IoT and connected devices on ESP32, STM32 or nRF — frequent prototypes, no dedicated test engineer.
Trigger: an engineer who owns both the schematic and the bench.
You hand the board to a client and are not there for the first power-on. The plan is part of the handover.
Trigger: the client's own team will power it up without you.
Experimental boards, rotating students, and the knowledge living in whoever graduates next.
Trigger: last year's bring-up notes are gone.
WHAT IT IS NOT
Not a simulator, and not a circuit analysis engine — it does not replace SPICE or a design review.
It does not power your board for you. It tells you where to put the probe and what to expect; you still hold the supply.
No automated measurement rig required: a multimeter and a current-limited bench supply are enough. A scope adds steps, not dependency.
It is not a safety case or a compliance record. Medical, high-voltage and safety-critical designs are out of scope.
THE PILOT
We are not launching a product on this page. We are running ten real projects through the tooling with an engineer on the call, because the only thing that makes this worth building is that it survives your board.
There are no case studies on this page, no client logos and no testimonials, because there are no customers yet. The pilot exists to produce them — and to tell us whether this should be a product at all.
The pilot is free. Starting the paid plan now charges $29/month today, cancels any month, and locks the price for the first ten teams.
QUESTIONS WE ACTUALLY GET
No. The rails, enables, straps, pull-ups and test-point access are extracted from your netlist deterministically — that part is parsing, not generation. Expected values come from the part's own datasheet table or from your measured baseline, and each line is tagged with which. Anything that cannot be derived is marked as needing your input rather than filled in with a plausible number. If a line has no source, it does not ship.
FIGURES AND SOURCES
Lifecycle Insights — PCB design respin study
Cited in Design World Online, "Modern Practices in PCB Design Verification" (2018): the typical PCB project averages 2.9 respins at $44,000 per respin with 8.5 days of schedule impact. Quoted here as published; it is an average across commercial projects, not a claim about your project.
PCBWiki — PCB Bring-Up Checklist: First Power Without Guesswork (2026)
The ordered workflow this product structures: identity and revision, unpowered inspection, planned first power, rails, code, interfaces — with stop conditions at each stage. Their caveat is the product thesis: expected values are design-specific.
KiCad 8 documentation — netlist and BOM export
Netlist and BOM export are documented, standard outputs of the CAD tool. That is the entire integration surface: if your tool can export them, it works here.
Qoitech Otii Arc Pro — source-measure unit for bring-up
A current-limited supply with logging is the instrument class this workflow assumes. The plan sets the limit and the expected startup current; the instrument records what actually happened.
The sample plan rows are illustrative examples on a sample board, not output from a customer design. Bring-Up Assistant is in pilot: plans are produced by an engineer using our tooling while the product is built. Figures are cited as published and may be revised by their authors.
NEXT BOARD
Send the netlist and the BOM. We will send back the plan for that board — in order, with the numbers, and with the stop conditions written down before you apply power.
Get a plan for my boardFounding pilot · 10 boards · free · reply within one business day