Every board house receives the same thing from you: a set of files. Not your design, not your intent — files. The checks below are the twelve verifications that separate a set that will come back as the board you designed from a set carrying a silent defect, and every one of them is a concrete action in Gerbv, the free viewer for Gerber RS-274X files, Excellon drill files and CSV pick-and-place data. Run the list in order, on the extracted set, exactly as it will be uploaded. It takes minutes; a respin takes weeks. New to the viewer? How to open and view Gerber files covers the mechanics, and Gerbv is on the download page.
Completeness: the set itself (checks 1–3)
1. Every expected layer is present. Open the extracted folder in Gerbv and count the loaded layers against your board’s structure: copper for every copper layer, mask and silkscreen for the sides you use, paste if a stencil is involved. A missing file is the most common export accident and the least visible — nothing anywhere reports it.
2. The board outline exists, is closed and is the right size. Display the outline layer alone: one unbroken shape. Measure its extents as in measuring distances in Gerbv and compare them with the mechanical design. An absent or open outline leaves the factory guessing at your board’s edge.
3. The ZIP re-extracts to the set you checked. If you will upload a ZIP, extract a fresh copy of that exact archive into a new folder and open it. What loads is what the board house gets. This test catches the wrong archive, the stale revision and the file that never made it into the ZIP.
Geometry and registration (checks 4–7)
4. Layers are stacked and aligned. Build the stack as in loading and aligning Gerber layers — copper, mask, silkscreen, paste — and confirm features register: mask openings centred on pads, paste openings matching their pads, nothing shifted relative to everything else.
5. Drill hits land centred in their pads. Overlay the Excellon file as in viewing drill files, across the whole board, on both copper sides. A block offset, a miniature pattern in one corner, or holes uniformly the wrong size each point at an export setting to fix — before ordering, not after.
6. Units and scale are consistent. Measure one known dimension — the outline, a connector pitch — and confirm it is the designed value, not a value scaled by a unit conversion. Then confirm a second dimension far from the first: consistent smallness or largeness that grows with distance is the signature of a format or unit mismatch in the export.
7. Bottom layers are not mirrored. View the bottom copper and bottom silkscreen as the factory will use them. Mirrored data betrays itself through asymmetric features sitting on the wrong side and any bottom-side text reading backwards. If your workflow presents bottom layers as seen through the board, know exactly which convention your files use before uploading.
Readability and manufacturing sanity (checks 8–12)
8. Silkscreen does not print over pads. Show silkscreen over the mask and pads: ink belongs on the board surface, not across exposed pads it would contaminate. Check designators are complete, legible and the right way up.
9. Minimum clearances survive visually. Zoom into the tightest routing — between pads, around fine-pitch parts — and measure the smallest trace width and the smallest gap. Gerbv measures; it does not run design-rule checks, so this is your confirmation that what you routed is what was exported, comfortably inside what your board house can etch.
10. Apertures render as expected. Scan each layer at working magnification for shapes that render wrong: pads that became outlines, flashes that vanished, pours with ragged edges. Gerbv’s aperture-usage reporting helps here — the set of apertures and their counts should match the board you think you exported.
11. Nothing lives outside the outline. With all layers visible, look beyond the board edge: stray copper, a silkscreen legend, or a drill hit outside the outline is either data that will be clipped unpredictably or a symptom of a wrong-origin layer in the set.
12. The checked set is the uploaded set. Note the folder and archive you verified — name, date, revision — and upload exactly that. If anything changes after checking, re-export the full set and re-run this list; partial re-exports are how checked and shipped sets quietly diverge.
Where to order
Once the set passes, any board house that accepts standard Gerber uploads can build it. Three widely used options, with the referral relationships stated plainly: JLCPCB (referral link — this site may receive coupons when new users register and complete a first order; coupons offset order amounts and cannot be redeemed for cash); NextPCB (referral link — if you place a first order, this site receives a benefit as site credit that can only be spent with NextPCB, not cash); and PCBWay (plain link, included on its merits). For regional manufacture, OSH Park in the United States and Eurocircuits in Europe are long-established options. Using a referral link never changes the price you pay. Full details are on our Affiliate Disclosure page. Destination-specific walkthroughs: checking before a JLCPCB upload and checking before a PCBWay upload.
Frequently asked questions
How long does this checklist take?
Minutes, once the layer stack is built — the checks are mostly toggling layers you have already loaded, plus two or three measurements. It grows faster with practice; the first run on a new board design is the slowest and the most valuable.
Can Gerbv run these checks automatically?
No. Gerbv displays and measures fabrication data and reports statistics about the codes in use, but it has no design-rule or DFM checking. The checklist is a guided visual and numerical pass — which is also why it catches wrong-file and wrong-revision problems that automated rule checks cannot see.
Which check catches the most problems?
In practice, the completeness checks (1–3) and the drill overlay (5): a missing layer, a missing outline and a mismatched drill file are the failures that most often reach factories, and all three are unambiguous in the viewer.