Before releasing SMT placement files, confirm three things with the assembler: what each X/Y pair represents, how the board coordinate system is defined, and how component rotation is interpreted on each side. Then compare an imported placement preview with the released assembly drawing. A CSV that opens correctly can still put the right component at the wrong position or orientation.
The review below is for the engineering handoff, before production programming and first-article approval. It helps expose mismatched export and import assumptions. It does not establish placement accuracy, validate a machine program or replace package-specific assembly instructions.
1. Identify what the coordinate fields actually mean
Request the assembler’s expected file format before choosing export columns. Useful handoff fields usually include reference designator, position, rotation and board side. Record the units in a header or accompanying note, and explain which position fields the recipient should import. Avoid sending several coordinate pairs without naming the intended pair.
For example, Altium’s official assembly-output documentation distinguishes center coordinates, footprint-origin coordinates and pad-1 coordinates. These describe different reference points. For an asymmetric footprint, choosing a different pair can change the imported location even when the board origin and units are correct.
Use a connector or another visibly asymmetric footprint as a review sample. Compare its chosen reference point in CAD with the point shown in the imported preview. If the positions disagree, establish which fields were used before changing coordinates. Moving every row by the same amount may hide the error for one component while making others incorrect.
Keep footprint identity tied to the approved manufacturer part number and BOM. A name such as “QFN32” does not explain pin numbering or a library’s zero-degree orientation. Include the assembly drawing and relevant package drawing so the reviewer can judge the intended orientation independently of the numerical angle.
2. Compare three landmarks to separate origin and scale problems
Define the exported origin relative to the board outline or another unambiguous released feature. A screenshot alone is insufficient if it omits the origin marker or drawing orientation. State whether the file describes an individual board or a panel, and whether the supplier will apply any panel-level transformation.
In the KiCad 9.0 placement-export documentation, the selected origin can be the drill/place origin or the page origin, and location units are configurable. Record the actual export settings for your release; another software version may present them differently.
Choose three separated, non-collinear landmarks that can be identified in both datasets. These can be agreed component reference points or dedicated alignment features, provided their coordinates use the same defined reference. Compare the board outline and the imported locations together. One matching point cannot reveal every scale, mirror or rotation error.
An illustrative offset check
Assume a hypothetical 50 mm × 30 mm board and an agreed top-view coordinate system, with X increasing right and Y increasing up. The three landmarks below are illustrative review points, not fabrication requirements or actual production measurements. The preview has been expressed in that same system.
| Landmark | Released X/Y, mm | Imported X/Y, mm | Difference, imported minus released |
|---|---|---|---|
| A | 5 / 5 | 15 / 9 | +10 / +4 mm |
| B | 45 / 5 | 55 / 9 | +10 / +4 mm |
| C | 5 / 25 | 15 / 29 | +10 / +4 mm |
All three differences are the same, while the distances between landmarks remain unchanged. That pattern is consistent with a translation in this simplified example. Check the export origin and importer offset first. It is not proof that every component is correct: the wrong coordinate column or a package-specific issue can still affect other rows.
If separations change by a common factor, inspect the units and importer scaling. If one axis reverses direction, inspect the view and mirroring assumptions. If only a few footprints disagree, review their coordinate reference points and library definitions. These patterns help select the next check; they should not become an automatic coordinate-repair recipe.
3. Treat bottom-side coordinates and viewing direction separately
Write down whether a bottom assembly drawing is shown from the bottom or viewed through the board from the top. Ask the recipient to confirm the preview’s viewing direction too. Two correct representations can appear mirrored when they use different views. A screenshot marked only “bottom” leaves that distinction unresolved.
Export settings can also change the numbers, independently of how a drawing is displayed. KiCad 9.0 documents an option that negates bottom-layer X coordinates. Altium’s assembly-output page describes a bottom-side Y-flip option affecting Ref-Y and Pad-Y values. These are software-specific settings, not interchangeable instructions to mirror every bottom component.
Agree a single transform with the assembler and check that it is applied once. An already-transformed export followed by an extra importer mirror can defeat the intended handoff. Verify an asymmetric bottom-side component against its numbered pads and drawing, then check a second component elsewhere on that side. Record both the selected coordinate fields and the transformation used.
4. Resolve rotation against pin numbering, not a visual guess
A rotation number needs a reference orientation and a direction convention. Obtain the supplier’s interpretation, including bottom-side treatment. Compare a polarized diode, an IC with a clear pin-1 reference and an asymmetric connector. A nonpolarized rectangular passive is a weak sole check because a half-turn can look equivalent.
An illustrative rotation mismatch
Suppose a hypothetical U1 is exported at 0°, while the released top assembly drawing places pin 1 at the northwest corner. The supplier preview puts pin 1 at the northeast corner. The placement center matches the intended point. This example establishes an orientation disagreement; it does not establish that adding or subtracting 90° from every row is safe.
First verify U1’s manufacturer part number, pad numbering, footprint reference orientation and imported side. Then have the supplier document how that footprint maps to its component definition. Approve the corrected preview using the pin locations. Check another instance of the same approved footprint and representative different footprints before accepting any broader mapping change.
If a correction is required, regenerate controlled outputs or document a supplier-side programming correction with its exact scope and approval. Keep the original release and the approved correction record. An undocumented spreadsheet edit is difficult to reproduce when the same design is reordered or moved to another assembler.
5. Release the checked package as one manufacturing handoff
Reconcile placement rows with the fitted BOM and assembly variant. Decide how DNP items, fiducials and non-SMT parts should be represented. An omitted row may be intentional, but that intent needs to be visible. KiCad’s documented exclusion options illustrate why an export’s contents should be checked against the release rather than assumed complete.
Use this practical signoff list:
- Identify the board revision, assembly variant, CAD version and export date.
- Specify the imported X/Y columns, units, origin and board-versus-panel scope.
- Confirm top/bottom labels, drawing views and any bottom-side transformation.
- Compare separated landmarks against the released outline and intended reference points.
- Verify orientation with pin numbering for representative polarized and asymmetric parts.
- Reconcile fitted and omitted rows with the BOM, recording exceptions.
- Save the reviewed preview, resolved questions and approval owner with the release.
This review belongs inside the broader PCB assembly RFQ package, alongside material ownership and acceptance requirements. Use the SMT engineering guide to organize related process questions and the SMT assembly scope to define the build. Placement data establishes intended locations; the actual assembly still needs its agreed setup verification, first-article inspection and testing.