Skip to content

PCB & manufacturing

Pick-and-Place Files: Why Components End Up Rotated the Wrong Way

Two hundred boards with every electrolytic in backwards, and nobody made a mistake. How origins and rotation references disagree, and how to catch it first.

Carlos Antonio Benito 5 min read

Two hundred boards come back from assembly. Every electrolytic capacitor is in backwards.

The BOM was right. The Gerbers were right. The parts were right. What was wrong was a number in a text file: the rotation column said 0 where the assembler's machine expected 180, because your ECAD tool and their line measure rotation from different references.

Nobody made a mistake in the ordinary sense. The two systems simply did not agree on what zero degrees means, and nothing in the workflow checked.

What a pick-and-place file contains

Also called centroid, CPL, XY data or placement file — the same thing under different names.

One row per placed component:

Designator  X        Y        Rotation  Side
U1          45.720   31.750   0         top
C12         52.400   28.900   90        top
J1          12.700   55.880   180       top
R22         33.020   41.910   270       bottom

Five columns, and four of them can be wrong in a way that is invisible until boards come back.

Coordinates: the origin problem

X and Y are measured from an origin. The file does not always say which one.

Your ECAD tool may use the sheet origin, a drill origin, the board's lower-left corner, or a reference point you set months ago and forgot. If the centroid is exported relative to one origin and the fabricator panelises relative to another, every part is displaced by the same offset.

The symptom is distinctive: nothing is randomly wrong. Everything is wrong by exactly the same amount. If an assembler sends you a placement preview with all parts shifted uniformly, this is why.

Prevention: set the drill/place origin explicitly, usually to the board's lower-left corner or a fiducial, and state it in the release note. Do not rely on the default.

Rotation: why there is no universal convention

This deserves the most space because it causes the most expensive failures.

Rotation is an angle measured from a reference, and there are three independent things that can differ between your tool and your assembler's:

1. Which direction is positive. Counter-clockwise is the common convention in ECAD tools. Not universal.

2. What zero means for a given package. Is a two-pin part at 0° horizontal with pin 1 on the left, or vertical with pin 1 at the bottom? Different component libraries answer differently for the same package.

3. Where the component origin sits. Most placement machines want the body centre. Some footprints — especially ones inherited from third-party libraries — put the origin at pin 1 instead. A part whose origin is at pin 1 will be placed offset by half its body, and rotation will pivot around the wrong point.

There is no standard that resolves these. Anyone who tells you "0° always means X" is describing their own toolchain.

What to do instead: send the assembler your centroid and your assembly drawing together, and ask them to confirm the convention before the first build. Once confirmed for a given tool and assembler, it holds for future jobs — but it has to be established once, in writing.

Which parts actually care

Rotation errors only matter for components that are not symmetric. Worth knowing which, because it tells you where to focus a review:

SensitivityPartsConsequence if rotated wrongly
Criticalelectrolytics, tantalums, diodes, LEDsreverse polarity — damage, often immediate
CriticalICs, connectors with keyingpin 1 in the wrong corner — nothing works
Moderatepolarised connectors, crystals with pin orderassembly or function failure
Nonesymmetric resistors, ceramic capacitors, inductorsworks either way

A board with 142 components may have only fifteen that care. Review those fifteen on the assembler's preview rather than trying to verify all 142.

DNP parts in the placement file

A part marked DNP on the BOM should not have a row in the centroid. If it does, the machine places it.

The two files come from different documents — the BOM from the schematic, the centroid from the layout — so they can disagree silently. Compare the designator lists; the centroid should be the BOM list minus the DNP entries.

Two minutes, and it is one of the few checks that reliably catches a real error.

The review the assembler offers you

Most assemblers will send a placement preview or a first-article inspection before running the full batch. Use it.

It is the only point in the process where a human looks at a rendering of what the machine is about to do, and it catches exactly the class of error this article is about: uniform offsets, rotated polarised parts, DNP positions that should be empty.

Check your fifteen orientation-sensitive parts and the board outline alignment. Ten minutes against the cost of a scrapped batch.

A pre-send check

  • [ ] Origin set explicitly, and stated in the release note
  • [ ] Rotation convention confirmed with this assembler, in writing
  • [ ] Centroid designators = BOM designators minus DNP
  • [ ] Orientation-sensitive parts listed, so they can be reviewed quickly
  • [ ] Pin 1 and polarity shown on the assembly drawing, not only the silkscreen
  • [ ] Footprint origins checked for any third-party library parts
  • [ ] First-article or placement preview requested

When a batch comes back wrong

Diagnose in this order, because it narrows fastest:

  1. Is everything displaced by the same amount? → origin mismatch.
  2. Are only polarised parts wrong? → rotation convention mismatch.
  3. Is one package type wrong and others fine? → footprint origin or library

issue for that package.

  1. Are parts present that should be empty? → DNP leaked into the centroid.
  2. Is it apparently random? → more likely a parts or feeder problem than a

data problem.

The first four are data errors you can fix before the next run. The fifth is a conversation with the assembler.


Getting the centroid right for one build is a matter of checking four things. Keeping it matched to the BOM and the Gerbers across revisions — so that the placement file in a folder belongs to the board revision beside it — is the part that outlives anyone's memory.

In XKEMA the manufacturing package belongs to a PCB revision, and an approved version is frozen.

Hardware documentation BOM, Gerber, CPL and drill files * DNP and DNI components

Where to go next

PCB & manufacturing

PCB Fabrication vs Assembly: Who Needs Which Files?

Fabrication and assembly are usually two different companies doing two different jobs. Which files each one needs, what they share, and where the handover fails.

XKEMA, in one sentence

Engineering knowledge management software for companies that develop electronic products: the product structure, its bills of materials, its documentation and its revisions, all in one place.