Direct Part Marking and Why It Is Usually Not a QR Code

A stainless coupling 12 mm across needs an identifier that survives passivation, ten years in a plant room and a wipe with solvent. No label lasts that long, and there is nowhere to put one. The mark has to be in the metal.

Direct part marking puts a machine-readable code onto the item itself rather than onto something stuck to it, and almost every rule from printed codes changes. Contrast stops being a property of ink, module size stops being a print resolution question, and the symbology usually stops being QR. Micro QR and rMQR opens on this exact situation, a serial in a few millimetres on a circuit board, and the smaller QR variants are one answer. Rarely the first one.

The mark has no colour

A printed code works because dark ink absorbs light and light substrate reflects it. A direct part mark has no pigment. What it has is a difference in surface texture, oxide layer or depth, and whether that reads as dark or light depends entirely on how you illuminate it.

Take a dot-peened mark: small conical indentations punched into metal. The metal inside a dimple is the same metal as the surface around it. Under diffuse light from above the dimple looks slightly dark, because it shadows itself. Under low-angle light from the side, the rim catches the light and the dimple reads bright while the flat surface goes dark, so the polarity of the whole code inverts.

So a reader here has to handle both polarities, which is why industrial imagers decode light-on-dark marks that a phone camera app refuses outright; inverted codes explains why that refusal is normal on the consumer side. The same physical mark can be an excellent code or an unreadable one depending on nothing but the lamp.

Three ways to make the mark

Laser. The most controllable and the most common. On metal, a fibre laser can ablate material to leave a rough matte area, or anneal the surface to grow a dark oxide layer without removing anything. Annealing is the choice for surgical instruments and food equipment, because it leaves no crevice to trap contamination and survives repeated autoclaving. On plastics, foaming lightens and carbonising darkens depending on the polymer.

Dot peen. A carbide stylus driven pneumatically or electromagnetically, indenting one dot at a time. Rugged, cheap, tolerant of rough cast and forged surfaces, and coarse: module size is the stylus diameter and the cells come out round. That rules it out for small marks, and the impact stress rules it out for thin-walled parts.

Chemical and electrochemical etch. A stencil plus electrolyte and current, or a photochemical process, removing a shallow layer to leave a matte recess. Common where thermal input must stay near zero. Slower per part, and it needs a stencil per design, which pushes it toward fixed rather than serialized identifiers.

Why Data Matrix usually wins, and when QR still does

The smallest ECC 200 Data Matrix symbol is a 10 by 10 grid holding six digits, and it needs a quiet zone of one module on each side, giving a footprint of 12 by 12 module widths. The smallest QR symbol is version 1 at 21 by 21, and the four-module quiet zone brings it to 29 by 29. At an identical module size that is 2.4 times the width and 5.8 times the area for the QR version, and on a 12 mm coupling that is the whole decision. Data Matrix also loses fewer cells to finder geometry, needing one L-shaped border and a clocking track rather than three 7 by 7 finders.

Two further reasons are institutional. Data Matrix ECC 200 is the symbology named in the marking standards industrial buyers work to, including MIL-STD-130 for United States defence item marking and the unique device identifiers on medical equipment, so readers, verifiers and quality specifications already exist for it. And its readers assume difficult surfaces, where consumer QR assumes paper. The wider symbology comparison covers the rest of the field.

QR earns its place in one common situation: when a human with an ordinary phone has to read it. Service technicians in the field. A machine nameplate somebody photographs to get the manual, which is what maintenance codes are built around. Warranty registration by an end customer. Nothing else here is read natively by a stock camera app. There is a payload argument too: for a URL rather than an identifier, QR's byte-mode capacity beats squeezing the same string into a small Data Matrix, and for a structured product identifier in retail, GS1 Digital Link is a QR-based scheme with an existing ecosystem.

Micro QR sits between the two. One finder instead of three and a two-module quiet zone instead of four bring its largest version down to a 21 by 21 footprint, the size of a bare version 1 QR grid before that code's own margin. It carries short identifiers only, and reader support is a per-device, per-firmware question, so it belongs in a closed loop with known hardware and nowhere a phone is involved.

Lighting decides everything

More read failures come from illumination than from anything else.

• Diffuse dome light for shiny and curved surfaces, flooding the mark from every angle at once and killing the specular hot spot a polished part throws back at the camera.
• Low-angle dark field for engraved and peened marks on flat surfaces. Light grazes the surface, flat areas reflect away from the lens and go dark, the mark scatters and goes bright. Inverted polarity, strong contrast.
• Coaxial or bright field for matte-on-shiny marks such as laser ablation on polished steel, where the flat surface returns light to the lens and the rough mark does not.

A fixed-mount camera changes every rule because it fixes all of those at once: distance, angle, focus, exposure, illumination, and exclusion of ambient light. You stop designing for the worst case a stranger might create and design for one geometry you guarantee. Minimum module size drops to whatever your optics resolve, usually expressed as pixels per module, with three the common working figure and five preferred for anything graded.

Circuit boards have their own rules

A mark on a board has to survive reflow at up to 260 degrees, possibly wave solder, and a cleaning step. Paper labels do not; polyimide labels do. Laser marking directly onto the solder mask works well, usually at an ultraviolet or green wavelength chosen so it marks the mask without cutting into the substrate.

Placement is constrained in ways artwork is not. The mark has to sit in a keep-out area with no copper beneath it, stay visible after components are placed, and be flat enough for a fixed camera at a known working distance. Contrast on a green or black solder mask is better than expected, which is why board marks are frequently light-on-dark by design, and why a phone is the wrong thing to test them with.

Verify to a grade, not to a scan

A successful read is not the acceptance criterion here, and that is the habit separating direct part marking from ordinary printing. A mark that scans today on your one reader can be marginal in every measurable respect, and it will fail next year on a different unit under different light.

Verification is a calibrated measurement against defined parameters, reported as a grade: contrast, modulation, grid uniformity, damage to the fixed patterns, and how much error correction was consumed. ISO/IEC TR 29158 defines that methodology for direct part marks specifically, because ISO/IEC 15415, the general print quality standard for 2D symbols, assumes a paper-like surface.

Set a minimum grade in the specification, verify to it on the real substrate before the design is frozen, and sample against it in production. A mark that grades well has margin. A mark that merely scans has none.