How to choose 2D code readers for automotive traceability

Automotive traceability lives on the part itself: an ECC 200 DataMatrix laser-etched into a knuckle housing, a QR label on a battery module, a VIN dot-peened on a rail. Start from the marking drawing, not a data sheet: design for 3-5 pixels per module — a 10 mil (0.254 mm) module needs a pixel of 0.085 mm or less at the part — size the field from the mark plus placement tolerance, then let focal length, lighting angle and strobe time follow. This guide is for the engineer who has to pick the fixed code reader that reads those marks every cycle. You will finish with a design you can order against and an acceptance test to run on real parts.

Label or direct part mark: which code are you reading?

Decide this first, because it sets every later choice: a printed label code reads with almost any reader under a plain white light, while a direct part mark on bare metal needs glare-controlled lighting and a decoder tuned for low contrast.

Most automotive marking specifications call for ECC 200 DataMatrix — laser-etched or dot-peened on powertrain and chassis parts, printed on labels for modules and packaging. Get the marking drawing before anything else: it fixes the symbology, the symbol size in modules, the module pitch (10 mil, 0.254 mm, is a typical DPM pitch) and the marking method. Every 2D imager decodes the common symbologies, so the code type rarely decides the purchase; the surface and mark quality do. An etched mark is low contrast on a reflective surface — a lighting and optics problem, which the next sections size. For label codes at a fixed stand-off, the MV-ID3013PM-12M-WBN is the compact shape of that job: 1.3 MP with a 12 mm lens.

How many pixels per module does a 2D code reader need?

Design for 3 to 5 pixels per module: 2 px/module is the survival floor where decode confidence collapses, and below it no amount of decoder tuning buys the reads back.

The 2 px/module floor is long barcode practice; the 3-5 target covers focus drift, edge rounding from etching and part-to-part contrast spread. For a 10 mil module (0.254 mm), 3 px/module needs a pixel of 0.254 / 3 = 0.085 mm or less at the part. Across the 2560 pixels of a 5 MP sensor that permits a field up to 217 mm wide — at automotive module sizes, sampling almost never limits the read. Contrast, focus and lens sharpness do, so spend the pixels on a tight field and a stand-off inside the lens's rated range rather than on megapixels.

What field of view and working distance should you design around?

Size the field from the mark plus part-position tolerance, then let the focal length follow: a 25 mm pad with ±10 mm placement tolerance wants a 50 mm field, which is a 13 mm lens at 100 mm working distance on a 1/1.7 in sensor.

Worked example, one line per decision:

• Mark: ECC 200, 16 × 16 modules at 10 mil pitch → 16 × 0.254 = 4.1 mm across.
• Field: 25 mm pad + 2 × 10 mm tolerance = 45 mm, designed up to 50 mm.
• Sensor: 1/1.7 in format, about 7.5 mm wide, 2560 pixels across (5 MP class).
• Magnification: m = 7.5 / 50 = 0.15.
• Focal length: f = 100 mm × 0.15 / 1.15 = 13.0 mm; the nearest common lens, 12 mm, lands the same field at WD = 12 mm × 1.15 / 0.15 = 92 mm.
• Pixel at the part: 50 mm / 2560 = 0.0195 mm → 0.254 / 0.0195 = 13 px/module, four times the target.

A 5 MP global-shutter reader with an 8 mm lens covers this exact design:

The standard 8 mm lens reaches the same 50 mm field at 8 × 1.15 / 0.15 = 61 mm working distance. Part height is the other constraint: stopping down from f/4 to f/8 roughly doubles the depth of field, and past about f/11 diffraction gives the gain back on pixels this small.

Which lighting reads an etched DataMatrix on bare metal?

Light the mark by geometry, not by brightness: low-angle bars raking the surface at roughly 30° turn etched and dot-peen marks into dark-on-bright contrast, a coaxial light wins on flat polished faces, and a white ring at 30-45° reads printed labels.

An etched mark reads because it roughens or darkens a mirror-flat surface. Light it straight on and the surface specularly reflects the LEDs into the lens and blanks the code; rake the light and the flat field reflects away while the disturbed pixels scatter light back. That one idea — contrast from angle — settles most cases of a reader that refuses machined parts. Polarization attacks the same glare at the source:

On a moving line, exposure has to freeze the part: keep blur under one pixel, so t ≤ pixel size / speed. At 0.0195 mm per pixel and a 100 mm/s transfer, t = 0.0195 / 100 = 0.0002 s, about 0.2 ms — strobe territory, not a shutter setting. Fire the light from a 4-channel, 200 W strobe controller for 24 V lights such as the LS-CTL-0070:

For why one narrow colour band beats broadband white on bare metal, see monochromatic lighting in machine vision.

How much read rate does a traceability station need?

Rate follows the takt, not the catalogue: one part per second needs one confirmed read per part, so a reader rated 60 codes/s carries a 60-fold margin, and the binding constraint on a moving line is exposure time, not decode count.

Stop-and-shoot stations — the common automotive layout — trigger per part, strobe, decode and report well inside the takt; budget under 100 ms from trigger to result and you have room for a retry. Continuous motion instead obeys the exposure limit from the lighting section, and a rolling shutter that is fine on stopped parts smears moving ones, so choose global shutter there. When one station must cover several marks — build code, VIN and date stamp — a wide field eats pixels per module quickly; the MV-ID5120RM-08M-RBN holds sampling headroom with 12 MP over GigE on stopped parts. For continuous-belt reading the numbers change again: the high-speed conveyor guide walks through that case.

When should a smart camera replace a dedicated code reader?

Choose a smart camera when the code is one check among several — presence, orientation, label skew, a date field — and keep a dedicated reader when decoding hard marks is the whole job.

A smart camera runs the decode and the inspection logic on the device: one mount, one Ethernet drop, one station in the cell. A dedicated reader brings a decoder stack built for marginal marks, plus lighting control and electrical isolation options. If the station must read the code and also verify placement or orientation in the same image, that combined job is what a vision sensor is for:

One image, one result string to the PLC, one station to maintain.

How does the read result reach your PLC and MES?

Plan the hand-off before you order the reader: trigger in on a discrete input, part ID out over TCP or serial to the PLC, and a message format agreed with the controls engineer — integration, not optics, is where traceability stations stall.

Fixed readers come with Fast Ethernet, GigE or serial interfaces depending on the model; shortlist by interface in the code reader finder before optics. Decide who opens the connection: the TCP client versus listener guide settles the roles, and the PLC communication plan covers retries and heartbeats.

What should you check before you order the reader?

Buy on evidence from real parts: five samples — new, worn, oily, re-marked and worst-case — each read 30 times in a row at line rate is the minimum acceptance test before the reader ships.

Around that test:

• The marking drawing: symbology, symbol size, module pitch, quiet zone, marking method.
• A lighting trial on the actual surface at the actual stand-off, with the polarizer swept through its angles.
• Trigger and message format agreed with the controls side: discrete in, TCP out.
• Rate margin: the decode path at five times the takt, so a retry still fits the cycle.
• Housing ratings for the cell — dust, oil mist, washdown — settled before the bracket is designed; the hazardous-area install guide covers the harshest version of that problem.
• For label codes with grade requirements, the label verification guide extends this list.

Run the trials inside the 90-day return window, so a reader that fails on real parts goes back instead of into the panel.

Why does a reader that passes the numbers still fail on the line?

Field failures are optical before they are digital: specular glare on a curved face, a chamfer eating the quiet zone, part height outside the depth of field — each has a ten-minute test and a known fix.

Work them in order. Glare: read the part again, rotated 90°; if the decode follows the rotation, the failure is specular, so rake the light harder or turn the polarizer. Curved faces: one low-angle bar lights one arc and blanks the next, so light from two sides or shorten the stand-off. Quiet-zone infringement: a chamfer, engraving or second mark inside the quiet zone is unfixable in optics — the correction belongs in the marking process. Intermittent decodes are the worst kind; judge them by read confidence rather than by a lucky read before anyone signs the station off.

When the samples read and the numbers hold, pick the model in the fixed code readers category, then size the rest of the station in the configurator below. The code-reading topic page gathers the decoder and integration guides that continue this one.