One camera for pick and place and barcode scanning: the shared setup

One overhead camera, one six-axis robot, two jobs: find each loose part on the table, then read its code before the part ships. The instinct is to serve both from one snapshot, and that instinct is right — as long as you size the pixels for the barcode first, because the code asks for roughly ten times the resolution the pick does. This how-to gives you the pixels-per-module math, the three layouts that work, a worked example on a 400 mm table, and the exposure and bench checks that finish the build.

What the two jobs demand

A decoder rebuilds bars and dots from pixels, and it needs at least 2 pixels per module; below that, reads go intermittent. Design to 3 pixels per module and treat 4 as comfortable. For a 0.25 mm module that is a pixel scale of 0.083 mm or finer, and on a 2448-pixel sensor the field of view then caps at 204 mm.

Locating a part for a pick is an easier ask. Sub-pixel edge tools place an edge to about a tenth of a pixel, so the same 400 mm field on 2448 pixels — 0.163 mm/px — still locates features to a few hundredths of a millimetre. The robot's own repeatability, ±0.05 to ±0.1 mm on a typical six-axis unit, dominates that stack-up. The design rule: the code sets the pixel scale, the pick only sets the field of view.

Step 1: Choose the layout before the camera

Three layouts make one camera serve both jobs. The pixels-per-module number at the table plane picks between them.

The refocus line is not optional. With the lens focused for a table 580 mm below the lens, a part presented at 300 mm sits far outside depth of field and blurs about 18 pixels — enough to kill both the code and the edges. Fixed focus serves one plane; the refocusing lens, or a second reader, buys the second plane.

Running the pick and the scan as two separate locates with different accuracies, rather than one generic pick, has control-side consequences too; this camera pick-and-place two-locates guide walks that part of the build.

Step 2: Worked example: 400 mm table, 0.25 mm DataMatrix code

Inputs as they come off the bench:

• Table working area 400 × 300 mm, parts loose on the surface.
• Code: DataMatrix, 0.25 mm modules, target 3 px per module, floor 2.
• Bench camera: 5 MP, 2448 × 2048 pixels at 3.45 µm, sensor 8.45 mm across.
• Upgrade candidate: 20 MP, about 5472 pixels at 2.4 µm, sensor 13.1 mm across.

Rule in words: pixels per module = module size × pixels across / FOV width; magnification m = sensor width / FOV; focal length f = WD × m / (1 + m).

Table plane on the 5 MP camera: 0.25 × 2448 / 400 = 1.53 px per module. Below the floor of 2 — the wide shot cannot decode this code. It still locates: 0.163 mm/px leaves edges good to a few hundredths of a millimetre, and the 400 mm field puts a 12 mm lens at WD = 12 × 1.021 / 0.0211 = 580 mm.

Table plane on the 20 MP camera: the field needs 400 / 0.0833 = 4800 pixels and the sensor offers 5472, so 0.25 × 5472 / 400 = 3.42 px per module — above target. Optics: m = 13.1 / 404 = 0.0324, and f = 510 × 0.0324 / 1.0324 = 16.0 mm — a 16 mm lens at 510 mm working distance frames the 404 mm field. Depth of field at f/4 with a 2-px circle of confusion of 0.0048 mm: 2 × 4 × 0.0048 / 0.0324² = 36 mm, about ±18 mm around the table surface.

Scan pose keeping the 5 MP camera: the 0.0833 mm/px target caps the field at 2448 × 0.0833 = 204 mm, so m ≥ 8.45 / 204 = 0.0414 and WD ≤ 12 × 1.0414 / 0.0414 = 302 mm. Hold the label plane 300 mm under the 12 mm lens: 3.02 px per module. That is a 280 mm lift, about 0.7 s at 500 mm/s plus a 100 ms settle, and a liquid or motorized lens refocus on every part.

Decision: the 20 MP camera decodes the code in the same exposure that locates the part — no lift, no refocus, cycle cost zero. Keep the 5 MP scan-pose branch only when the robot can present every part at the pose and you accept a refocus per shot; if parts cannot be lifted, the code earns its own reader at the station. The same pixels-to-millimetres chain, worked end to end on a measuring task, is in coin size sorting with a single camera.

Step 3: Pick the camera to match the branch

For the single-shot branch, 20 MP carries the table alone: the MV-CS200-10UM is a 20 MP mono camera on a 1" sensor with USB3 and a C-mount. Its rolling shutter costs nothing here because the parts sit still during the exposure.

On tables between the 204 mm five-megapixel cap and the full 400 mm width, 12 MP is the balance point. The MV-CH120-11UC runs 12 MP with a global shutter on USB3 in a 1.1" format, so a moving part in the wide shot stays unskewed.

Without a PC in the cell, a smart camera runs locate and decode on the device: the MV-SC5060XM-08M-WBN carries 6 MP mono, an 8 mm M12 lens and GigE. About 3000 pixels across a 6 MP sensor decode 0.25 mm modules on tables up to about 250 mm in one shot; on wider tables the smart camera keeps the pick and the code moves to a reader.

When parts cannot be lifted and the table is wide, mount a dedicated reader where the robot presents each part: the MV-ID3013PM-12M-WBN is a compact 1.3 MP reader with a fixed 12 mm lens on Fast Ethernet, and the robot triggers it as its own cycle step. The fixed code readers page lists short-range and logistics variants.

The PC-based branches draw from the area scan cameras page; the camera finder filters by pixels, interface and shutter, and the smart cameras page holds the no-PC line.

Step 4: Time and light the two exposures

Give every exposure a motion budget: keep blur under a third of a pixel, so exposure time ≤ pixel scale / (3 × speed). Parts at rest on the table tolerate 10-20 ms under steady light. At the scan pose, dwell 100 ms after the in-position signal, then fire 10 ms: a 1 mm/s drift then blurs 0.01 mm, about 0.12 px at the 0.083 mm/px scan scale.

Set f/4 as the working aperture from the example: depth of field ≈ 2 × f-number × CoC / m², and at m = 0.032 that lands within about ±18 mm of the set plane — your pose height tolerance. Stop to f/5.6 only if poses wander; past that, diffraction starts to soften 2.4 µm pixels. The depth of field calculator runs your own numbers.

One diffuse white light over the table serves both exposures; add polarization if labels glare. When the jobs want different settings, hold two exposure sets on the camera and alternate per trigger — two exposure times, one camera shows the mechanism. For light that drifts through the shift, ambient light variation in pick and place vision gives the containment steps.

Step 5: Verify on the bench before wiring the cell

1. Etch or print a golden 0.25 mm DataMatrix and place it at the four table corners and the centre.
2. Require 100 consecutive decodes per spot; intermittent decodes mean pixels per module sit under target, not that the decoder needs tuning.
3. Log the confidence value with every read and judge it the way barcode read confidence sets out.
4. Constrain the decode ROI to the presented part; setting an ROI for inline code reading shows the speed and reliability gains.
5. Calibrate the pixel frame to the robot, then structure the program so each located part maps to a drop target — structuring a vision-guided pick and place program around drop targets is the pattern.

Mistakes that cost weeks

• Decoding from the 400 mm wide shot on a 5 MP camera because one good label decoded once. At 1.5 px per module the read rate collapses under print variation; the math above is the check.
• One fixed focus across two working distances. Focused for 580 mm, a 300 mm pose blurs about 18 px — codes fail and edges smear together.
• Firing the camera before the robot settles. Vibration for 50-100 ms after a move is enough to blur a code.
• Full-frame decoding in the cycle. An ROI sized to the part cuts decode time by an order of magnitude and drops false reads from the neighbouring part.

One camera, two jobs, three layouts — the numbers decide. Take the table size, module size and cycle target into the system configurator, which walks camera, lens and light as one build; more builds of this kind sit on the systems topic page. If a branch stays open, send the numbers through the contact page and we will work the example for your parts.