A 100% inspection camera is only as useful as the master and tolerances loaded into it. If every harmless variation creates an alarm, operators learn to widen limits or ignore the screen. If the limits are too broad, significant defects pass. Build the setup from the approved artwork, customer requirement and real defect risk, then challenge it before the roll is released.
Start from the approved revision
Load the final approved PDF or a verified golden image and confirm the job revision, repeat, lane count and orientation. Do not teach the system from an uncontrolled production sample that may already contain a defect.
Lock master creation or approval to authorised roles. Record who approved the master and when.
Divide the label by risk
Create inspection zones for critical text, barcodes, variable data, logos, colour areas, varnish, foil and non-critical background. Different defects and tolerances may apply in each zone. A pharmaceutical code and a decorative texture should not share one generic sensitivity.
Exclude only genuine process features such as planned gaps or web edges. Every exclusion needs a documented reason.
Define defect classes in plain terms
List missing print, spots, streaks, register shift, colour deviation, matrix waste, splice, missing label and variable-data errors relevant to the job. Define the minimum size or tolerance and the required machine response for each class.
Avoid labels such as 'major' without a measurable acceptance rule. The customer specification should drive the class, not the camera's default menu.
Control normal process variation
Account for substrate texture, holographic effects, reflective foil and transparent constructions through correct lighting, masking and zone settings. Do not simply lower sensitivity across the entire label to quiet one difficult area.
Inspect camera focus, web stability, lighting cleanliness and encoder performance. Mechanical movement can appear as an image defect.
Challenge the setup
Run known defect samples or a controlled challenge roll covering the important classes. Confirm detection, alarm, stop, roll-map or editing response and operator action. Test at normal speed and across lanes.
A pass on one obvious defect is insufficient. Retain challenge results with the job master and repeat the challenge after software, camera or lighting changes.
Review alarms as production data
Track alarm type, lane, position, operator disposition and false-alarm reason. Repeated nuisance alarms may show a setup or web-control issue; repeated real defects point upstream to print or converting.
Changes to limits should be approved and versioned. The final roll report should show what was inspected, which master was used and how defects were handled.
Run a controlled production trial
Use known missing print, streak, spot, register, code and matrix defects across the lanes. Agree the test roll, normal operating range and responsible operator before starting. Begin from verified mechanical condition and approved consumables. Change one variable at a time, allow the process to stabilise and retain samples from make-ready, steady production and the end of the roll. A short demonstration at an easy setting does not prove that the process will hold through a commercial job.
During the trial, monitor master revision, lighting, focus, zones, tolerance, encoder, speed, alarm and roll-map response. Mark every deliberate change and production event on the same timeline. Acceptance should require detection of agreed defects without a nuisance-alarm rate that drives operators to bypass the system. If the result is outside the agreed limit, preserve the sample and machine state long enough to identify the cause instead of making several unrecorded corrections.
Turn the result into a pressroom standard
Save master version, defect class, zone, tolerance, challenge result, alarms, dispositions and approvals. Attach photographs, measurement reports and an approved finished sample where useful. Give the recipe a revision and link it to the exact material, artwork or machine configuration. Operators should be able to see what may be adjusted, the permitted range, the time allowed for stabilisation and who approves an exception.
Define a stop and escalation rule for a missed critical defect, untraceable master change, unstable image acquisition or an operator bypass. Review repeat faults by station, material and shift instead of treating each one as a separate surprise. A good standard does not remove judgement; it gives experienced operators reliable evidence and prevents the next shift from undoing a successful correction.
Questions from production and purchasing teams
What to ask about print inspection setup
Should every area of a label use the same sensitivity?
No. Use zones and tolerances based on the risk and appearance of each feature.
What is a golden image?
A verified defect-free reference image linked to the approved artwork revision and production setup.
How should false alarms be reduced?
Correct lighting, focus, web stability, zones and feature-specific tolerances instead of broadly lowering sensitivity.
What should be changed first during the trial?
Start from verified mechanical condition and approved consumables, then change only one controlled variable at a time.
What makes the result repeatable on the next shift?
A revision-controlled recipe with component identity, settings, measurements, approved samples and clear stop rules.