A brighter LED does not automatically produce a stable inspection image on a fast conveyor. The product must reach the field of view, the camera must expose at the right moment, the light must actually emit during that exposure, and the LED and driver must remain within their thermal limits as the next products arrive. A setup that looks excellent in a static demonstration may yield intermittent dark frames or bands at production speed. This guide helps production engineering, quality and maintenance teams in Thailand specify the lighting head and controller as one synchronized imaging system.
This article addresses the strobe control decision after the inspection task and camera class are broadly defined. For ring, dome, coaxial and other general lighting geometries, see our machine vision lighting selection guide. For camera resolution, shutter and interface selection, see our industrial camera selection guide. Here the question is when the chosen light emits, for how long, at what drive level, and with what repeatability.
Buy an imaging chain, not an isolated light head
Maximum brightness, LED count and nominal wattage are poor stand-alone comparison points. Assess the illuminator, pulsed-current controller, camera, product sensor, PLC, wiring, shielding and mounting together. Does the head permit the proposed peak current? Can the controller deliver that current for the specified pulse width at the worst production frequency? Does the actual light output overlap the useful camera exposure window? A quotation that merely says “strobe capable” leaves these questions unanswered.
Before requesting proposals, put five input sets on one sheet: minimum and maximum part pitch and line speed; smallest relevant defect and acceptable motion blur; camera shutter type and exposure mode; measured or specified delays from detection to exposure and illumination; and operating temperature, runtime and permissible duty cycle. Mark unknown values “to be measured” rather than filling them with catalogue defaults. Advanced Illumination describes the dependence of overdrive on the light head, current, pulse width and thermal design. CCS publishes distinct strobe and overdrive modes with limits tied to a particular controller. Neither vendor’s product figure should be used as a universal limit.
Draw the timing chain before choosing a pulse width
A photoelectric sensor may detect a product upstream of the camera. The product then travels a known nominal distance, while PLC processing, camera trigger acceptance and lighting-controller response add different delays. Let d be the distance from the detection point to the target position and v the line speed: d/v is the nominal travel time. It is not a sufficient fixed delay if belt speed varies, products slip, or sensing position changes. An encoder-based trigger, a relocated sensor or extra field-of-view margin may be needed. Count missed and duplicate acquisitions as well as successful images.
Camera exposure time and LED pulse command duration are different quantities. The controller has trigger-to-output latency; the LED has rise behavior; the camera has a defined exposure-active or flash-window signal. Draw separate traces for Sensor, Camera Trigger, Exposure, Controller Input and actual Light Output. Use a scope or other suitable timing instrument to verify edges on the machine rather than assuming that identical setting numbers mean overlap. Basler’s light-control documentation describes flash-window-based control and compensation for turn-on delay, illustrating why a timing diagram is essential.
Poor overlap gives a dark image even at high drive current. Variable overlap gives frame-to-frame brightness changes that destabilize a threshold or an AI model. Record input polarity and voltage, programmable delay range and increment, minimum and maximum pulse width, and whether the camera or controller is the timing master. Tie each waveform capture to an image ID. The objective is a reproducible window in which the relevant surface is illuminated during every accepted exposure.
Start with acceptable motion blur
Strobing often aims to freeze a moving feature. If the acceptable movement at the workpiece is b and speed is v, b/v is a first estimate of the longest useful illumination interval. For example, a 0.05 mm blur allowance at 500 mm/s gives 0.1 ms. This is an engineering starting point, not a guarantee: magnification, vibration, angled travel, shutter behavior and required defect contrast still need real sample tests.
Record an image with the strobe switched off. If ambient lighting still exposes the moving workpiece, the remaining camera exposure can create a ghost or smear even when the pulse is short. Enclosing the cell or shortening exposure may matter more than increasing peak brightness. Compare day, night and maintenance conditions when doors or other lights can change ambient reflections. Use production-good samples and borderline defect samples, not only an attractive demonstration part.
Shorter illumination requires sufficient photons, but “set maximum overdrive” is not a design method. Check working distance, beam placement, lens aperture, sensor sensitivity, wavelength and gain as one optical budget. Increasing digital gain makes the display brighter while also raising noise; it may not preserve the contrast needed for a small defect. Compare local defect contrast, saturation, uniformity and repeatability with fixed camera settings.
Global and rolling shutters require different synchronization checks
With a global shutter, all pixels generally share an exposure interval, although the exact output definitions depend on camera model and mode. With a rolling shutter, rows begin exposing at different times. A short pulse synchronized to only one row may produce a bright band and dark remainder. Check whether the camera offers a common all-row flash window, how long that window is, and whether the required light dose fits inside it. Do this from the selected camera’s documentation and real images.
Basler warns that particular automatic strobe and exposure modes must not be combined in its light-control implementation. That is a model-specific warning, not a rule to apply blindly to every camera. It shows why the RFP should name model, firmware, shutter type, trigger mode, exposure mode, line output and timing master. At maximum speed and minimum product pitch, log frame mean, defect-region contrast, banding, missing frame IDs and the camera’s re-arm behavior. A fast light cannot compensate for a camera that cannot accept the next trigger.
Calculate duty cycle under the worst sequence
Duty cycle is the proportion of time the light is on. For one pulse per period, an initial calculation is pulse width in seconds multiplied by flashes per second, then multiplied by 100%. A 300 µs pulse at 20 Hz gives 0.6%; at 200 Hz it gives 6%. If there are several flashes for each product, add their on-times. Use the shortest product pitch and sustained peak flow, not only an average daily throughput. A recipe change that adds images also changes the load.
The calculation does not itself tell you allowable peak current. The limit depends on LED head, wavelength, heat sinking, ambient temperature, controller and pulse width. Request the manufacturer’s compatibility table and derating rules for the exact combination. CCS’s CN controller brochure lists limits for its particular modes; those are examples of model-specific constraints. Advanced Illumination’s practical guide explains how light-head design and thermal management affect safe overdrive. Use neither as a substitute for the selected equipment’s manual.
Do not measure only the outside of the light housing. LED junction temperature and controller protection can limit operation before the housing seems unusually hot. Follow manufacturer limits, then record cabinet temperature, light-housing temperature, alarms, brightness drift and missed pulses in the real installation. An air-conditioned FAT bench and a warm Thai production enclosure are different environments. If brightness falls over time, first check dirt, alignment and protection logs before increasing current.
Some controllers skip output pulses when trigger frequency exceeds a safe duty condition. Advanced Illumination’s DCS manual explicitly describes this behavior and the dark images it can cause. A sporadic dark frame is therefore not necessarily a camera failure or LED ageing. Correlate the frame ID, trigger frequency, pulse settings, controller alarm and temperature on a common time line. Avoid a setting that automatically raises current after a bad image and hides the underlying protection event.
What belongs in the controller comparison table
For every candidate, request head compatibility, channel count, current and voltage limits, pulse-width and delay ranges, trigger voltage and polarity, camera synchronization output, protection behavior, recipe storage, readable settings, event logs and replacement procedure. A matching connector is not evidence of electrical and thermal compatibility. If multiple lights fire in sequence, specify whether channels can be set independently and how each flash maps to a camera frame and product ID.
| Decision item | Supplier evidence | FAT evidence |
|---|---|---|
| Timing route | I/O drawing, polarities, voltage and timing diagram | Scope trace tied to image IDs |
| Drive envelope | Head-specific current, width, rate and duty limits | Continuous run at peak speed |
| Image stability | Reference part and fixed camera settings | Start-up and warm-state distributions |
| Protection | Overcurrent, overtemperature and excess-trigger behavior | Alarm and recovery demonstration |
| Service | Backup file, spares and calibration procedure | Reference-image recovery after replacement |
Multiple illumination directions may require one image per channel. A controller’s built-in sequence can simplify wiring, but it also raises frame count, processing load and the need to keep every frame linked to its workpiece. Advanced Illumination’s DCS-400E product page provides an example of sequenced channels and camera triggering. Decide how many images the inspection actually needs before paying for more channels or assuming the PLC can meet the timing.
Write an RFP that a supplier can price and test
Replace “provide bright lighting” with a measurable request: under the specified maximum speed, minimum pitch, part variation, camera model, space and ambient temperature, propose the light-head and controller models; state pulse width, current, frequency, calculated duty cycle, manufacturer-supported envelope, protection response, wiring and settings; and demonstrate them on the supplied good, borderline and defective samples. State which measurements are still open and who will close them during the PoC.
Include material, surface finish, oil film, height variation and ambient-light conditions. If basic illumination geometry is still undecided, make that a separate PoC item. Keep geometry as constant as practical while comparing controller timing; otherwise a different light angle can mask a synchronization problem. Define image acceptance with defect-region contrast, saturation, field uniformity, frame-to-frame variation and dark-frame count, not a single average brightness figure. Lock ROI, exposure, gain and file format before testing. A demo screenshot with automatic enhancement is not an acceptance record.
FAT: link waveforms to images
At factory acceptance, run the product at a representative production speed. Record sensor input, camera trigger, exposure-active output, controller input and light output if measurable. Save each waveform together with workpiece ID and image ID. Test normal and maximum speed, shortest pitch, restart after stopping, consecutive arrivals, missing parts, channel sequencing and power cycling. Keep every failed frame, not only the best examples. Define acceptance for acquisition completeness, defect visibility, protection alarms and recipe recovery.
Run long enough at the agreed frequency and thermal conditions to observe warm-state behavior. Log temperature and image brightness over time. Use the chosen product manual to set absolute current and temperature limits; do not copy a generic temperature from an article. Do not disable thermal protection just to pass FAT. If a single missed flash occurs, preserve its waveform and alarm state, identify the cause and repeat the test after correction.
SAT: repeat the test in the installed environment
Site acceptance in Thailand adds daylight, night-shift lighting, motor and power noise, final cable routing, enclosure heat, cleaning and oil contamination. Re-measure synchronization after production wiring replaces test leads. Compare SAT images with FAT reference images at the same settings. When they differ, inspect alignment, shielding, dirt and trigger signals before increasing current. Document connector checks and failure diagnosis for maintenance staff.
The flash may be uncomfortable or hazardous to view directly. Identify operator sight lines, reflective surfaces and maintenance access with covers open. Decide shielding, indications and interlocks through the equipment risk assessment. IEC 62471 provides a framework for evaluating photobiological hazards of non-laser lamps including LEDs, but citing a standard number alone does not establish safety for the installed machine. Request the manufacturer’s optical-safety information for the actual head and operating condition, and evaluate exposure in the real workplace.
Hand over controller settings and version, wiring, I/O map, waveforms, reference images, sample parts, inspection thresholds, cleaning interval, spare parts, alarm interpretation and a replacement verification procedure. A spare controller is useful only if technicians can restore the settings and prove that the reference image and defect response are reproduced.
Diagnose common failure patterns systematically
Suppose dark frames occur only after several products arrive at the shortest pitch. First match frame IDs to sensor and trigger times. If the sensor fired but no camera trigger arrived, investigate PLC logic and wiring. If the trigger arrived but exposure did not start, inspect camera re-arm time and trigger mode. If exposure started but the light did not flash, inspect controller duty protection, trigger limits and recipe state. If the light flashed, examine overlap, optics, contamination and product height. Raising gain before locating the missing stage can conceal the fault.
If day shift passes and night shift fails, take strobe-off and strobe-on images under both ambient conditions. Exposure outside the short pulse can still admit room light. Reflections from open guards or nearby equipment may change, even with an otherwise stable pulse. Fix shielding or exposure first, then repeat tests with borderline defect samples. Store image, waveform, temperature, alarm and workpiece position under one test number so intermittent faults can be reconstructed later.
For multiple flashes per part, calculate the full sequence time and verify that the next product cannot enter before its predecessor’s images are linked. Record encoder count, trigger number, camera frame ID and PLC workpiece ID. Verify total controller output limits as well as per-channel ratings, especially when channels overlap. Keep recipe files under version control and record the actual loaded recipe in image logs. A display showing the expected name is not proof that the expected pulse train ran.
Cost, acceptance data and maintenance
Compare installed cost: head, controller, mount, shield, wiring, added I/O, waveform tests, sample testing, commissioning, training, spares and service. An inexpensive head that cannot run at line speed can make later camera or conveyor changes far more costly. Use current dark-frame rate, escaped defects, reinspection time and stoppage costs as plant-specific business inputs. There is no defensible universal payback period for strobe lighting.
Create a test ledger that ties test number, part lot, sample class, speed, pitch, ambient and cabinet temperature, camera and lens settings, pulse current, width, rate and delay to image and waveform files and approver. Use the same columns at FAT and SAT. Define ROI and statistical measure before seeing results. For defect tests, fix sample count, random order and workpiece ID matching. If a threshold is still unknown, mark it as a PoC decision with an owner and date. During maintenance, image the same reference workpiece periodically; check window cleanliness, alignment, supply and trigger before changing current. Record every approved setting change and retest borderline defects.
Purchase checklist
- Record the smallest defect, borderline samples, part variation, speed and shortest pitch.
- Estimate motion-blur allowance and confirm shutter mode and exposure window.
- Draw sensor, PLC, camera, controller and LED timing with polarities and delay.
- Verify model-specific pulse width, rate, current, duty and ambient limits from manufacturer documents.
- Agree on sustained worst-case tests for banding, missing frames and skipped flashes.
- Include site lighting, heat, cable route, cleaning and optical safety in SAT.
- Require setting backup and a reference-image test after replacement.
Frequently asked questions
Is strobing always brighter than continuous lighting?
No. A compatible head and controller may deliver more light within a brief interval under a low duty cycle, but permitted current and thermal behavior are product-specific. Poor timing can leave the image dark despite a bright flash. Test the required image quality at the actual production rate within the manufacturer’s operating envelope.
Is a pulse width equal to exposure time sufficient?
No. Start time, camera delay, controller latency, LED rise and shutter behavior all matter. A common trigger sent to both devices and a light triggered from Exposure Active have different delay paths. Save waveforms and images at both shortest and longest part pitch.
Are occasional dark frames evidence of ageing LEDs?
Not by themselves. Missed triggers, camera re-arm, controller protection, loose wiring, recipe switching and ambient changes are alternatives. Link the failing frame ID to sensor input, controller alarm, pulse frequency and temperature. Dark frames clustered at short pitch warrant an immediate duty and trigger-limit check.
What should a Thai factory prepare before seeking advice?
Prepare a line video, product drawing, good and borderline defect samples, speed and shortest pitch, current camera and lighting models, cabinet temperature, installation space, and any image or trigger logs. Missing measurements can be planned as part of a PoC. Production engineering, quality, maintenance and the machine builder should work from the same conditions sheet.
Discuss your application with TOMAS TECH
TOMAS TECH supports Thai manufacturing lines across cameras, lighting, PLCs, conveying and inspection-result integration. If you need to compare strobe controllers, stabilize a fast inspection line or specify RFP and FAT/SAT evidence, contact us with workpiece samples, current camera and line conditions. We can help define a realistic image trial after checking each candidate’s model-specific limits and safety information.
