Strobe vs continuous lighting and how to choose a lighting controller

Why the lighting mode is a system-level decision

In machine vision, lighting is not a passive accessory bolted on at the end; it is what determines whether your features are present in the raw image at all. The choice between continuous and strobed operation sits upstream of almost everything else: it interacts with the camera's shutter type and exposure time, with how fast the part is moving, with the lens aperture you can use, and with the controller hardware you must buy. Get it wrong and it shows up as motion blur, washed-out or starved images, flicker beating against the frame rate, or LEDs that dim and die months after commissioning.

The physics is simple to state. A sensor collects light only during its exposure (integration) time. To first order, the signal at a pixel is proportional to scene radiance multiplied by exposure time, lens transmission, and the aperture area, which itself scales with one over the f-number squared. On a moving part you must keep the exposure short to avoid blur, and that short window starves the sensor of light. You then have three levers to restore the signal: open the aperture (which costs depth of field), raise sensor gain (which costs noise), or deliver more light during the window. Strobing is the lever that adds light without those penalties, which is why it is the workhorse of high-speed inspection.

So the practical question for an integrator is rarely strobe or continuous in the abstract. It is: given my line speed, my required resolution and field of view, my chosen camera and lens, and my ambient environment, what mode gives a clean, repeatable image, and what controller do I need to drive it safely? The rest of this guide answers that in order.

Continuous, strobed, and overdrive operation

Continuous (DC) operation runs the LEDs at a steady current at or below their nominal continuous rating. The light is always on, brightness is constant, and there is nothing to synchronize. This is the right default for stationary parts, slow indexed stations where the part is at rest during exposure, manual inspection aids, and any case where you have ample light and no motion problem. Its weakness is headroom: the brightness you see is all you get, so on a fast-moving part a short exposure simply yields a dark image.

Strobed (pulsed) operation switches the LEDs on only for a brief window aligned to the exposure, then off again. Two benefits stack. First, a short, well-timed pulse freezes motion: if the light is on only while the part moves a fraction of a pixel, the image is sharp regardless of how dark the room is, because almost no light reaches the sensor outside that window. That same property suppresses ambient and overhead lighting, which is often as valuable as the motion freeze. Second, because the duty cycle is low, you can drive the LEDs in overdrive, pushing the pulse current well above the continuous rating; LEDs tolerate brief, high-current pulses far better than sustained overcurrent, so you can extract substantially more instantaneous brightness, often several times the continuous output, during the exposure.

A subtle point that trips people up: strobing does not by itself increase brightness. Pulsing an LED at its continuous current just makes it blink at the same peak level. The brightness gain comes from overdrive, which the low duty cycle of strobing makes thermally safe; the motion-freeze and ambient-rejection benefits come from the short on-time. These are independent advantages that strobing unlocks together, and conflating them is the source of most confusion about why strobed systems look brighter.

Overdrive is therefore best understood as a property of the controller-and-light pair, not of the light alone. The safe overdrive depends on pulse width, repetition frequency, and the LED and fixture thermal design, which is exactly why manufacturers publish overdrive curves. A passive light pulsed by an unaware power supply, with no knowledge of those limits, can be destroyed in seconds.

• Continuous (DC): simplest, constant brightness, no timing; best for stationary or slow indexed parts and already-bright scenes.
• Strobed: short on-time freezes motion and rejects ambient light; requires synchronization to the exposure.
• Overdrive: pulse current above the continuous rating for brief, low-duty windows; multiplies brightness but only within the published limits.
• Strobing alone does not brighten. Overdrive brightens, the short on-time freezes motion, and the low duty cycle makes overdrive safe.

Synchronizing the strobe with camera exposure

A strobe is only useful if the light is fully on for the entire time the sensor integrates, and off otherwise. The cleanest way to achieve this is to let the camera drive the timing: the camera's exposure-active (strobe-output) signal triggers the controller, which fires a pulse wide enough to cover the integration window. The alternative is a single line or PLC trigger that fires both the camera and the light; this works, but you must account for the camera's trigger-to-exposure delay so the pulse and the integration window actually overlap.

The dominant gotcha is shutter type. A global-shutter sensor starts and ends integration on every pixel at the same instant, so a single short pulse during that common window illuminates the whole frame uniformly. That is what makes global shutter the natural partner for strobing and high-speed work. A rolling-shutter sensor exposes rows sequentially over a longer interval, so a short pulse fired mid-frame brightens only the rows that happened to be integrating then, producing a bright band across the image. You can strobe a rolling shutter only if the pulse covers the full readout window, which sacrifices the motion-freeze benefit, or by using a global-reset mode where the sensor offers one and the ambient light during readout is negligible. For most strobed motion applications, choose a global-shutter camera.

Timing quality comes down to two numbers: latency and jitter. Latency is the fixed delay from the trigger edge to light-on; a constant latency can be compensated with a programmable delay so the pulse centers on the exposure. Jitter is the shot-to-shot variation in that delay, and it cannot be compensated; it eats directly into your timing margin and, at high line speeds, causes either partial illumination or brightness variation from frame to frame. When parts move fast, also budget for the controller's minimum pulse width and for the encoder or trigger resolution that positions the part.

Set the pulse slightly wider than the exposure, never narrower, with a small lead and lag so that even with jitter the light is solidly on for the whole integration. Then tune brightness by adjusting exposure or current rather than by trimming the pulse, because a pulse narrower than the exposure leaves part of the integration dark, while a pulse much wider than needed wastes duty-cycle budget and raises thermal load for no extra signal.

• Prefer camera-driven triggering (exposure-active or strobe output) so the pulse tracks the real exposure window.
• Use global-shutter sensors for strobed motion; rolling shutters band unless the pulse covers the whole readout.
• Compensate fixed latency with a programmable delay; jitter cannot be compensated and must fit inside your timing margin.
• Make the pulse slightly wider than the exposure, with lead and lag, so jitter never leaves the frame partly lit.

Duty cycle, overdrive limits, and LED lifetime

Duty cycle is the pulse width multiplied by the repetition frequency: the fraction of time the LED is on. It is the single most important number for protecting the light. LED brightness and lifetime are governed by junction temperature, and junction temperature tracks average power. A high pulse current is tolerable as long as the average, set by the duty cycle, keeps the junction cool enough. This is why a light that survives a given overdrive at a low duty cycle will fail at the same current once you raise the frequency or widen the pulse: the peak did not change, but the average, and therefore the heat, did.

Two distinct limits apply, and both must be respected. The peak-current (or peak-power) limit caps the instantaneous current in any single pulse; exceed it and you risk immediate, catastrophic LED failure even at a low duty cycle. The duty-cycle (thermal) limit caps the time-averaged current; exceed it and you cook the junction, which shows up as gradual lumen depreciation, color shift, and shortened life rather than a sudden failure. Good controllers enforce both, often interlocking pulse width and frequency so you physically cannot program a setting that exceeds the safe overdrive curve for the connected light.

There is also a real trade between brightness and longevity even inside the safe envelope. Running near maximum overdrive accelerates the dimming that forces you to recalibrate exposure over time; running with margin keeps output stable far longer. For a 24/7 line it is usually wiser to specify a slightly larger or brighter fixture run at moderate overdrive than a small light flogged at its limit, so that aging does not silently erode image quality and trip false rejects months later. Always cross-check the fixture's published overdrive curve against your worst-case pulse width and the highest frame rate the machine can reach, including transient bursts, not just the nominal cycle time.

• Duty cycle (pulse width times frequency) sets average power and therefore junction temperature and LED life.
• Respect two separate limits: peak current (instant failure) and duty-cycle/thermal (gradual degradation).
• Prefer a controller that interlocks pulse width and frequency to enforce the light's overdrive curve automatically.
• Run with margin below the maximum for stable output over time; a small light flogged at its limit drifts and causes false rejects.

Sizing and selecting a lighting controller

A lighting controller is a constant-current driver with timing and intensity control, not a generic power supply. Specify it from the lights outward. Start by counting channels: each independently driven light or independently controlled zone needs its own output, because that is what lets you set intensity, pulse, and delay per light. A four-light station that needs separate brightness per light needs four channels, not one channel fanned out.

Then size current and voltage per channel against the worst case, not the average. Each channel must supply the light's peak overdrive current, which can be several times the continuous rating, and it must have enough output voltage headroom to drive the LED string's total forward voltage plus the driver's own overhead at that current. A common mistake is to specify the channel for the continuous current and then find it current-limits or droops at the pulse the application actually needs, quietly costing you brightness. Confirm both the peak current per channel and the maximum string voltage the channel supports.

Triggering is the next axis. You want enough isolated, optocoupled trigger inputs to accept signals from cameras, a PLC, or encoders without ground loops, plus programmable per-channel delay and pulse width so each light can be timed to its own exposure. Check the specifications that actually bound performance: minimum pulse width, trigger latency and jitter, and maximum trigger (retrigger) frequency. These must comfortably exceed your line's demands, or fast motion will blur and brightness will wander frame to frame.

Finally, match the control and intensity features to the application. Per-channel analog or digital intensity setting is essential for repeatable, recallable recipes across product changeovers. RGB or multi-channel control matters when you switch wavelengths to maximize contrast on different features or surfaces. Field-bus or serial control (for example, over an industrial network or a simple serial link) lets the line recall lighting recipes automatically rather than relying on a manual potentiometer that anyone can knock out of calibration. Specify these deliberately so the controller, the lights, and the camera form one timed, repeatable system.

• One channel per independently controlled light or zone; do not fan one channel across lights you must set separately.
• Size each channel for peak overdrive current and for the LED string's forward voltage plus driver overhead, not the average.
• Require enough isolated trigger inputs and confirm minimum pulse width, latency, jitter, and maximum retrigger rate against your line.
• Insist on per-channel intensity (and RGB or multi-wavelength control if needed) plus recipe storage for repeatable changeovers.

A decision path: which mode for your application

Begin with motion. If the part is stationary or fully at rest during the exposure, continuous lighting is almost always the simplest correct answer: no timing to get wrong, constant output, and a cheaper driver. If the part is moving while you image it, ask how far it travels during the exposure you would need. Once that motion approaches a fraction of a pixel, you need either an extremely short exposure with a great deal of light or, more practically, a strobe that fires a brief, intense pulse and freezes the motion.

Next, weigh ambient light and contrast stability. Even on slow lines, strobing with overdrive can be worth it when overhead or sunlight pollutes the image, because the short window plus a high peak swamps the ambient and makes the inspection immune to changing room light. If you are fighting flicker, inconsistent exposure, or daylight leaking into the cell, a synchronized strobe often solves it more cleanly than chasing the ambient with enclosures and shrouds.

Then confirm the rest of the chain can support the mode you want. Strobing presumes a global-shutter camera, a trigger source (camera, PLC, or encoder), and a controller whose minimum pulse width and jitter fit your line. If any of those is missing, that is a hardware decision to make now, not a setting to tune later. When the application is genuinely borderline, prototype both: continuous is trivial to try, and a single strobed exposure on the real part at line speed will tell you immediately whether you have frozen the motion and held the contrast.