Lighting controllers for line scan cameras: what they fix

A line scan station already has three devices that must agree thousands of times a second: a camera that acquires each line, a light that must be identical on every one of them, and an encoder that says where the web actually is. A lighting controller is the fourth box, and it is the one that makes the other three behave. This explainer covers the four jobs a controller does — holding light level steady as LEDs age, turning coarse encoder pulses into a trigger for every line, shaping strobe pulses to fit microsecond exposure windows, and sequencing several lights through one camera — with the arithmetic to size each. By the end you will know whether your station needs one, and which outputs and channels to order.

What does a lighting controller fix in a line scan system?

Four things: light level that drifts as LEDs age and warm, encoder pulses far too coarse to trigger every line, strobe pulses that merge together above a few kHz, and one camera that must see the web under two different lights.

Traditional line scan wiring gives the camera an encoder signal and the light a power supply, then hopes both stay put. That works until the light dims, the encoder cannot keep up, or the job grows a second view. Each failure maps to one controller capability:

Every row maps to a filter in the light controllers category, and the sections below walk them in the order a line scan project meets them. One neighbouring question sits outside this article: how much strobe current a light can take is covered in LED strobe duty cycle: matching lights and controllers.

How do you keep the light output stable as LEDs age?

Hold the current, not the voltage: a constant-current controller pins the LED current while output falls with age and heat, and a 5 % loss of light is then a 5 % trim of current instead of a re-tuned threshold.

Line scan shows LED drift before any other imaging mode because the hours pile up: a line that runs 24/7 logs 8,760 hours a year against thresholds you set on commissioning day. Two effects are at work. Thermal droop appears within minutes of switch-on as the junction warms — LED flux falls of order 0.1–0.3 % per °C for white and red LEDs, so a light that settles 30 °C hotter loses several percent by lunchtime. Aging then moves the same direction over thousands of hours. Current sag inside a strobe pulse is a separate, faster effect with its own diagnosis in LED strobe current droop: diagnose pulse startup.

Constant current is the minimum fix. It holds current while the diode's forward voltage wanders (about −2 mV/°C), which is exactly what a plain voltage supply lets run away. Where the threshold budget is tight, closed-loop control watches a photodiode and moves current to hold measured output — the method and its wiring are covered in closed loop lighting control for repeatable vision images.

The same per-channel control fixes line uniformity. A long line light reads darker at its ends: resistive drop along the LED chain and fewer diodes at useful angles to the edges. Split the light into segments on separate channels and trim the ends up against a uniform white target until the flat-field profile sits within ±5 %; residual variation past that point is usually lens vignetting, not the light.

A four-channel current source covers a segmented line plus spares:

The LS-CTL-0180 is a 4-channel constant-current controller rated 120 W for 24 V lights. Drift matters most when you hunt rare defects: if the occasional fault is the product, selecting cameras for recording intermittent line faults is the companion decision to stabilising the light.

Why can't the encoder trigger every camera line directly?

Because one encoder pulse spans many pixels of web: a 2048 pulse-per-revolution encoder on a 409.6 mm wheel reports every 0.2 mm, a 50 μm line pitch needs a report every 0.05 mm, and the controller supplies the missing factor of four.

An encoder is the only honest clock in a web system — triggering from wall-clock frequency assumes the web never slips, and it always does. But the encoder is bolted to the transport, not to the pixel grid, so its pulse rate is rarely the line rate. A timing controller conditions the encoder signal (differential RS-422 or HTL into clean logic), squares it, and multiplies or divides its frequency by settable ratios. Quadrature decoding, wiring and line-signal details are covered in encoder-triggered line scan capture: wiring quadrature inputs.

Worked example — a 16K camera across an 819 mm web at 2 m/s. The MV-CL166-91F2C is a 16K line scan camera with 5 μm pixels and an 82 mm image circle. Its 16,384 pixels at 5 μm make an 81.92 mm sensor. From there:

• Magnification: 81.92 mm ÷ 819.2 mm = 0.1×.
• Object pixel: 5 μm ÷ 0.1 = 50 μm on the web.
• Line rate: 2000 mm/s ÷ 0.05 mm = 40,000 lines/s = 40 kHz, a 25 μs line period.
• Encoder: 2048 pulses over 409.6 mm of wheel travel = one pulse per 0.2 mm = 10 kHz at 2 m/s.
• Multiplication: 0.2 mm ÷ 0.05 mm = 4, so the controller emits four camera lines per encoder pulse, 40 kHz total.

Divide instead when the encoder is faster than the line rate — the same ratio logic runs both ways. The camera side of the arithmetic is what how to set the line rate of a line scan camera walks through, and the line scan cameras finder filters by line length and pixel size so the sensor matches the web first time.

How wide should the strobe pulse be at high line rates?

Set the pulse to 60–80 % of the line period — 15 to 20 μs inside the 25 μs line of the 40 kHz example — wide enough to fill the exposure window, narrow enough that edge jitter can never bridge two lines.

The failure mode at each extreme is different. A pulse wider than the line period merges with the next one and the light becomes continuous: exposure then equals the line period whatever the camera asks, motion freezing disappears, and each line's brightness bleeds into its neighbour. A pulse much narrower than the exposure throws away the light you paid for and forces current up to compensate. LED rise and fall times eat the budget too — reserve a few microseconds at each edge and measure them rather than assume, as LED strobe optical rise time: measure usable light shows.

Repeatability is the other half of shaping. A sub-microsecond, line-synchronised pulse places the light in the same part of every exposure. A PLC output refreshing on its scan cycle places it anywhere inside a millisecond, and at 2 m/s a millisecond is 2 mm of web — 40 pixels at 50 μm — arriving under random lighting.

Duty cycle closes the argument. An 18 μs pulse in every 25 μs line is 72 % duty, which is continuous operation in a strobe costume: overdrive lives below about 10 % duty, so above a few tens of kHz you buy light power instead of overdrive current.

How do you run two lights through one line scan camera?

Double the line rate and alternate: the controller fires light A on odd lines and light B on even lines, software splits the interleaved stream into two images each still sampled at 50 μm, and one camera station replaces two.

The worked example carries over. The single-light rate was 40 kHz, so the camera runs 80 kHz with each light strobed every second line: 40 kHz effective per light, an 18 μs pulse in every 50 μs, 36 % duty and half the thermal load per light. Keep the doubled rate — at the original 40 kHz each image's web-direction sampling would halve to 100 μm. The doubling is what preserves resolution, and it needs a camera with line-rate headroom, which current 16K models have.

What you do with the second light is the point. Bright field plus dark field on one sensor catches gloss defects and print defects in a single pass. Colour sequences — red, blue, white — turn one mono camera into a multispectral inspection, the technique detailed in multi wavelength light: sequence colours to get more from one camera and sequential machine vision lighting for multi-view inspection.

Exact line-locked timing is not optional here. One misrouted line pairs a flash of the wrong light with the wrong slice of web, and the reported fault is wrong, not just the image. Enable the sequence from a stable state, because the first frames after a light turns on can be dark: first camera frame dark: exposure settling or light timing? covers the startup order.

Sequencing needs one strobe-capable channel per light:

The LS-CTL-0322 is a 4-channel digital controller with 8 A total output; the LS-CTL-0070 a 4-channel strobe control box with RS232; the LS-CTL-0487 a 2-channel 24 V unit rated 240 W.

Constant current or constant voltage for a line scan light?

Match the controller's output to the light's input: a light specified by drive current needs constant current, a light with its own onboard driver needs constant voltage, and long high-power lines move to 48 V to keep cable current reasonable.

The line scan lights category splits the same way the controller category does — 24 V and 48 V dominate. Never cross the types: a current output forced into a light whose internal regulator sets its own current will fight it, and a voltage output into a raw LED chain gives you the thermal runaway that constant current exists to prevent.

Length and power pick the rest. The LS-LSC-0131 is a 975 × 22 mm line light that runs on 24 V. Its 975 mm of emitting length comfortably overfills the 819 mm field of the worked example, which is exactly what you want — a line light should be longer than the field so the sensor sees the flat part of the profile. A light that size draws real power, and the controller must deliver it without sagging at the strobe edge.

For one long line plus headroom, high power on few channels:

The LS-CTL-0264 is a 2-channel high-power digital controller rated 400 W. How much power your line actually needs at your line rate is its own calculation — line scan light: how much power a fast line really needs — and it should come before the controller order, because watts, not channels, choose the model.

When is a lighting controller the wrong fix?

Below roughly 5 kHz with one continuously-lit light and a camera whose built-in encoder multiplication covers your pulse spacing, a clean 24 V supply and the camera's own strobe output do the whole job.

The decision per case:

• One light, slow line, no drift specification: skip the controller. Filter the supply, drive the light continuously, let the camera time everything.
• One light, coarse encoder, camera without multiplication: controller, for the trigger arithmetic alone.
• Two or more lights on one camera: controller, always — no camera sequences external lights on its own.
• Thresholds that must hold for months: constant-current control at minimum, closed loop if the budget allows.
• No encoder at all: a controller cannot add geometry. Fit the encoder first; free-run triggering measures a web that slips.

One trap worth naming: driving camera triggers from a PLC high-speed output. The arithmetic above already priced it — a millisecond of scan-cycle jitter is 2 mm of misplaced web at 2 m/s. Keep the PLC upstream of the controller, never between encoder and camera.

What should you check before ordering?

Six numbers settle the order: the light's input type and wattage, the controller's output voltage, the channel count now and next year, the encoder's signal level, the pulse-width step, and who multiplies — camera or controller, never both.

• Light input type — constant current or constant voltage, read straight off the light's data sheet; it halves the controller list on its own.
• Total wattage at full strobe — per-channel and total ratings must clear it with margin; watts choose the model.
• Output voltage — 24 V or 48 V to match the light, decided with the light, not after it.
• Channels — one per light, plus segments for uniformity trimming, plus one spare.
• Trigger input level — the controller must accept your encoder's output, differential RS-422 or HTL; an adapter in the timing path is a jitter source.
• Pulse-width resolution — sub-microsecond steps for a 25 μs line; coarse steps show up as brightness banding.
• Multiplication owner — camera or controller, one of them, never both.

Then take the numbers through the light controller finder or the controller step of the system configurator. If the station is anything beyond the simple single-light case, talk it through with us before the order.