LED strobe rise time: why microseconds of light delay blur your image

Every strobed machine-vision light takes a few microseconds to reach full intensity, and a few more to decay. At exposure times of 10 to 100 µs those edges are part of the light budget: an exposure that opens on the rising edge records a dimmer image, and one that straddles the falling edge drifts darker still, frame by frame, as trigger timing wanders. This page gives the numbers: typical rise times for strobed constant-current outputs, the pulse-width and trigger-lead rule that holds the shutter on the flat top, and a bench method that measures what your own light and controller do. It is written for anyone pairing a fast camera with a strobed light through a dedicated controller.

How fast does a strobed LED light actually turn on?

The LED die itself emits within tens of nanoseconds of current arriving; what you measure is the driver, so a well-driven strobed constant-current output reaches full intensity in roughly 1 to 10 µs, while a supply that must wake a soft-start loop can take hundreds of microseconds. The junction is never the bottleneck. The optical rise time you can see, measured between the 10 and 90 percent points of full intensity, is the time the driver needs to push the set current into the LED string and its wiring. Three contributions stack up:

• Output stage. A switched constant-current stage slews to the set point in about 1 to 10 µs. A supply that must first wake its regulation loop can take 100 µs to a few milliseconds, which is why a strobe output and a dimmable power supply are different products.
• Trigger path. Input filtering and isolation in the controller add a trigger-to-output delay; on digital strobe outputs that is normally single-digit microseconds. Budget 10 µs until you measure yours.
• Wiring. The inductance of the controller-to-light loop sets the current slew; the arithmetic is further down.

Delay and rise time are different faults with different fixes. Delay slides the whole pulse later; you cancel it with the trigger offset and never think about it again. Rise shapes the pulse; only a wider pulse or a faster output stage fixes it. When a strobed image comes back black or half-lit, the first suspect is the offset, and strobe light on but camera image is black: fix the timing walks that diagnosis.

When should the strobe fire relative to the camera exposure?

Fire the strobe early enough that the exposure opens on the flat top: set the pulse width to the exposure plus rise plus fall plus about 10 µs of margin, and lead the exposure start by the rise time plus half that margin. Written as two lines you can put in the commissioning sheet:

t_lead is how far before the exposure start the strobe command must fire, and the 10 µs margin absorbs trigger jitter on both the camera and the controller side. If the controller cannot be made to lead the camera, invert the arrangement: trigger the light first and delay the exposure start by the same t_lead. The camera's own trigger-to-exposure delay, a few to tens of microseconds on global-shutter CMOS sensors, is a constant; measure it once with the method below and fold it into the offset.

The exposure itself comes from the motion budget, not from the light: exposure time vs shutter speed for moving-part inspection shows that arithmetic.

Worked example: a barcode passing at 2 m/s

Assume an allowed blur of 0.2 mm on the print, and a measured 8 µs rise and 8 µs fall on the light:

• Exposure ceiling: 0.2 mm ÷ 2,000 mm/s = 100 µs. Run at 100 µs.
• Pulse width: 100 + 8 + 8 + 10 = 126 µs, set 130 µs.
• Lead: 8 + 10/2 = 13 µs, set 15 µs before exposure start.
• Duty at 500 fps: 130 µs ÷ 2,000 µs = 6.5 percent; at 1,000 fps, 13 percent.
• Pulse trimmed to the bare exposure, with both edges inside the window: 1 − (8 + 8) ÷ (2 × 100) = 92 percent of the light, about 0.12 stops dark, and the error moves with every microsecond of trigger jitter.

High speed barcode verification: size the strobe pulse sizes the same pulse against decode grade rather than brightness, and the duty side of the setting is covered in LED strobe duty cycle: matching lights and controllers. Pulse width and trigger offset both live in the controller, so keep them in software rather than on a trim pot: a 2-channel 24 V digital LED light controller with an RS-232 interface.

How much light do I lose when rise time eats into the exposure?

With the pulse the same width as the exposure, budget the usable fraction as 1 minus rise plus fall over twice the exposure: 8 µs edges on a 100 µs exposure cost about 8 percent of the light, a tenth of a stop. The formula assumes linear ramps and both edges landing inside the exposure window, which is the conservative budget:

Hold total edge time under a fifth of the exposure and the loss stays under 10 percent. For scale, one gray level out of 256 is 0.4 percent, so a 10 percent loss shifts the histogram visibly on an eight-bit camera.

The wall is harder than the slope. When rise + exposure + fall no longer fits the window you actually have, the line period on a line scan camera or the maximum pulse width the light allows, no trigger offset rescues the image: the exposure integrates a ramp, the frame is dim, and brightness walks with every microsecond of jitter. That case, not the gentle percentage loss, is where rise time decides the architecture.

What sets the rise time — the LED or the driver?

The driver and its wiring set it: the die switches in nanoseconds, but the current loop must slew from zero to the set current against the loop inductance, so rise time tracks the voltage headroom divided by that inductance. The loop inductance of the power wiring is the number to know. Two metres of controller-to-light cable runs near 2 µH, a wire pair lands around 1 µH per metre as an estimate, and the current slew into that inductance is fixed by how much voltage headroom the driver holds above what the LED string needs:

With 5 V of headroom: di/dt = 5 V ÷ 2 µH = 2.5 A/µs, so a 2 A strobe current is reached in 2 ÷ 2.5 = 0.8 µs. Squeeze the headroom to 1 V and the same loop takes 4 µs. Short, twisted leads and a driver that keeps headroom in reserve are the two levers; raising current without headroom lengthens the edge rather than shortening it. High-current flash drivers commonly parallel several smaller current sources so the edge stays fast while the flat-top current adds up.

Output type decides whose rise time you are buying. A constant-current strobe output regulates the LED current directly, so the numbers above are the controller's. A constant-voltage 24 V output feeding a light with its own internal driver hands the edge to the light; treat the light's strobe specification as the controlling one. Three light controllers span the choice, from fast digital strobing to simple analog switching:

Values as printed in each product's data sheet; open a part for its drawing and full specifications.

The LS-CTL-0487 is the 2-channel 24 V digital unit above, rated 240 W, listed with both constant-current and constant-voltage 24 V output modes. The LS-CTL-0054 is the high-current end: 2 channels, constant current, 240 W per channel and 320 W total, analog control. The LS-CTL-0011 is the simple end, an analog 4-channel 24 W unit for lights that stay lit between frames.

How do I measure rise time and trigger delay on the bench?

Trigger the controller from a pulse generator at the real repetition rate, aim a photodiode at the lit face from about 300 mm, and read both numbers on one oscilloscope: delay from trigger edge to 10 percent light, rise from 10 to 90 percent. The procedure in full:

1. Set the controller to the current, pulse width and repetition rate you will actually run; rise time stretches when the loop runs out of headroom at high current.
2. Trigger it from a pulse generator, not from software, so the electrical edge is clean and known.
3. Aim a photodiode, a plain photodiode into 50 Ω is enough, at the lit face from about 300 mm, and put it and the trigger line on one timebase.
4. Read delay from the trigger edge to the 10 percent point and rise from 10 to 90 percent. Scope bandwidth should be at least 0.35 divided by the rise you expect: 350 kHz resolves a 1 µs edge, 3.5 MHz a 100 ns edge.

The first pulse after a long idle can differ from the steady ones, and on a one-shot inspection station that first pulse is the image: first camera frame dark: exposure settling or light timing? separates the camera-side causes. The full method with the usable-light integration is in LED strobe optical rise time: measure usable light. If the flat top sags at the start of a hard pulse, that is current droop rather than rise time: LED strobe current droop: diagnose pulse startup.

When does rise time not matter?

Above about 1 ms of exposure it stops mattering: a 10 µs rise inside a 1,000 µs window costs about 1 percent of the light, so spend the money on output, uniformity and thermal design rather than edge speed. Stationary parts, 30 fps inspection and exposures of 1 to 10 ms all sit in that regime; a 10 µs rise vanishes into the noise. Choose the light on output, wavelength and uniformity, and which LED colour gives the best contrast in machine vision is the decision that matters there.

Line scan is the far pole. A 4,096-pixel camera across an 800 mm web gives 0.195 mm per pixel; at 1 m/s the line rate is 1,000 mm/s ÷ 0.195 mm = 5,128 lines/s, one line every 195 µs, and a 100 µs exposure with 16 µs of edges and margin fits inside it. At 4 m/s the period falls to 49 µs: a 100 µs exposure cannot fit at all, and a 20 µs exposure would need 46 µs of pulse inside 49 µs of period, 94 percent duty, which the light will not survive. The right answer there is a light sized for continuous running, as in line scan light: how much power a fast line really needs, on a controller rated for the current, the role lighting controllers for line scan cameras: what they fix describes.

What should I check before ordering a light and controller pair?

Check five numbers in this order: the exposure your motion budget allows, the light's maximum strobe current and pulse width, the controller's trigger-to-output delay, the duty limit, and the sum of rise, exposure and fall against the shortest window you will ever run. As a list for the requisition:

• Exposure ceiling from motion. Allowed blur ÷ part speed. Below about 50 µs, edge speed becomes a first-order specification.
• The light's strobe ratings. Maximum pulse current and maximum pulse width bound the flat top you can buy, and both come from the light's data.
• Output type. Constant-current strobe output: the edge is the controller's. Constant-voltage 24 V output into a light with its own driver: the edge is the light's.
• Trigger-to-output delay. Budget 10 µs, then measure it once and treat it as a fixed offset in the trigger setting.
• Duty at rate. Pulse width × repetition rate: a 130 µs pulse at 1,000 fps is 13 percent. If that exceeds the limit, shorten the exposure or move to continuous light.
• The sum check. Rise + exposure + fall + margin must fit the shortest window you will ever run, line period or frame period. If it does not fit, no setting fixes it.

On the light side, small fields strobe well from a spot light and wide stations and webs from a bar light; both families carry models built for hard pulse duty:

The LS-BAR-0400 is a 312 x 30 mm high luminance bar light; the LS-SPT-0175 is a 6 mm high luminance spot light, power 5 W.

Work the numbers in the light controller finder or start at the controller step of the configurator. If your motion budget lands in single-digit microseconds, send the part speed and the camera model through contact and we will size the pulse and the lead with you.