A collimated backlight turns a soft silhouette into a black-and-white edge you can threshold to a single pixel — but only when the geometry is right. The same narrow cone that sharpens edges will wrap light around the flank of a round part and blur it if the panel sits badly. Here you get the working-distance rule behind that wrap angle, the light-sizing decision, and a complete worked setup for a 12 mm round part gauged to ±0.05 mm with a standard C-mount lens. If you have not settled collimated versus diffuse for edge gauging, start with the collimated vs diffuse edge-gauging comparison; everything below assumes collimation is the choice.
Why do edges wrap around round parts under a collimated backlight?
The wrap is geometric, not a defect: a collimated panel still emits across a ±30° cone, with output falling from maximum to 5% somewhere in the 15-30° band, so rays leaving the panel's outer edge strike the flank of a round part at a tilt and smear the silhouette.
Output holds near full strength out to roughly ±15°, so the rays you must control are the strong ones. A ray leaving the panel edge sits (D_L/2 − a/2) horizontally outside the field-of-view edge; that offset over the sample-to-light distance WD_sl is the wrap angle x°. A diffuse panel throws light from every direction, so wrapping is built in — the same off-center geometry behind edge shadows on off-center parts — and the formula below removes it.
How far should the collimated backlight sit from the part?
Size the sample-to-light distance so the wrap angle stays under 10°: WD_sl ≥ 5.7 × (D_L/2 − a/2), where D_L is the panel's emitting width and a your field of view — a 35 mm panel over a 16 mm field therefore sits 54 mm or more from the part.
D_L is the emitting width, a the field of view, WD_sl the gap from part to panel. Worked row: the 35 mm-over-16 mm pair gives (17.5 − 8) mm = 9.5 mm, and 9.5 mm ÷ 0.176 = 54 mm. Mount at 60 mm for mechanical margin and the real angle is atan(9.5 ÷ 60) = 9.0° — still inside the limit.
| Panel width D_L | Field of view a | D_L/2 − a/2 | Minimum WD_sl at 10° |
|---|---|---|---|
| 35 mm | 16 mm | 9.5 mm | 54 mm |
| 90 mm | 60 mm | 15 mm | 85 mm |
| 120 mm | 100 mm | 10 mm | 57 mm |
| 120 mm | 80 mm | 20 mm | 113 mm |
Hold 10° when you gauge edges; presence checks and coarse profiles tolerate more, with visibly rounder corners. The formula assumes the worst case — a part flank at the field edge — so a part that wanders off-center stays covered as long as it remains inside the declared field.
With a fixed, the camera side follows from sensor width and lens working distance:
How large should the collimated light be?
Choose the smallest panel whose emitting width still covers the field of view plus positioning tolerance: every 10 mm of extra panel width adds about 28 mm to the distance you need at the 10° limit, and oversizing is the most common cause of wrapping.
The factor is 1 ÷ tan 10° = 5.67, so each extra 5 mm of half-gap costs 28 mm of distance. Size down before you space out. Fields under about 30 mm fit a collimated spot; fields near 100 mm suit a flat collimated panel; matched telecentric pairs shrink the light to the lens front diameter, covered below.
The LS-CLM-0051 is a 35 mm collimated spot light on 5 V. Fields up to roughly 30 mm sit inside that beam with room to center the part.
The LS-BKL-0830 is a 120 x 120 mm collimated back light on 24 V. Fields around 90-110 mm match it, needing 28-85 mm of distance at the 10° limit.
- Size
- Under 50 mm
- Voltage
- 5 V
- Size
- 100-200 mm
- Type
- Collimated
How do you set the system up step by step?
Six steps take you from part print to lit image: fix the field of view, pick the smallest covering panel, set the 10° distance, size the lens from sensor width and magnification, start at f/5.6, then verify the edge ramp at the worst-case part position.
- Fix the field of view. The part is 12 mm across and the guide holds it to ±2 mm: a = 12 + (2 × 2) = 16 mm.
- Pick the panel. That spot covers the field with 9.5 mm of margin per side — the tightest sensible fit.
- Set the gap. Half-gap is (35 − 16) ÷ 2 = 9.5 mm; 9.5 ÷ 0.176 = 54 mm; mount at 60 mm, giving x° = 9.0°.
- Size the lens. Example camera: 5 MP mono, 2448 × 2048 pixels at 3.45 µm, sensor width 8.45 mm. Magnification m = 8.45 ÷ 16 = 0.53. With a 35 mm lens: WD = f × (1 + m) ÷ m = 35 × 1.53 ÷ 0.53 = 101 mm. Pixel scale is 16 mm ÷ 2448 px = 6.5 µm/px, so a 0.1 mm tolerance band spans about 15 px.
- Set the aperture. Start at f/5.6. Depth of field ≈ 2 × N × c × (1 + m) ÷ m² with c = 0.015 mm (2 px): 2 × 5.6 × 0.015 × 1.53 ÷ 0.28 = 0.9 mm total, about ±0.45 mm — enough for a capped part on a hard stop; stop down if part height wanders.
- Set exposure, then strobe if needed. Continuous-on at a 60 mm gap typically exposes in single-digit milliseconds through f/5.6; if the part moves or the gap grows toward the brightness limit below, pulse 100-500 µs at the panel's rated strobe current.
- Verify at the worst case. Park a part at the field edge, threshold it, and check the gray ramp across each edge spans 2-3 px or less. Re-gauge a reference part 10 times and confirm the spread stays within 1/10 of tolerance — the bar set by the 10:1 accuracy and repeatability rule.
Does a telecentric lens remove the wrap angle?
Yes — a telecentric lens accepts only rays parallel to its optical axis, so the wrap angle x° drops to 0° at the sensor, the light can shrink to the lens's front diameter, and panel-to-part distance stops mattering for edge quality.
That is why matched telecentric illuminators exist, and it is the configuration for the best edge precision. What the lens cannot change is brightness: the aperture-filling limit in the last section still applies. Pay for telecentric at the sub-0.01 mm classes, on thick parts whose silhouettes must not shift with height, or whenever the telecentric gauging decision guide says magnification must hold; the matched-light method for round shiny parts is in the telecentric illumination guide.
The LS-CLM-0079 is a telecentric collimated light with a 90 mm beam diameter, 223.9 mm long, on 3 V. That beam covers fields up to about 60 mm at the 10° distance (85 mm), and closer once a telecentric lens above it only sees parallel rays.
- Size
- 50-100 mm
- Voltage
- 3 V
- Type
- Telecentric
Does the same setup work for transparent parts?
Yes, and it works better than on opaque parts: parallel rays enter clear plastic or glass and refract out of the lens path, so the body reads near-black against a bright field while edges, cracks and extrusion lines keep full contrast.
A diffuse panel floods the same features from every angle and washes them out. The full technique — dark silhouettes, wall-thickness ghosts, and where thin film defeats the effect — is in the collimated backlighting guide for transparent parts; for glass sitting inside a clear housing, the glass-in-plastic-enclosure lighting answer covers the extra reflections. Keep the 10° geometry anyway: on transparent parts, wrapping shows up as bright rims rather than gray gradients.
Which colour should the collimated backlight be?
Pick green: typical CCD and CMOS sensors respond most strongly around 480-560 nm, so a green panel reaches the same pixel levels at the shortest exposure — exactly what a measurement station wants.
Red is the fallback when the part transmits green strongly or you need to suppress fluorescence; blue buys a small diffraction benefit on the finest edges at a sensitivity cost. The contrast logic matches any backlight colour decision — the LED colour and contrast guide walks it — and once a fixture is qualified, hold the colour: LED binning between orders can shift brightness and wavelength enough to move a threshold.
How do you keep the image bright when the panel sits far away?
Respect the aperture-filling limit: through f/5.6 a lens accepts rays over about ±5°, and the spot panel over the 16 mm field fills that cone only out to roughly 105 mm — past it the image dims steeply even though the panel still looks bright to your eye.
The mechanism: from a point at the field edge, the panel subtends atan((D_L/2 − a/2) ÷ WD_sl) — 9.0° in our 60 mm setup — while the lens cone is set by NA = 1 ÷ (2N): 5.1° at f/5.6, 3.6° at f/8. Rays outside that cone are emitted but never collected. Closing to f/8 lifts the distance limit to about 150 mm, trading a slightly softer, diffraction-broadened edge for the reach.
When the machine fixes the gap farther out, buy the signal back with strobe current: pulsing at rated strobe current for 100-500 µs drives the LEDs harder than continuous-on and returns several stops of headroom, while rejecting plant lighting at the same time — worth documenting if your customer specifies ambient light rejection for acceptance tests. Trigger timing decides whether the pulse lands inside the exposure; the strobe output timing guide lines up camera exposure, light pulse and timestamp.
The LS-CTL-0070 is a digital control box with four strobe channels, 200 W total output on 24 V, set over RS232.
- Output
- Constant voltage 24 V
- Channels
- 4 channels
- Control
- Digital
- Feature
- Strobe
Next step: size the panel and gap with the formula above, then filter the collimated lights by emitting width and colour, compare flat panels under backlights, and match voltage and strobe channels in light controllers. The light finder and the light configurator pair the panel with your camera, and the applications team will check the geometry against your part print before you order.
Questions
How far can the panel sit before the image starts to dim?
A standard lens only collects rays inside its acceptance cone, so the panel must subtend more than the lens half-angle from every field point. At f/5.6 that half-angle is about 5°, and a 35 mm panel over a 16 mm field satisfies it out to roughly 105 mm. Stopping down to f/8 (3.6°) stretches the limit to about 150 mm; past it, strobe at higher current or fit a wider panel.
What working distance does ±0.01 mm accuracy need on a round part?
Tighten the wrap-angle limit from 10° to about 5°. The distance doubles, because tan 5° is 0.0875: the same panel-field pair then sits at 9.5 mm ÷ 0.0875 ≈ 109 mm. At that accuracy class also fit a telecentric lens, since the residual ray cone of a standard lens starts to dominate the error budget.
Can a collimated backlight replace a telecentric lens-and-light pair?
Often, yes. The panel supplies the parallel illumination that sharpens the silhouette, so a standard C-mount lens can gauge round profiles to a few hundredths of a millimetre. The telecentric pair still wins at the ±0.01 mm class, when parts tilt or vary in height, or when magnification must hold constant across the depth of the part.
Do I need a special driver to run a collimated backlight?
Not for constant-on use: flat collimated backlights accept plain 24 V, while the small collimated spot family runs at 3-5 V, so match the channel output to the light. A strobe-capable controller earns its place when the part moves or the panel sits at the brightness limit, because pulsing at rated current for 100-500 µs restores the lost signal.