When to use a line scan camera: motion, sync and light

A line scan camera builds its picture one row of pixels at a time while the part travels past the lens. That single difference drives every choice around it: how fast the camera must clock, why an encoder belongs in the loop, and how bright the light has to be. This guide gives you the selection rule first, then the line-rate, exposure and data-rate formulas with worked belt numbers, the encoder check that keeps pixels square, and the interface math that follows from them. By the end you can commit to a camera, lens, light and interface for your belt, with a number behind each choice.

When does a moving part make line scan the right choice?

Choose a line scan camera when the part or web moves continuously past a fixed camera and the finished image must run without gaps: paper, foil, steel strip, glass, printed web, coated electrodes, or a conveyor of parts in free flow.

An area scan camera captures frames, and every frame has a boundary. A part longer than the frame, or moving faster than the frame rate allows, lands across two frames with a seam in the middle. The line scan cameras family removes that boundary in the travel direction: the sensor is one row, motion lays down the rows, and a 500 m roll of foil and a 5 mm label obey the same logic. Cross-web resolution scales cheaply too, from 2K to 16K on a single row.

Line scan also unwraps round parts. Rotate a roller, bottle or printed cup under the camera and you get a true 360° view with no perspective squeeze at the edges, which a fixed area camera cannot deliver in one image.

Three quick cases:

• Continuous web or endless material — line scan, every time.
• Part stops at the station — area scan; nothing moves, so there is nothing to build rows from.
• Discrete parts moving with gaps — line scan with a part-present trigger, so each part starts its own image block.

The third case has its own walkthrough in the sensor-triggered, encoder-driven capture guide, and coated webs get a full worked example in the battery electrode web inspection guide.

How do I size the line rate to belt speed and resolution?

Divide web speed by the pixel size you need on the part: a belt at 300 mm/s inspected at 0.1 mm/pixel needs a 3,000 lines/s line rate, because 300 ÷ 0.1 = 3,000.

Work Case A on paper — a 300 mm/s belt inspected at 0.1 mm/pixel:

• Line rate: 300 ÷ 0.1 = 3,000 Hz.
• Field of view on a 2,048-pixel line: 2,048 × 0.1 mm = 204.8 mm.
• Data rate in mono 8-bit: 3,000 × 2,048 × 1 B = 6.1 MB/s.
• Exposure cap for blur inside half a pixel: 0.5 × 0.1 ÷ 300 = 0.000167 s, or 167 μs, against a line period of 333 μs.

Every number follows from speed and pixel size alone. Halve the pixel size and the line rate doubles; double the belt speed and it doubles again. Leave margin on the camera side: a camera whose maximum clock sits within 25% of your requirement will crowd the exposure and the interface at the same time.

The LS-LCM-0016 is a 2K GigE line scan camera with a 7 μm pixel, a 58 kHz line rate and a TDI mode on an M42 mount.

Case A needs 3 kHz, so this camera runs at about 5% of its clock ceiling — the headroom you want, not the constraint.

The camera-side settings that follow from the rate are walked through step by step in how to set the line rate of a line scan camera.

Why does an encoder decide whether the image is square?

An encoder fires one camera line per measured step of travel, so belt-speed wobble changes the line rate with the part instead of stretching it: without that link, a 5% speed swing at a fixed line rate makes every row 5% too wide or too narrow.

Pixels are square only when one row lands per pixel-width of travel. Fix the line rate to a clock and the wobble goes straight into the geometry of the image. Tie the clock to a roller-mounted encoder and every row is recorded exactly where the part is at that moment, whatever the belt does between rows.

The check is ticks per pixel. Mount a 2,048 pulses-per-revolution encoder on a drive roller of 200 mm circumference and you get 2,048 ÷ 200 = 10.24 ticks/mm. Case A needs 10 pixels/mm (1 ÷ 0.1), so the rig delivers 10.24 ÷ 10 = 1.02 ticks per pixel. Triggering one line per tick gives a travel pitch of 0.098 mm against a 0.100 mm optical pitch, about 2.4% anisotropy — invisible for defect detection. For gauging, set the camera's line divider so rows accumulate per exact pixel-width of travel.

Below one tick per pixel you cannot divide your way out: the fix is a higher pulse count or an electronic multiplier. The encoder ticks per pixel guide covers the divider math, and the quadrature wiring guide covers the A/B signals into the camera or frame grabber.

What data rate does the line rate create, and which interface carries it?

Multiply line rate by pixels per line by bytes per pixel: the slow example makes 6.1 MB/s (3,000 × 2,048 × 1), which any GigE port carries easily, while the fast example makes 328 MB/s (40,000 × 8,192 × 1), which rules GigE out.

Bytes per pixel is 1 for mono 8-bit and 3 for 8-bit RGB. Case B is a 2,000 mm/s foil line at 0.05 mm/pixel on an 8,192-pixel camera: it clocks at 2,000 ÷ 0.05 = 40,000 Hz and produces 40,000 × 8,192 × 1 B = 328 MB/s in mono. GigE delivers roughly 100 MB/s of usable payload after protocol overhead, so this one multiplication decides the interface.

Colour triples every figure: Case B in RGB moves 983 MB/s (3 × 327.7), past a single 10GigE port's comfort zone and into CoaXPress, Camera Link or XoFLink territory. The GigE Vision selection guide covers the network side, and the Camera Link frame grabber compatibility guide covers the grabber side.

How much light does the short exposure need?

Blur, not the sensor, caps exposure: hold motion within half a pixel and the 3 kHz line allows 167 μs, while the 40 kHz line allows 12.5 μs — over thirteen times less light per row, which is why fast webs live on strobed line lights.

Case B's cap works out at 0.5 × 0.05 ÷ 2,000 = 12.5 μs, exactly half its 25 μs line period. Per-row light at 12.5 μs is 13.4 times scarcer than at 167 μs, so the light must pay that factor back in irradiance or the rows go dark. Three levers exist, in the order to try them: a brighter strobed light, TDI, then binning.

Run the light in strobe mode from the camera's exposure signal, through a lighting controller built for short pulses. At 12.5 μs the LED's optical rise time is no longer free — a few microseconds of ramp eats a visible slice of the window — so measure the usable light the way the LED strobe rise time guide shows.

A line light must reach across the whole field, or the ends of every row go dark. Case A's field is 204.8 mm, and the line scan lights range runs from under 50 mm to over a metre of lit length. The LS-LSC-0527 is a 300 x 17 mm line light source rated at 87 W in red or 108 W in white.

That 300 mm of lit length covers the 204.8 mm field with margin for mounting tolerance.

Case B's field is 409.6 mm. The LS-LSC-0017-L25 is a 24 V line scan light, 1092 mm long.

Its 1092 mm length covers the 409.6 mm field with room to spare for lens clearance over the web. Geometry is a separate choice from power: bright field for print and surface colour, oblique dark field for embossed scratches. The illumination techniques guide lays out both arrangements, and the line-light power guide does the irradiance arithmetic for fast lines.

When the light still cannot close the gap, TDI shifts charge along the direction of travel so several rows add their signal — the LS-LCM-0016 carries a TDI mode for this duty — and pixel binning trades resolution for light at the same clock.

Which camera resolution fits the web width?

Work from the smallest defect, not the catalogue: catching a 0.3 mm defect reliably takes about 3 pixels across it, which sets 0.1 mm/pixel and makes a 2,048-pixel line cover 204.8 mm — exactly Case A. The three-pixel rule comes from detection practice: a one-pixel spot is one noisy pixel among thousands, while a 3 × 3 patch stands clear of the noise floor. Tighten the spec to 0.15 mm defects and the same field needs 4,096 pixels.

Case B works from the other end: 409.6 mm ÷ 0.05 mm/pixel = 8,192 pixels, an 8K camera by definition. At the 40 kHz clock it feeds the 328 MB/s worked out in the interface section, so resolution and interface lock together.

The LS-LCM-0099 delivers an 8K mono line over 10GigE on an M72 mount.

The field of view guide runs this calculation for other web widths and pixel pitches.

Sensor width, not pixel count alone, decides the lens. A 2,048 × 7 μm row is 14.3 mm wide and any 1-inch C-mount lens covers it; an 8K 7 μm-class row is 57.3 mm wide and needs a large-format lens in M58, M72 or V-mount. The LS-LSL-0023 is a 16K 5 µm line scan lens with a 116 mm focal length, a 127.2 mm working distance and a 60-90 mm image circle on a V-mount.

Keep 10-20% of the image circle in reserve at the row ends, or vignetting darkens them; the vignetting guide separates that falloff from natural cosine falloff, and the line scan lens selection guide matches lenses to 4K, 8K and 16K rows. You can also browse the line scan lenses range by image circle and mount.

When should you not use a line scan camera?

Skip line scan when the part is stationary or the machine stops parts in front of the camera: with no travel there is nothing to build rows from, and an area scan camera with a fixed frame is simpler at every level.

• Stationary or indexed part: area scan, with exposure sized from the same blur formula; the camera fps and shutter speed guide covers the frame-rate side.
• Speed varies with no encoder and no motion control: rows distort, and no software fix reconstructs them reliably, so the encoder is not optional for measurement work.
• Slow, low-resolution jobs below roughly 100 mm/s: the blur cap sits at 500 μs (0.5 × 0.1 ÷ 100), short enough for any area camera, and many area cameras offer a line scan mode for slow webs as a middle path.
• No rigid bridge for camera, lens and light at a fixed working distance: line scan optics have no autofocus to fall back on, so the mechanical frame is part of the instrument.

One more honest limit: line scan raises the floor of the whole station. Encoder, lighting controller, strobe wiring and a stiff mount are not upgrades you bolt on later — they are the system. If the part is small, still and needs 5 μm measurement, a telecentric area-scan setup beats a line scan rig.

What should you check before ordering?

Check five numbers on paper before you order — line rate, ticks per pixel, data rate, exposure cap and light length; each is a minute of arithmetic here and each costs days to unwind during commissioning.

1. Line rate at maximum belt speed, plus 25% margin, against the camera's maximum clock (Case A: 3 kHz; Case B: 40 kHz).
2. Ticks per pixel at least 1, and 2 or more for gauging: encoder PPR ÷ roller circumference, divided into pixels per mm.
3. Data rate in mono, then × 3 for colour, against the interface's usable payload with 20% headroom.
4. Exposure cap at half a pixel of travel, with a strobe pulse width and rise time that fit inside it.
5. Light length at field width plus ~15%, wavelength chosen for the surface, controller rated for the strobe current at that duty cycle.
6. Lens image circle over sensor width with reserve, mount matching the camera, working distance fitting the mechanical bridge.

When the numbers fight — colour at 40 kHz is the usual fight — the trade is real and worth a human: send the belt speed and defect spec through the contact page and we will work it with you. To go from numbers to parts, start the camera step of the configurator, browse the line scan camera finder, or use the lights finder and lens finder once the widths and clocks above are settled. More reading along the same path lives on the cameras topic page.