Line scan cameras

When line scan beats area scan

The deciding factor is geometry combined with motion. If your product is a continuous web or a long, moving strip with no natural frame boundary, an area scan camera forces you into overlapping captures with stitching artefacts and seams. A line scan camera removes that problem entirely: the conveyor or web motion is the second axis of the image, so you get one unbroken image of arbitrary length at perfectly consistent resolution down the run.

The second strong case is resolution across a wide product. A line scan sensor concentrates all its pixels into a single very long row - commonly 2K, 4K, 8K, 12K or 16K pixels wide - so it can resolve fine defects across a wide web that would need several tiled area scan cameras to match. One 16K line scan camera can cover a width that might otherwise demand four or five area scan cameras plus the calibration and overlap management that comes with them.

Line scan also shines on round or cylindrical parts. By rotating the part exactly once under a fixed line, you image the entire circumference as a flat, undistorted rectangle - useful for inspecting bottles, cans, shafts, fruit and other rotationally symmetric objects where an area scan camera would only ever see one face at a time.

• Choose line scan for: continuous webs (paper, film, foil, textile, nonwovens), wide sheet/coil inspection, print quality, and 360 degree imaging of rotating parts.
• Stay with area scan for: discrete parts that fit one frame, intermittent or unpredictable motion, and applications where lighting and encoder integration overhead is not justified.
• If a single area scan camera cannot resolve your smallest defect across the full width, a wide line scan sensor is usually simpler than tiling multiple cameras.

Line rate, speed and resolution-over-width

The core performance number is line rate - how many lines per second the camera can expose and read out, quoted in kHz. Line rate, web speed and your required along-the-web resolution are locked together: line rate (lines/s) must equal web speed (mm/s) divided by the desired pixel size along the travel direction (mm/line). Push the web faster or ask for finer resolution and the required line rate climbs in direct proportion. It is common to spec the line rate first from the fastest production speed and smallest defect, then work backwards to the sensor and interface that can sustain it.

Across the web, resolution is set by sensor width and optics: pixel size across the web equals the field of view width divided by the number of sensor pixels. This is where the resolution-over-width advantage matters - an 8K or 16K sensor lets you hold a small pixel pitch across a wide product without tiling. Keep the across-web and along-web pixel sizes roughly equal (square pixels in the final image) unless the application deliberately favours one axis, so defects look the same regardless of orientation.

At high line rates, light becomes the limiting factor before the sensor does. The exposure window per line shrinks as line rate rises, so you need disproportionately more illumination to keep signal-to-noise acceptable. When light runs out, TDI (time delay and integration) line scan sensors are the answer: they integrate the same point on the web across multiple sensor stages, multiplying effective sensitivity and enabling very high speeds in low-light conditions, at the cost of higher price and tighter mechanical and encoder tolerances. Catalog line scan families from Basler (racer 2, including TDI variants), iRAYPLE (L Series) and Omron/Sentech (STC/FS line scan) span this range from cost-effective CMOS to high-sensitivity TDI.

• Required line rate (kHz) is fixed by max web speed and smallest along-web defect - calculate it before anything else.
• Across-web resolution comes from sensor pixel count and lens magnification; wide sensors avoid multi-camera tiling.
• Aim for near-square effective pixels so defect detection is direction-independent.
• Reach for TDI when speed is high and you cannot add enough light any other way.

Encoder synchronisation and uniform geometry

Because the image is built line by line, any variation in material speed distorts the picture - speed up and the image stretches, slow down and it compresses. The standard solution is to trigger each line from a rotary encoder coupled to the web or conveyor, so one line is acquired per fixed increment of travel rather than per fixed unit of time. This keeps the along-web scale constant regardless of speed changes, accelerations, stops and starts, giving you measurable, repeatable geometry for metrology and reliable defect sizing.

Specify the encoder resolution against your along-web pixel size: you want enough encoder pulses per millimetre that each acquired line maps cleanly to one pixel row, often using the grabber's tick-multiplier or divider to match exactly. Plan for direction handling and jitter on lines that start and stop, and confirm the camera or frame grabber accepts the encoder input directly. Encoders, trigger cables and brackets from suppliers such as Datalogic/Datasensing are part of the system, not an afterthought - the encoder is as much a precision component as the camera.

• Encoder-triggered acquisition fixes the along-web scale so the image does not stretch or compress with speed.
• Match encoder pulses-per-mm to your target pixel pitch, using the grabber's multiply/divide to land exactly one line per pixel row.
• Free-run (timed) triggering is only acceptable when web speed is genuinely constant - rare on real production lines.

Interface, frame grabber and line lighting

Line scan generates a continuous, high-bandwidth data stream, so the interface and acquisition hardware are critical. GigE Vision (often with PoE) suits lower and mid-range line rates and simplifies cabling, while Camera Link and especially CoaXPress (CXP) carry the sustained bandwidth that fast, wide sensors demand. Camera Link and CXP cameras require a matching frame grabber in the host PC - the grabber handles deterministic encoder triggering, line-rate timing and DMA transfer that a software-only interface cannot guarantee at high throughput. Basler and iRAYPLE both offer grabbers aligned to their line scan ranges; size the grabber and PCIe bandwidth to the camera's peak data rate, not its average.

Lighting is where line scan projects most often fail. All of the camera's pixels look at one thin line, and at high line rates the per-line exposure is extremely short, so you need an intense, narrow, highly uniform line of light precisely aligned and focused onto that sensor line. Line (bar) lights and focused-line illuminators are purpose-built for this; uniformity across the full width and stability over time directly determine defect contrast and false-reject rates. Geometry matters too - bright-field for general surface, dark-field or grazing-angle line lighting to pop scratches, embossing and topographic defects on metal, film and print.

Finally, treat optics and mechanics as part of the spec. Wide, high-resolution line scan sensors are physically long and often need large-format F-mount or M42/M72 lenses with enough image circle and resolving power to cover the sensor corner to corner without falloff. Rigid mounting, vibration control and accurate perpendicular alignment of the camera, light and web keep the single imaged line sharp and consistent across the entire width.

• GigE/PoE for moderate line rates; Camera Link or CoaXPress (with a frame grabber) for fast, wide sensors.
• Size frame grabber and PCIe bandwidth to the camera's peak line-rate data, with headroom.
• Use focused line (bar) lights with high uniformity; choose bright-field vs dark-field by defect type.
• Specify a large-format lens with sufficient image circle and MTF to cover the full sensor width.