Monochrome and colour line scan camera sensor types explained

Two line scan cameras with the same pixel count can behave like different instruments. The sensor architecture — not the headline resolution — decides how much light each line collects, whether colour is measured or interpolated, and how long the line is. That length in turn fixes the lens mount. This guide walks the seven designs you will meet: single line, dual and quad line, TDI, dual-line Bayer, trilinear, prism 3-CMOS and SWIR. Take the light budget, the colour requirement and the line length in that order, and the sensor type, the mount and the lens magnification all fall out of the arithmetic.

Which sensor types does a line scan camera come in?

Monochrome designs are single line, dual or quad line, and TDI with 16 to 128 stages; colour designs are dual-line Bayer, trilinear (optionally quadlinear with NIR) and prism 3-CMOS; InGaAs SWIR lines extend coverage to 900-1700 nm.

All the visible-band types ride silicon that responds from roughly 400 to 1000 nm, with RGB filters narrowing a colour camera's useful band to about 400-700 nm. What separates them is how many rows they expose per line and how the rows are combined:

The next sections give the numbers behind each row.

How much more light does a bigger pixel buy you?

Light collection scales with pixel area, so a 14 µm pixel gathers 196 µm² against 25 µm² for a 5 µm pixel — about 8 times the charge in the same exposure — and summing lines helps the same way: dual doubles, quad quadruples.

Twice the area means the same grey level at half the exposure or half the light, which is why the fast end of line scan favours larger pixels. When the lamp is already at full power, summing two or four rows raises signal 2× or 4× without thinning resolution along the line. That is the across-motion version of pixel binning. The small-pixel selection guide covers the resolution side of that trade.

When is a TDI camera the right answer?

Choose TDI when a single line cannot collect enough light at your line rate: a 16-stage TDI sums 16 rows for 16 times the signal, a 64-stage for 64 times, but it needs encoder-locked motion and a flat part.

TDI — time delay and integration — moves each pixel's charge through the stacked stages in step with the part, adding photons stage by stage; Wikipedia's TDI page shows the mechanism. Signal grows with stage count N while shot noise grows only as √N, so 64 stages lift signal-to-noise 8 times over one line. That is what holds a usable grey level once the line rate has climbed past what a single row can expose.

The catch is kinematics. The charge must advance in step with the part, so the line rate follows the encoder exactly — the encoder ticks per pixel guide sets that scaling up. The surface must hold one working distance, and stop-start motion defeats the design. Size the lamp too: how much power a fast line needs tells you whether the shortfall is in the light or in the sensor.

The sensitivity ladder in practice:

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

Which colour sensor design gives true colour?

Prism 3-CMOS cameras give true colour because red, green and blue light hit three separate sensors at the same instant, while trilinear rows sit 1-3 pixels apart and only a prism design survives start-stop or uneven web motion without colour fringes.

Dual-line colour is the entry tier: one row reads green, the other alternates red and blue behind a mosaic, and the camera interpolates the missing values. It is sensitive and affordable, with the same softness any Bayer-style interpolation brings.

Trilinear puts a full row on each of red, green and blue — full resolution and clean colour — at the price of row spacing. The camera delays the trailing rows by one to three line periods to rebuild one composite line. Any wobble in web speed misplaces the colours and edges fringe. The prism versus Bayer comparison quantifies that in measurement terms.

A prism camera splits the beam once and focuses it onto three sensors aligned to sub-pixel accuracy, so all three bands sample the same instant and each channel runs its own exposure and gain. Quadlinear adds a fourth, NIR-sensitive row when the job needs infrared beside colour. At the top of the colour range:

The MV-CL166-91F2C reads a 16K colour line at 5 µm in RGB/4TDI form on XoFLink, with an 82 mm image circle behind an F-mount. Where approximate colour is enough, a monochrome camera under RGB strobes rebuilds colour in three exposures — the RGB light sequencing guide walks that cheaper route.

When do you need a SWIR line scan camera?

SWIR line scan sensors use InGaAs pixels sensitive from 900 to 1700 nm and earn their place where silicon cannot separate the scene: silicon wafers turn transparent past about 1100 nm, water absorbs strongly near 1450 nm, and hot surfaces image by their own emission.

Everyday mono and colour silicon lines stop around 1000 nm, so anything needing behaviour beyond that moves to InGaAs: wafer and solar inspection, hot-process monitoring, moisture and adhesive gradients, plastics and textiles that look identical in visible light. Expect larger pixels and gentler line rates — the design goal is photons per line, not resolution.

A typical InGaAs line:

The LS-LCM-0038 reads a 1K line of 12.5 µm pixels over 900-1700 nm at 29.4 kHz on GigE.

How does line length decide the mount?

Line length is pixel count times pitch, and the mount must physically clear it: 2048 × 14 µm is a 28.7 mm line — past a C-mount — while 8192 × 7 µm reaches 57.3 mm and needs an M72 or V-mount lens with a matching image circle.

A C-mount is a 25.4 mm thread, so it clears at most a 20-22 mm line. From there M42 carries lines to about 45 mm, and longer lines move to M58, M72 or large V-mounts — exactly the 30-45 mm, 45-60 mm and 60-90 mm image-circle classes the line scan lenses range is sorted by. The lens finder walks them in order. The 57.3 mm class looks like this:

The MV-CL081-41CM needs a 57 mm image circle, which is why this 8K, 7 µm Camera Link camera rides an M72 mount. A 16K line at 5 µm measures 81.9 mm, so it needs a lens in the 60-90 mm image-circle class. The LS-LSL-0023 sits in that class: a 116 mm lens on a V-mount with a 60-90 mm image circle. Check the circle and the MTF before committing — the vignetting guide and the 4K/8K/16K lens selection guide take those checks one at a time.

What does a real selection look like in numbers?

A 120 mm colour web inspected at 0.03 mm per pixel and 1 m/s lands on a 4K, 7 µm colour camera at about 34 kHz, an M42-mount 25 mm lens at roughly 130 mm working distance, and a 29 µs exposure that decides the lighting budget.

1. Pixels across the part. 120 mm ÷ 0.03 mm/px = 4000 px, so a 4K sensor covers it with margin.
2. The camera. Colour at this width means a colour line sensor:

3. Sensor length. 4096 × 7 µm = 28.7 mm — already past C-mount, matching the camera's M42 mount and 29 mm image circle.
4. Magnification and lens. m = 28.7 mm ÷ 120 mm = 0.239. The LS-LSL-0215 matches it: a 25 mm, 0.24x lens on M42 covering a 30-45 mm image circle. Working distance then follows: WD = 25 × (1 + 0.239) / 0.239 ≈ 130 mm — re-run it at your own width with the lens focal length calculator.
5. Line rate and exposure. mm per pixel = 7 µm ÷ 0.239 = 29.3 µm. At 1 m/s: 1000 ÷ 0.0293 = 34.1 kHz, so exposure ≈ 29 µs.

6. Light budget. 29 µs is a short, hard exposure, and on a colour sensor every band still gets that same 29 µs. If grey level comes up short, add strobe power first, and let how to set the line rate keep the speed side matched; step up to a multi-row colour design like the RGB/4TDI camera above only when the lamp is at its limit.

Decision for this case: 4K colour camera, 25 mm 0.24× M42 lens at about 130 mm working distance, 34 kHz line rate, 29 µs exposure — with the lighting sized to that 29 µs before anything is ordered.

Which checks settle the choice before you order?

Six numbers settle it: pixels across the smallest defect (3 or more), mm per pixel, line length and mount, line rate with a 20-30% margin under the camera's maximum, exposure per line, and the light budget that names the sensor family.

• Pixels per defect. Keep at least 3 pixels across the smallest defect you must catch; below that, defects average into the background.
• mm per pixel and line length. Pixels × pitch gives sensor length; the mount and the lens image circle follow from it.
• Line rate. Web speed ÷ mm per pixel; leave 20-30% headroom below the rated maximum for ramps and jitter.
• Exposure and light. One line period is the whole exposure; if the lamp cannot hold the grey level in it, climb the sensitivity ladder — dual, quad, TDI, or a multi-row colour design — before buying a bigger light.
• Colour fidelity tier. Mosaic for appearance, trilinear for clean colour on steady motion, prism for measurement or uneven motion.
• Spectrum. Silicon 400-1000 nm covers most jobs; InGaAs 900-1700 nm only where the physics needs it.

Settle those six and the sensor type names itself. To turn the numbers into a parts list, start at the line scan camera finder or the configure a camera step and let the lens follow; expert support comes with every order if you want the exposure and light budget checked before you commit.